Dissecting multi‐rust resistance in wheat through genome‐wide association study, haplotype analysis, and marker validation

Abstract

Wheat is a major global staple food affected by three diseases: leaf rust (LR), stem rust (SR), and stripe rust (YR), all of which can cause substantial yield losses. Identifying genotypes with broad‐spectrum resistance to diverse pathotypes of all three rusts remains a major challenge. In this study, we examined the genomic basis of resistance to three rust diseases LR, SR, and YR in a diverse panel of 346 bread wheat (Triticum aestivum) accessions. The seedling stage phenotypic evaluation was performed for 2 years using prevalent and virulent pathotypes. Based on best linear unbiased estimators, LR and YR displayed right‐skewed distributions, whereas SR showed a bimodal pattern. Genotyping with the 35K Axiom Wheat Breeders Array, followed by quality control, yielded 11,910 high‐quality single nucleotide polymorphisms (SNPs). Population structure analysis revealed five subpopulations and a whole genome linkage disequilibrium decay of 3.49 Mb. Multi‐trait genome‐wide association studies identified 11 significant SNPs distributed on chromosomes 3A, 3B, 3D, and 7B, which were associated with 47 disease resistance genes, 22 of which were highly expressed in at least one condition. The haplotype analysis revealed eight different haplotypes, where H006 and H007 were superior in terms of multiple rust resistance (MRR). Note that 17 elite accessions, including IC427824 and HGP1‐359, were selected using multi‐trait genotype ideotype distance index analysis. Three key Kompetitive allele specific polymerase chain reaction (KASP) markers, AX94381808, AX94874313, and AX94807942 were developed and validated. This integrated genomic approach advances the identification process and can accelerate the breeding of wheat cultivars with durable MRR.

Plain Language Summary

Wheat is a globally important staple crop but is highly vulnerable to leaf rust, stem rust, and stripe rust, which together cause significant yield losses. This study investigated genetic resistance to all three rust diseases in a panel of 346 bread wheat (Triticum aestivum) accessions evaluated over 2 years. High‐density genotyping with 11,910 single nucleotide polymorphism markers identified 11 genomic regions associated with rust resistance, with chromosome 3D showing the strongest marker–trait association. Haplotype analysis revealed two groups with superior multiple rust resistance. Note that 17 elite accessions, including IC427824 and HGP1‐359, were identified as valuable sources of durable resistance. Furthermore, three KASP markers were developed and validated to enable rapid selection of resistant genotypes. These results provide genomic insights and practical tools to accelerate breeding of wheat cultivars with broad‐spectrum and durable rust resistance, contributing to global food security.

Abbreviations

AM

association mapping

BLUE

best linear unbiased estimator

CG

candidate gene

FC

fold change

GWAS

genome‐wide association study

IC

indigenous collection

IT

infection type

KASP

Kompetitive allele specific polymerase chain reaction

LD

linkage disequilibrium

LOD

logarithm of odds

LR

leaf rust

MAS

marker‐assisted selection

MGIDI

multi‐trait genotype‐ideotype distance index

MRR

multiple rust resistance

MTA

marker–trait association

NGS

next generation sequencing

PCR

polymerase chain reaction

QTL

quantitative trait locus

QTN

quantitative trait nucleotide

SNP

single nucleotide polymorphism

SP

subpopulation

SR

stem rust

YR

stripe rust

1. INTRODUCTION

Wheat (Triticum sp.) is the most extensively cultivated and commercially exchanged cereal globally (USDA, 2025). The consumption of wheat‐based food products is increasing globally because of changes in dietary habits caused by urbanization and income growth (Krishnappa et al., 2023). It has been estimated that 840 million tonne of wheat will be needed by 2050 to feed the increasing population of world. This means that an extra 80 million tonne of wheat grain will be needed in 2050, over the current 759 million tonne produced (https://www.statista.com/statistics/263977/world‐grain‐production‐by‐type/). Globally, wheat yield production has decreased from 3% during the Green Revolution era to 0.9% annually (Ray et al., 2013; Yadav et al., 2021). The loss of arable land due to urbanization and industrialization has contributed to declining wheat production trends (Malik, 2021; D. Mohan & Krishnappa, 2020;). Climate change has further increased the susceptibility of cultivars to biotic and abiotic stresses, with limitations in grain yield gains (Grigorieva et al., 2023; Pourkheirandish et al., 2020). On a global scale, wheat yield is estimated to reach 805 million tonnes in 2024/2025 with a slight increase in the previous year (FAO, 2025; USDA, 2025). In India, it is estimated that the harvested area is 31.8 Mha, production is 115–117.5 Mt, and average productivity is ∼3.56 t/ha, which constitutes the highest national output to date (FAO, 2025; Indian Space Research Organisation, 2025; Reuters, 2025; USDA, 2025). Hence, continuous efforts are needed to increase wheat production at the national and global levels.

Global wheat production is increasingly threatened by both biotic and abiotic stresses, posing serious challenges to food security worldwide. Among the major biotic threats, fungal rust diseases remain a significant concern. These include stem rust (SR), stripe rust (YR), and leaf rust (LR), which are caused by Puccinia graminis f. sp. tritici, Puccinia striiformis f. sp. tritici, and Puccinia triticina, respectively. These rust diseases can spread rapidly under favorable conditions and severely damage wheat crops. Wheat rust pathogens are difficult to manage because of their aerial mode of spread; they produce urediospores rapidly and evolve new pathotypes. Identifying and developing resistant cultivars remains a sustainable, cost‐effective, and reliable strategy (S. C. Bhardwaj et al., 2019). Wheat is resistant to rust diseases at different stages of growth. Seedling‐stage resistance (all‐stage resistance) protects wheat from rust infections from early growth through maturity and is typically governed by single, race‐specific major gene (Jan et al., 2025). In contrast, adult plant resistance (APR) is expressed after flowering, is usually nonspecific, is more durable, and involves multiple genes (Lagudah, 2011). Markers identified at the seedling stage can be valuable for breeding programs, as some of these are associated with loci that also contribute to APR. Several studies have demonstrated that a subset of seedling resistance loci is consistently detected in adult plants, and that pyramiding these with APR genes enhances overall resistance and durability (Dakouri et al., 2013; Delfan et al., 2023; Gao et al., 2016; R. Mohan et al., 2025; Vikas et al., 2022; P. Zhang et al., 2021).

LR, identified as orange–brown uredinia on leaves, is widespread and remains a major concern in wheat despite being less destructive than SR and YR (Bolton et al., 2008; Huerta‐Espino et al., 2011). Approximately 100 Lr genes and allelic variants have been reported across bread, durum, and diploid wheat, with 11 (Lr1, Lr9, Lr10, Lr13, Lr14a, Lr21, Lr22a, Lr34, Lr42, Lr58, and Lr67) cloned and characterized (Dracatos et al., 2023; Mapuranga et al., 2022). SR severely affects grain filling, causing stem breakage and shriveled grains, with potential yield losses of up to 100% in susceptible varieties (Kayim et al., 2022; Prank et al., 2019). Over 188 Sr quantitative trait loci (QTLs), 28 QTL‐rich clusters, and key loci such as Sr22, Sr24, Sr26, Sr36, Sr40, Sr57, Sr58, Sr55, and Sr56 have been mapped (Juliana et al., 2017; Tong et al., 2024), although many have lost effectiveness in parts of Asia, Europe, and North Africa due to rapid pathogen evolution.

YR, which appears as yellow–orange pustules in stripes on leaves and other aerial tissues, can cause up to 100% yield loss under severe infection (Ali et al., 2017; Chen et al., 2014; Wellings, 2011). Once restricted to cool climates, new virulent races now thrive in warmer regions (Godoy et al., 2018; Muleta et al., 2017). To date, 86 Yr genes (Yr1–Yr86) have been catalogued, with 12 cloned (Yr5, Yr7, Yr10, Yr15, Yr18, Yr27, Yr28, Yr36, Yr46, YrU1, YrNAM, and YrSP) and widely used in breeding programs (Jan et al., 2025; S. Kumar et al., 2023; Long et al., 2024; Mapuranga et al., 2022). Among these strains, Yr36 provides broad‐spectrum resistance, whereas Yr18/Lr34/Sr57/Pm38 and Lr67/Yr46/Sr55/Pm39 confer partial multi‐pathogen resistance, offering durable defense against the three wheat rusts and powdery mildew (Krattinger & Keller, 2016; C. Wang et al., 2025).

To solve the complex dynamics between wheat and rust pathogens involving both susceptibility and resistance, identification/mapping of resistance is a valuable tool. This approach facilitates the identification of both major and minor resistance genes and helps in pinpointing associated molecular markers, which are crucial for stacking multiple resistance genes in breeding programs aimed at achieving durable rust resistance (Soriano & Royo, 2015). Mapping has historically been applied to biparental mapping populations to find underlying genetic variation that co‐segregates with a trait of interest (Zhu et al., 2008). However, it has several drawbacks, such as low resolution, limited genetic diversity, and the considerable time needed to develop biparental mapping populations (P. Gupta et al., 2013; Zhu et al., 2008; Das et al., 2022;).

The advent of next generation sequencing (NGS) technologies has produced a vast amount of genome‐ and molecular‐level data for various species. In addition, the availability of low‐cost and high‐throughput single nucleotide polymorphism (SNP) genotyping, combined with ongoing advances in NGS, has greatly accelerated the number of molecular studies such as QTL mapping and genome‐wide association studies (GWASs). These developments have made it possible to identify and introduce reliable molecular markers for use in marker‐assisted breeding of wheat (Marone et al., 2024). The further discovery of QTLs in a natural population through GWAS provides greater mapping resolution by utilizing all the historical recombinations that have been gathered over generations, which makes it otherwise difficult to exploit QTL mapping for this purpose (Brachi et al., 2011; Mir et al., 2012; Kulwal & Singh, 2022). Since population structure might lead to spurious associations in GWASs, it is taken into consideration by incorporation as a covariate in the association model (Segura et al., 2012). Therefore, GWAS can be applied to any current panel that has trait variability and can be regarded as a useful alternative for biparental linkage mapping. Several GWASs have been conducted for the identification of novel genes/QTLs in germplasm lines for LRs (Kertho et al., 2015; P. Zhang et al., 2021; Vikas et al., 2022; Kaur et al., 2023; Delfan et al., 2023; Lakkakula et al., 2025; Czembor et al., 2025; Gardner et al., 2025), SRs (Kankwatsa et al., 2017; Mourad et al., 2018; Pradhan et al., 2023), and YRs (Pradhan et al., 2020; P. Zhang et al., 2021; Lin et al., 2023; Khan et al., 2024; Qiao et al., 2024; El Messoadi et al., 2024; Sharma et al., 2025; Atsbeha et al., 2025; Gardner et al., 2025). In addition, Pal et al. (2022) conducted a large‐scale meta‐QTL analysis in wheat by integrating 1146 QTLs from 152 mapping studies targeting resistance to LR, SR, and YR, and identified 86 meta‐quantitative trait loci—including 71 robust multi‐disease resistance regions. However, there is still room to discover new marker–trait associations (MTAs), genes, and haplotypes utilizing more germplasm lines, as the current repertoire of QTLs/MTAs is not saturated. Owing to the persistent threat posed by rust diseases, developing an ideal wheat cultivar requires incorporating durable resistance against all three major rust types via the use of genomic tools.

Traditional GWAS analyses for wheat rusts are typically conducted separately for each disease, identifying trait‐specific quantitative trait nucleotides (QTNs) but often overlooking the shared genetic basis of resistance across rust types. However, since LR, SR, and YR frequently co‐occur under field conditions, a multi‐trait GWAS approach offers greater power to detect pleiotropic loci that contribute to broad‐spectrum resistance. In this context, the present study employed a multi‐trait GWAS using a panel of 346 bread wheat (Triticum aestivum) accessions to dissect the genetic architecture underlying resistance to all three major rusts. By integrating phenotypic data for LR, SR, and YR into a single analytical framework, we aimed to identify shared QTNs, candidate genes (CGs), and haplotypes conferring seedling‐stage resistance across multiple rust types. We hypothesize that a multi‐trait GWAS approach would reveal common and stable loci associated with multi‐rust resistance, thereby facilitating the development of durable, broad‐spectrum resistant cultivars and supporting sustainable wheat production.

2. MATERIALS AND METHODS

2.1. Evaluation of bread wheat germplasm for multiple rust resistance

The present study was conducted at the ICAR—Indian Institute of Wheat and Barley Research (IIWBR), Regional Station, Flowerdale, Shimla, to evaluate the association mapping (AM) panel consisting of 346 bread wheat (T. aestivum) accessions against three rust diseases—LR (P. triticina), SR (P. graminis f. sp. tritici), and YR (P. striiformis f. sp. tritici) at the seedling stage. The AM panel consists of a mixture of land races (174), advanced breeding lines (30), exotic collections (50), and cultivars (92). Under controlled greenhouse conditions, screening was performed over two consecutive years (2020–2021 and 2021–2022). Predominant pathotypes of YR (110S119, 238S119, 46S119, and 47S103), LR (12‐5(29R45), 77‐5(121R63‐1), 77–9(121R60‐1) and 104‐2(21R55), and SR (11(79G31), 40A (62G29), 40–3(127G29), and 117‐6(37G19) were used for evaluation. These pathotypes were selected based on their prominence and virulence. The screening was conducted following the methodology of S. Bhardwaj (2011). The details of the accessions used in the study are given in Table S1.

Seedlings, along with the checks, were grown in spore proof glasshouse chambers using trays. At 7 days of age, fully expanded primary leaves were inoculated with a glass atomizer containing 100 mg of spores from specific pathotypes suspended in 10 mL of light‐grade mineral oil. After inoculation, the seedlings were sprayed with a fine water spray and placed in dew chambers under controlled conditions for 48 h at 20  ±  2°C for brown and black rust and 16  ±  2°C for YR, with over 90% humidity and 12 h of daylight at 15,000 lux. After 48 h, the plants were transferred to a glasshouse and maintained under the same temperature and light conditions, with a relative humidity ranging from 50% to 70%. Fine elemental sulfur was dusted on the plants to prevent powdery mildew without affecting rust development. The infection types (ITs) were scored 14–16 days post inoculation via a 0–4 scale modified from Stakman et al. (1962). IT 0 had no visible symptoms; IT 1 had minute uredia with chlorosis; IT 2 had small to medium uredia with chlorosis; IT 3 had similar uredia with possible chlorosis; and IT 4 featured large, profusely sprouting uredia, forming rings. The resistant types include IT e 0 to IT 2, whereas the susceptible types include IT 3 and IT 4. The scores were later converted to a 0–9 linear scale (Peterson et al., 1948; Riaz et al., 2016; Ziems et al., 2014).

2.2. Phenotypic data analysis

Phenotypic data were analyzed using the best linear unbiased estimator (BLUE) values to estimate resistance levels across four pathotypes for each rust disease. In our analysis, genotype and pathotype were treated as fixed effects to estimate the specific contribution of each genotype and pathotype to rust resistance, whereas the genotype × pathotype interaction and replication were considered random effects to account for variability across interactions and experimental repeats. This distinction ensures that the BLUE estimates reflect the overall resistance levels of each genotype across the four pathotypes while appropriately modeling the random sources of variation via the linear mixed effects model implemented via the R package lme4 (Bates, 2010). Furthermore, a histogram with a frequency curve was generated via the R package ggplot2 (Wickham, 2011) to determine the disease trends among the genotypes. Further genetic divergence among the 346 genotypes was assessed via Mahalanobis’ generalized distance (D ²) via the stats package (Mahalanobis, 1936). Clustering analysis was performed to group the genotypes into distinct clusters, reflecting their phenotypic similarities and differences.

2.3. SNP genotyping and quality control

Genomic DNA was extracted from 346 wheat lines from 15‐day‐old seedlings via the cetyl trimethyl ammonium bromide method (Doyle, 1990). The quality and purity of the extracted DNA were evaluated via a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific), and only high‐quality DNA samples were selected for downstream genotyping. SNP genotyping was performed via the 35K Axiom Wheat Breeders Array, following the Affymetrix Axiom 2.0 Assay protocol (P/N 703154 Rev. 2), resulting in the identification of 35,143 SNPs. To increase the robustness of the dataset, SNP markers with more than 10% missing data, call rate > 90%, and a minor allele frequency less than 10% were excluded. Consequently, a total of 11,910 high‐quality SNP markers were retained for association studies and subsequent analyses.

2.4. Population structure and linkage disequilibrium analysis

The population structure was analyzed via a model‐based clustering approach implemented in STRUCTURE v2.3.4 (Pritchard et al., 2000). The analysis was conducted with a parameter value of genetic cluster K ranging from 1 to 10, employing 100,000 burn‐in iterations followed by 200,000 Markov Chain Monte Carlo replicates. The optimal value of K was determined via Structure Selector (Evanno et al., 2005). Linkage disequilibrium (LD) was estimated as the squared allele frequency correlation (r ²) between pairwise SNP markers via TASSEL v5.0 (Bradbury et al., 2007). Genome‐wide LD decay was assessed by plotting r ² values against the physical distance between markers. To refine the estimation, LD decay was modeled via locally weighted scatterplot smoother regression (Cleveland, 1979), allowing for the determination of the threshold at which LD arises from true physical linkage. The critical r ² threshold was defined as the 95th percentile of the square root–transformed r ² values of unlinked markers (Breseghello & Sorrells, 2006).

2.5. Multi‐trait GWAS analysis

A multi‐trait GWAS was performed to identify genomic regions associated with resistance to multiple rust diseases via the R package statgenQTL×T (Zhou & Stephens, 2014). To correct the bias introduced due to multiple hypothesis testing, that is, controlling for Type I errors and mitigating false‐positive associations, the Bonferroni correction was applied. The analysis integrated genome‐wide SNP markers and phenotypic rust resistance scores to detect loci that significantly influence resistance to multiple strains of rust. To increase the robustness of SNP‐trait associations, a multi‐trait mixed model approach was employed, incorporating genetic correlations among traits to improve detection power and accuracy. The effects of significant SNPs were estimated across three rust diseases to quantify their contributions to resistance. A stringent threshold, defined by a logarithm of odds (LOD) score of ≥5.45 (p value: 3.5 × 10⁻⁶), was applied to identify significant MTAs for further downstream analyses, such as CG identification, gene expression, and haplotype analysis. For clarity, the identified MTAs were designated as QTNs and were visualized through a Manhattan plot. Single‐trait analysis was also conducted using the fixed and random model circulating probability unification and Bayesian‐information and linkage‐disequilibrium iteratively nested keyway implemented in R using GAPIT software package (Lipka et al., 2012).

Further resistance checks HD 2888 for LR and SR, FLW29 for SR, and susceptibility checks (HD 2733 for all three rust diseases) were sequenced via the same 35K array along with the wheat accessions under study. These significant QTNs were assessed for polymorphisms between the resistant and susceptible checks for all three rust diseases. This comparative analysis identifies polymorphic resistant QTNs and is used for allele screening of the remaining 346 wheat accessions to determine the number of resistant alleles present in them.

2.6. Identification, annotation, and in silico expression analysis of CGs

To identify CGs associated with the identified QTNs linked to resistance to multiple rust strains, a systematic approach was employed. SNP probe sequences that exhibited significant associations through multi‐trait GWAS were extracted along with their genomic locations and corresponding annotations. The physical coordinates of each SNP tag were used to query the Ensembl Plants database. For each QTN, the corresponding chromosomal region was extended by 300 kb upstream and downstream, resulting in a 600 kb genomic window. This region was scanned to identify CGs potentially associated with disease resistance. Protein coding information for the identified genes was retrieved via the BioMart utility available on the Ensembl Plants platform. The functional annotations of the CGs were further cross‐referenced with published literature to validate their putative roles in plant immunity and disease resistance. The identified putative CGs associated with the QTNs were further subjected to gene expression analysis via the wheat omics expression database (wheatomics.sdau.edu.cn) for LR, SR, and YR. Gene expression levels were quantified in transcripts per million. The fold change (FC) in expression was calculated under stress conditions relative to the control, and the genes with FC > 1.5 were identified as highly expressed genes. The results are presented through heatmap graphical representations via the R package ggplot2 (Wickham, 2011).

2.7. Haplotyping and identification of superior genotypes

To identify the superior haplotypes associated with the MTAs identified in the current study, we used the R package geneHapR (R. Zhang et al., 2023), a widely adopted tool for haplotype analysis and visualization. This approach has recently gained popularity and has been effectively applied in several genetic studies (Abbai et al., 2019; Sinha et al., 2020; K. Kumar et al., 2024). By using the geneHapR, we were able to define haplotype blocks on the basis of genotypic data and assess their distribution across the wheat panel. Furthermore, to explore the functional relevance of these haplotypes, haplotype–phenotype (haplo‐pheno) association analysis was conducted. This analysis enabled us to evaluate the relationships between different haplotype classes and corresponding phenotypic traits, thereby helping to identify haplotypes with favorable effects.

2.8. Multi‐trait genotype ideotype distance index

Multi‐trait genotype‐ideotype distance index (MGIDI) analysis was conducted via the metan R package (Olivoto & Lúcio, 2020), following the methodology proposed by Olivoto and Nardino (2021). A stringent selection criterion was applied, combining both phenotypic resistance scores and the number of resistant SNPs to ensure robust identification of disease‐resistant genotypes with a selection intensity of 5% to prioritize the top‐performing genotypes.

2.9. Validation by KASP markers

To validate the MTAs, we developed Kompetitive allele specific polymerase chain reaction (KASP) assays using the identified significant SNPs. First, we retrieved a 100‐base pair sequence from both sides of each significant SNP in the Chinese Spring reference genome (RefSeq v2.0). Primer design was carried out using the WASP tool (https://bioinfo.biotec.or.th/WASP/) (Wangkumhang et al., 2007) with the “no mismatch” option at the penultimate base. Allele‐specific primers were generated with the target SNP positioned at the 3′ end, and the standard FAM (5′ GAAGGTGACCAAGTTCATGCT 3′) or HEX (5′ GAAGGTCGGAGTCAACGGATT 3′) tails were added at the 5′ end (Table S2).

Polymerase chain reaction (PCR) was performed in 96‐well plates. The PCR mixture was prepared by adding 5 µL of KASPTM master mix, 1 µL of assay mixture (all provided by LGC), and 4 µL of DNA (∼5 ng/µL). PCR was performed in Real Time‐Polymerase Chain Reaction(RT‐PCR) (Biorad–CFX‐96). PCR was performed with one cycle of initial denaturation step at 94°C for 15 min, followed by 10 cycles of denaturation at 94°C for 20 s and annealing and extension step at 61°C for 1 min (drop 0.6°C per cycle), followed by 26 cycles of denaturation at 94°C for 20 s and annealing and extension at 55°C for 1 min. After completion of PCR, the plates were read via a (Biorad–CFX‐96) plate reader and the fluorescence readings were analyzed using KlusterCaller software (LGC Genomics).

The KASP assay for the significant MTAs was performed on validation set of independent population comprising of 46 new germplasm lines that has not been the part of AM panel (Table S3). This set was procured from ICAR‐National Bureau of Plant Genetic Resources, New Delhi, and categorized into resistant (22 accessions) and susceptible (24 accessions) groups. The validation set was screened under similar artificial conditions. Genotypic and phenotypic data were matched for the validation. The genotypic data from the homozygous alleles were analyzed using the Kruskal–Wallis test to identify statistically significant differences between the alleles.

3. RESULTS

3.1. Phenotyping analysis

346 wheat accessions were screened for LR, SR, and YR pathotypes under glasshouse conditions for two consecutive years (2020–2021 and 2021–2022) at the ICAR–IIWBR, Regional Station, Flowerdale, Shimla. The BLUE values provided an accurate estimation of phenotypic performance among different races of pathogens (Table S4). The LR infection scores show a right‐skewed pattern, indicating that most accessions are susceptible, with only a few showing resistance (Figure 1a). Genotypes such as IC290173 (0) (where IC is indigenous collection), IC416141 (0.25), IC427824 (0.25), CRP 165/53 (0.5), and IC290156 (0.5) were highly resistant. SR scores are more spread out and show a bimodal pattern, with many accessions showing moderate resistance and approximately 35 accessions were showing greater resistance of which a few accessions such as IC111851, IC111931, IC252415, and IC381168, show complete resistance (Figure 1b). YR scores are also right‐skewed, with most genotypes highly susceptible; however, IC336645 was highly resistant followed by CRP‐165/36 and IC393943 (Figure 1c).

FIGURE 1.

Distribution of disease infection scores across the accessions. (a) Leaf rust, (b) stem rust, and (c) stripe rust. The histogram represents the distribution of the infection scores across the accessions, with the x‐axis indicating the infection score and the y‐axis showing the frequency density.

Further clustering D ² analysis divided the 346 wheat accessions into four separate clusters according to their resistance levels against LR, SR, and YR and revealed specific distribution patterns among the four clusters for the disease infection scores (Figure 2). Furthermore, as expected within clusters, the accessions remained similar, but between clusters, differences in disease infection scores were observed (Figure 3). Cluster 1 (105 accessions) presented moderate resistance, with low scores for SR (2.233) and LR (6.730) but moderate susceptibility to YR (7.138). Cluster 2 (87 accessions) had the highest scores for all three rust types, indicating high susceptibility. Cluster 3 (44 accessions) presented the lowest scores for LR (2.480), SR (2.006), and YR (5.898), suggesting strong multi‐rust resistance. Compared with the other clusters, Cluster 4 (110 accessions) presented moderate resistance to LR (7.395) and SR (4.858) and the lowest YR score (4.469). This variation highlights the genetic diversity in rust resistance, with Cluster 3 containing the most promising accessions for breeding multi‐rust resistant lines. The details of the accessions in each cluster are provided in the Table S5.

FIGURE 2.

Dendrogram showing clustering of bread wheat accessions for multiple rust resistance.

FIGURE 3.

Genotypic variation in disease infection scores across the clusters. Boxplots illustrate the distribution of genotypic disease infection scores for leaf rust, stem rust, and stripe rust across four clusters.

3.2. SNP density, population structure, and LD analysis

A total of 11,910 polymorphic SNPs was used for GWAS analysis after filtering. The precise physical position of each SNP was mapped via IWGSC RefSeq v2.0 (https://www.wheatgenome.org/Projects/Reference‐Genome‐Project/RefSeq‐v2.1). Out of a total of 11,910 SNPs, 3679 were mapped to the A sub‐genome, 4737 on the B sub‐genome, and 3494 on the D sub‐genome (Figure 4). Furthermore, the number of SNPs on individual chromosomes varied from the highest, with 865 SNPs (2B), to the lowest, with 209 SNPs (4D and 6A). The SNP distribution at the chromosome level in the A sub‐genome revealed that the maximum number of SNPs was 2A (709), followed by 1A (681) and 7A (623). In the case of the B sub‐genome, 2B (865) had the maximum number of SNPs, followed by 1B (849) and 5B (798). The D sub‐genome presented the greatest number of SNPs in 2D (739), 1D (684), and 5D (526) (Figure 5).

FIGURE 4.

Distribution of single nucleotide polymorphisms (SNPs) across chromosomes and sub‐genome used for analysis.

FIGURE 5.

Single nucleotide polymorphism (SNP) density plot showing the distribution of SNPs on different chromosomes.

Population structure analysis grouped 346 accessions into five distinct subpopulations (SPs), SP1, SP2, SP3, SP4, and SP5, comprising 84, 14, 70, 111, and 67 accessions, respectively (Figure 6). SP4 was the largest among all five SPs, accounting for 32.1% of the total accessions, followed by SP1, SP3, SP2, and SP5. Individuals of each population were further classified into pure and admixture types based on membership proportions, with populations with ≥0.80 of their members considered pure and the rest considered admixtures. The composition of the five SPs was as follows: SP1 had 66.7% pure and 33.3% admixed individuals, SP2 had 100% admixed individuals, SP3 had 11.4% pure and 88.6% admixed individuals, SP4 had 32.4% pure and 67.6% admixed individuals, and SP5 had 22.4% pure individuals and 77.6% admixture individuals.

FIGURE 6.

Population structure plot, which shows the five major subpopulations of bread wheat accessions, and Delta K values for different numbers of subpopulations associated with population structure analysis.

A total of 11,910 markers were used to estimate LD decay values for the studied genotypic panel by plotting pairwise r ² values against physical distances. The background LD threshold was defined via LD data from the analyzed AM panel operated at r ² = 0.156, which served as the boundary value for LD estimation. Individually, the average R ² of genome‐wide LD was 0.11 for sub‐genome A, 0.14 for sub‐genome B, and 0.13 for sub‐genome D. SNP markers, with their assigned physical position on the map, were further used to estimate intrachromosomal LD. The whole‐genome LD was approximately 3.49 Mb, and sub‐genomes A, B, and D presented LD values of 2.35, 3.54, and 4.81 Mb, respectively. Within the AM panel used in our study, the A sub‐genome exhibited the most rapid LD decay relative to the B and D sub‐genomes (Figure 7). A detailed summary of the markers, including chromosome distribution, average LD score, and other associated information, is presented in Table 1.

FIGURE 7.

Linkage disequilibrium (LD) decay plot across the genomes (a) whole genome, (b) A sub‐genome, (c) B sub‐genome, and (d) D sub‐genome.

TABLE 1.

Distribution of 11,910 single nucleotide polymorphisms (SNPs) on 21 wheat chromosomes along with linkage disequilibrium (LD) and other associated statistics.

Chr	Size(Mb)	No. of polymorphic SNPs	Average number of SNPs per Mb	Chromosome LD	Number of marker pairs in perfect LD (R 2 = 1)
1A	594.10	681	0.87	0.27	6175
1B	689.85	849	0.81	0.36	8526
1D	495.45	684	0.72	0.33	6485
2A	780.8	709	1.10	0.26	6340
2B	801.26	865	0.93	0.15	5624
2D	651.85	739	0.88	0.19	5085
3A	750.84	483	1.55	0.18	6367
3B	830.83	683	1.22	0.17	6491
3D	615.55	445	1.38	0.13	5258
4A	744.59	425	1.75	0.16	5536
4B	673.62	320	2.11	0.13	5279
4D	509.86	209	2.44	0.11	2084
5A	709.77	549	1.29	0.17	6116
5B	713.15	798	0.89	0.17	6246
5D	566.08	526	1.08	0.18	5470
6A	618.08	209	2.96	0.17	3420
6B	720.99	654	1.10	0.16	6482
6D	473.59	405	1.17	0.13	5286
7A	736.71	623	1.18	0.15	5382
7B	750.62	568	1.32	0.18	5818
7D	638.69	486	1.31	0.14	5066

3.3. Multi‐trait GWAS Analysis

In the present study, a multi‐trait GWAS analysis was conducted to evaluate three traits, namely, LR, SR, and YR, using 11,910 SNPs. The marker effects were estimated via a generalized least squares approach with chromosome‐specific kinship matrices for a precise analytical approach and finer association of the traits. A total of 11 significant MTAs were identified above a threshold LOD value of 5.37665 (Figure 8) with the most significant p values ranging from 4.64 × 10⁻¹⁷ to 3.18 × 10⁻⁶. These MTAs were distributed on four different chromosomes (3A, 3B, 3D, and 7B). The details of the MTAs associated with all three rusts, along with their position, statistical significance, and other details, are presented in Table 2.

FIGURE 8.

Manhattan plot of multiple rust resistance single nucleotide polymorphisms (SNPs). The Manhattan plot displays SNP associations with rust resistance traits (leaf rust, stem rust, and stripe rust). Red colors denote significant SNPs associated with three rust types. The y‐axis represents logarithm of odds (LOD) threshold, while the x‐axis shows SNP positions across the genome.

TABLE 2.

List of significant single nucleotide polymorphisms (SNPs) for multiple rust resistance with their associated statistics.

No.	MTA	Allele	Chromosome	Position	p value	LOD	Allele frequency
1	AX94681641	G/C	3A (7)	739951129	1.78E‐12	11.75	0.344
2	AX94805563	C/T	3A (7)	742613066	2.60E‐09	8.585	0.403
3	AX94381808	C/A	3A (7)	746917385	3.12E‐06	5.506	0.171
4	AX94873612	G/A	3B (8)	810537107	2.12E‐09	8.673	0.191
5	AX94724171	A/G	3B (8)	818273099	8.55E‐14	13.068	0.108
6	AX94421161	G/A	3D (9)	2090331	1.51E‐07	6.822	0.11
7	AX94874313	A/G	3D (9)	598011345	4.50E‐17	16.347	0.09
8	AX94862539	A/G	3D (9)	606366123	1.20E‐11	10.921	0.124
9	AX94723806	T/C	3D (9)	607097496	1.09E‐09	8.962	0.145
10	AX94807942	A/G	3D (9)	607309483	5.33E‐07	6.273	0.468
11	AX94571854	G/A	7B (20)	743410903	8.73E‐07	6.059	0.301

Abbreviations: LOD, logarithm of odds; MTA, marker–trait association.

On chromosome 3D, the MTA AX94874313 displayed the strongest association (LOD = 16.35, p value of 4.50 × 10⁻¹⁷). Chromosomes 3A and 3B also carried strongly associated SNPs, including AX94681641 and AX94724171, both with LOD > 11 and highly significant p values. The SNPs associated with rust were predominantly located on chromosome 3 (10 out of 11 SNPs), with the highest number on chromosomes 3D, 3A, and 3B. On chromosome 7B, a single SNP (AX94571854) associated with rust resistance was identified.

Furthermore, the allele frequencies among the rust‐associated MTAs, varied from 0.090 to 0.468. Two SNPs, AX94874313 (allele frequency = 0.09) and AX94724171 (allele frequency = 0.108), were identified as rare variants with strong associations. On the other hand, SNP AX94807942, which had a higher allele frequency of 0.468, represents a common variant that produced a moderate LOD score with a significant p value.

To complement the multi‐trait GWAS, we also performed single‐trait analyses for each of the three rusts to identify trait‐specific MTAs and to assess whether any loci were consistently detected across both single‐trait and multi‐trait approaches. Three markers were found common between the single trait and multi‐trait GWAS analyses, namely, AX94681641 (Chr3A), AX94874313 (Chr3D), and AX94571854 (Chr7B) (Table S6). AX94681641 and AX94571854 were specifically detected for LR, and AX94874313 was detected for both LR and SR.

3.4. Identification, annotation of CGs, and in silico expression analysis

Approximately 47 CGs were found to be associated with the 11 QTNs. Genes located near the identified QTNs correspond to several functional domains relevant to plant defense. On chromosome 3A, the genes TraesCS3A03G1242500 (HMS domain) and TraesCS3A03G1243100 (MYB domain) were found in proximity to QTN AX94681641. Similarly, TraesCS3A03G1251300, which encodes an ankyrin repeat, was positioned near AX94805563. On chromosome 3B, genes encoding a winged‐helix DNA‐binding domain (TraesCS3B03G1457200) and a zinc‐finger domain (TraesCS3B03G1488700) co‐localized with AX94873612 and AX94724171, respectively. A cluster of genes on chromosome 3D containing transcription factor II A domains (TraesCS3D03G1172500, TraesCS3D03G1172000, and TraesCS3D03G1172400) was associated with AX94421161 and AX94723806. Additional nearby genes included an NADP‐binding domain gene (TraesCS3D03G1168400) associated with AX94862539 and a G3P dehydrogenase domain gene (TraesCS3D03G1169700) corresponding to AX94723806. Further annotation identified zinc‐finger (TraesCS3D03G1170800) and pentatricopeptide repeat (TraesCS3D03G1171700) genes within the intervals of these QTNs. On chromosome 7B, genes encoding LRR (TraesCS7B03G1340300), NAC (TraesCS7B03G1341000), and phosphatidylinositol‐4‐phosphate 5‐kinase (TraesCS7B03G1341500) domains were located near QTN AX94571854 (Table 3). Thus, these QTNs correlate with biologically significant CGs with domains commonly associated with plant defense and signaling. To gain a better understanding of the expression patterns of the CGs, a heatmap was generated (Figure 9).

TABLE 3.

List of candidate genes associated with the marker–trait associations (MTAs) identified in the present study.

No.	MTA	Candidate gene	Start position	Stop position	Domain	Function	Reference
1.	AX94681641	TraesCS3A03G1242500	741883382	741884296	HMS_domain	Key module in Nucleotide‐binding leucine‐rich repeat eceptors (NLRs) that detects effectors and triggers immune responses.	(X. Zhang et al., 2024)
		TraesCS3A03G1243100	741982453	741983958	Myb_domain	Involved in secondary metabolism and stress responses.	(Yu et al., 2023)
2.	AX94805563	TraesCS3A03G1251300	745013301	745015193	Ankyrin_rpt	Mediate protein interactions to regulate plant immunity and stress responses.	(Yan et al., 2002)
		TraesCS3A03G1251200	745009228	745011162	Cytochrome_P450 domain	Catalyze defense metabolite biosynthesis to enhance pathogen resistance.	(A. Wang, Ma, et al., 2022)
		TraesCS3A03G1249300	744476026	744478032
		TraesCS3A03G1249400	744498314	744500551
		TraesCS3A03G1250300	744581978	744586714	Myb_domain	Involved in secondary metabolism and stress responses.	(Yu et al., 2023)
		TraesCS3A03G1251000	744923406	744928150	LRR_domain	Recognizes pathogens and regulates activation of immune signaling.	(Belkhadir et al., 2004)
		TraesCS3A03G1249200	744465073	744471285	Kinase like domain	Phosphorylates downstream proteins to activate immunity, signaling after pathogen recognition.	(Tang et al., 2017)
		TraesCS3A03G1250200	744579318	744581582	Pentatricopeptide_ repeat	Regulate mitochondrial genes to control cell death and defense.	(Qiu et al., 2021)
		TraesCS3A03G1251100	744930582	744932263	Pectin esterase catalytic domain	Modify pectin to regulate plant defense against pathogens.	(N. Liu et al., 2018)
3.	AX94381808	TraesCS3A03G1269400	750517103	750517885	MADS box transcription factor	Regulate gene expression to enhance disease resistance.	(Ren et al., 2022)
		TraesCS3A03G1270000	750559995	750561831	SGNH Hydrolases	Modify cell walls and signals to enhance plant disease resistance.	(Anderson et al., 2022)
		TraesCS3A03G1269800	750533478	750535007	MIP_Aquaporin transporter	Enhance disease resistance by regulating water and reactive molecule transport.	(Hemmati, 2017)
4.	AX94873612	TraesCS3B03G1457600	829839727	829840907	Cytochrome_P450 domain	Catalyze defense metabolite biosynthesis to enhance pathogen resistance.	(A. Wang, Ma, et al., 2022)
		TraesCS3B03G1457300	829751051	829754255	RNA binding domain	Regulate RNA processing to control plant immune responses against pathogens.	(Woloshen et al., 2011)
		TraesCS3B03G1457700	829841951	829846698	Tify domain‐binding domain	Regulates jasmonate signaling and modulates defense gene expression via JAZ proteins.	(Dai et al., 2022)
		TraesCS3B03G1457200	829720151	829729193	Winged helix DNA‐binding domain superfamily	Facilitates DNA recognition and binding for gene regulation.	(Caruana et al., 2021)
5.	AX94724171	TraesCS3B03G1489700	839474392	839476906	Wall‐associated receptor kinase	Regulate basal defense, acting as both positive and negative disease resistance factors.	(Delteil et al., 2016)
		TraesCS3B03G1488700	839308706	839310381	Zinc finger domain	Aid structural stability and pathogen recognition in plant disease resistance.	(S. Gupta et al., 2012)
		TraesCS3B03G1487700	839172454	839174439	RNA binding domain	Regulate RNA processing to control plant immune responses against pathogens.	(Woloshen et al., 2011)
		TraesCS3B03G1487900	839191229	839192581	F‐box domain	Mediate protein degradation to regulate plant immune responses.	(H. Li et al., 2020)
6.	AX94421161	TraesCS3D03G1172500	611353749	611355192	Transcription factor II A	Regulates gene expression during plant immune responses.	(Dehury et al., 2013)
7.	AX94874313	TraesCS3D03G1136700	601342119	601343459	Plant_Fungal_Acyltransferase	Regulate disease resistance by controlling signaling and pathogen effectors.	(Ahmad et al., 2021)
		TraesCS3D03G1137000	601435168	601437131	Cytochrome b561 domain	Regulates ascorbic acid recycling to boost disease resistance.	(H. Liu et al., 2025)
8.	AX94862539	TraesCS3D03G1168000	610431088	610441089	LRR_domain	Recognizes pathogens and regulates activation of immune signaling.	(Belkhadir et al., 2004)
		TraesCS3D03G1168400	610694209	610698466	NAD(P) binding domain	Regulates redox balance and signaling for disease resistance.	(Pétriacq et al., 2013)
		TraesCS3D03G1167800	610385609	610391881	Zinc finger domain	Aid structural stability and pathogen recognition in plant disease resistance.	(S. Gupta et al., 2012)
		TraesCS3D03G1167300	610172792	610176509	Eukaryotic translation initiation factor 4E (eIF4E)	Role in the regulation of gene translation during plant disease infection.	(Song et al., 2013)
9	AX94723806	TraesCS3D03G1170800	611128693	611137614	Zinc finger domain	Aid structural stability and pathogen recognition in plant disease resistance.	(S. Gupta et al., 2012)
		TraesCS3D03G1170000	611037576	611041581	F‐box domain	Mediate protein degradation to regulate plant immune responses.	(H. Li et al., 2020)
		TraesCS3D03G1172500	611353749	611355192	Transcription factor II A	Regulates gene expression during plant immune responses.	(Dehury et al., 2013)
		TraesCS3D03G1172000	611229388	611232947
		TraesCS3D03G1172400	611335711	611337895
		TraesCS3D03G1169600	611019100	611021170	Pentatricopeptide_repeat	Regulates mitochondrial genes to control cell death and defense.	(Qiu et al., 2021)
		TraesCS3D03G1169500	611011745	611015625
		TraesCS3D03G1171700	611219792	611225087
		TraesCS3D03G1169700	611021950	611027201	Glycerol‐3‐phosphasate dehydrogenase domain	Regulates cellular redox balance and energy metabolism, influencing fungal development and virulence in plant–pathogen interactions.	(Shi et al., 2018)
		TraesCS3D03G1172200	611298723	611299220	CAP_domain	CAP proteins increases immunity by releasing defense peptides, forming protective protein complexes.	(Han et al., 2023)
		TraesCS3D03G1170100	611044654	611046759	START‐like_dom_sf	Aid lipid binding to support plant defense signaling.	(Wilson & Arunachalam, 2024)
10.	AX94807942	TraesCS3D03G1170800	611128693	611137614	Zinc finger domain	Aid structural stability and pathogen recognition in plant disease resistance.	(S. Gupta et al., 2012)
		TraesCS3D03G1171700	611219792	611225087	Pentatricopeptide_repeat	Regulate mitochondrial genes to control cell death and defense.	(Qiu et al., 2021)
		TraesCS3D03G1172000	611229388	611232947	Transcription factor II A	Regulates gene expression during plant immune responses.	(Dehury et al., 2013)
		TraesCS3D03G1172400	611335711	611337895
		TraesCS3D03G1172500	611353749	611355192
11	AX94571854	TraesCS7B03G1340300	763328872	763334281	LRR_domain	Recognizes pathogens and regulates activation of immune signaling.	(Belkhadir et al., 2004)
		TraesCS7B03G1341000	763526083	763528160	NAC_domain	Modulating defense gene expression through hormone and reactive oxygen signaling pathways.	(Dong et al., 2024)
		TraesCS7B03G1341100	763532434	763533711	F‐box domain	Mediate protein degradation to regulate plant immune responses.	(H. Li et al., 2020)
		TraesCS7B03G1341500	763574769	763580107	Phosphatidylinositol‐4‐phosphate 5‐kinase	Control membrane trafficking to enable plant disease resistance.	(Starodubtseva, 2022)
		TraesCS7B03G1341800	763601233	763603550	KH domain	Promote plant immunity by regulating ROS production.	(W. Wang, Wang, et al., 2022)
		TraesCS7B03G1341900	763675529	763681065	Zinc finger domain	Aid structural stability and pathogen recognition in plant disease resistance.	(S. Gupta et al., 2012)
		TraesCS7B03G1342000	763682182	763687540	RNA‐binding motif	Controlling RNA processing during plant disease responses.	(Woloshen et al., 2011)
		TraesCS7B03G1342500	763839549	763845028	ABC transporter‐like domain	Transporting jasmonates and antimicrobial compounds to activate defense responses.	(H. Zhang et al., 2020)

FIGURE 9.

Gene expression heatmap under leaf, stem, and stripe rust infections in wheat. *Highly significant genes (fold change [FC] > 1.5).

Among the genes linked to the identified MTAs, only 22 CGs exhibit the strongest and biologically significant transcriptional activation during rust infection and are associated with the defense response. The genes provided the strongest response to YR infection, as evidenced by exceptionally high FC values for TraesCS7B03G1342000 (2.161–2.280 FC) and TraesCS7B03G1341900 (1.876–1.236 FC). There was a distinct phase pattern of induction in LR, with TraesCS3A03G1250200 (5.723 FC at 12 hpi) and TraesCS3A03G1251300 (1.891 FC at 72 hpi) well‐timed to their early and late peaks, respectively. SR, on the other hand, induced a limited gene expression signature, and only TraesCS3B03G1489700 (1.681 FC at 2 dpi) showed significant gene expression. Notably, there was a set of D‐sub‐genome genes: TraesCS3D03G1172000 (2.633–2.912 FC), TraesCS3D03G1169500 (2.436 FC), and TraesCS3D03G1171700 (1.673 FC), which were expressed during SR only at later infection stages, and this underlined the sub‐genome‐specific specialization in the process of stress regulation. The key CG TraesCS7B03G1342000 proved to be the most potent broad‐spectrum candidate, with consistent, repeated upregulation in YR, LR, and SR, unlike the other loci. Collectively, these results suggest that wheat defense transcriptomic signaling is predominantly driven by a 7B hub SR‐responsive, followed by an isolated LR‐specific response on chromosome 3A and a single SR‐specific response on chromosome 3B, with TraesCS7B03G1342000 being the most promising multi‐rust resistance gene for further validation and implementation in breeding.

3.5. Favorable allele analysis

The sequencing data of resistant (HD 2888 for LR and SR and FLW29 for YR) and susceptible (HD 2733) strains were used for the identification of favorable alleles in the AM panel through comparative analysis. To identify favorable alleles, alleles linked to a reduction in rust response were considered favorable alleles for the respective QTN at each locus. The number of favorable alleles ranged from 0 to 11 in the AM panel. Among all the genotypes, HGP1‐359 presented the highest allele level of resistance due to the presence of 11 favorable alleles, and 19 genotypes (CRP‐165/46, CRP‐165/5, HGP1‐318, HGP1‐435, IC138426, IC252517, IC252655, IC279030, IC290192, IC290196, IC335534, IC443636, IC443704, IC524299, IC531833, IC531970, IC532037, IC547561, and J31‐170) presented 10 resistance alleles (Figure 10). The significant effects of all the pleiotropic SNPs are depicted in Figure 11, which shows that the superior alleles confer resistance, whereas the inferior alleles confer susceptibility. For each of the multiple rust resistance (MRR) disease scores, a negative correlation (p value < 0.0001) was observed (−0.44) between the number of favorable alleles in each genotype and the disease severity score. The same finding has also been confirmed by fitting simple linear regression with the number of favorable alleles as a regressor and the severity of the disease as a response (Figure 12).

FIGURE 10.

Resistant single nucleotide polymorphism (SNP) abundance among genotypes.

FIGURE 11.

Box plot showing the effects of significant pleiotropic single nucleotide polymorphisms (SNPs) on each rust types.

FIGURE 12.

Regression plot for favorable alleles, which shows regression of multiple rust resistance score with the favorable alleles. MRR, multiple rust resistance.

3.6. Haplotyping and identification of superior genotypes

To identify the superior haplotypes for stacking the MTAs associated with resistance to multiple rust pathotypes, haplotype analysis was conducted, and eight haplotypes were identified (Figure 13). Furthermore, to identify the distinct patterns of associations between specific haplotypes and phenotypic scores, a haplo‐pheno analysis was performed. The accessions harboring haplotypes H002, H005, H006, H007, and H008 presented consistently low disease scores for all three rust diseases, indicating a high level of resistance, and were the superior haplotypes. Among them, accessions IC290173 and CRP‐165/3 (H002), IC416141 (H006), IC535176 (H005), IC252655 (H008), and IC427824 (H007) were resistant to multiple rust resistant strains and can be used as donors for the introgression of MRR strains (Figure 14). Moreover, the accessions carrying haplotypes H001, H003, and H004 were IC138379, IC128283, and IC384533, which were highly susceptible, with higher disease scores for all three rusts.

FIGURE 13.

Haplotype analysis plot for identifying superior haplotypes. Rows represent distinct haplotypes (H001–H008), and columns correspond to significant single nucleotide polymorphism (SNP) positions. The frequency column indicates the occurrence of each haplotype.

FIGURE 14.

Haplotype‐based breeding for multiple rust resistance in bread wheat by stacking the superior haplotypes.

Using disease scores from all three rusts and the number of favorable alleles, the MGIDI index was applied with a 5% selection intensity to identify accessions with MRR. As a result, 17 out of 346 accessions were identified: J31‐170, IC427824, IC416141, IC290173, IC443636, IC252655, IC524299, HGP1‐318, CRP‐165/27, IC535176, CRP‐165/3, IC290156, CRP‐165/23, IC290192, IC290143, IC527448, and HGP1‐435. These accessions showed strong resistance both phenotypically and genotypically, as shown in Figure 15. The resistant accessions obtained using the MGIDI index were always characterized by low levels of rust severity, with the mean scores of 1.0–2.0 in the case of LR, 1.0–2.0 in the case of SR, and 4.0–6.0 in the case of YR, which reveal high and constant resistance to rust infection. All identified resistant accessions carried all 10 favorable resistant alleles, thereby conferring both the phenotypic and genotypic level resistance. For example, the accession HGP1‐318 recorded LR: 1.38, SR: 1.75, and YR: 4.88; the accession IC443636 recorded LR: 1.25, SR: 1.75, and YR: 4.63; and the accession J31‐170 recorded LR: 1.75, SR: 1.25, and YR: 4.25, and LR: 0.75, SR: 0.75, YR: 5.00 was significantly lower in the accession IC2526555.

FIGURE 15.

Multi‐trait genotype‐ideotype distance index (MGIDI) plot, which displays the ranking of accessions based on our selection criteria. Here, green color dot represents the selected genotypes for multiple rust resistance.

3.7. Validation of QTNs through KASP assay

A set of 11 KASP primers were designed for the significant QTNs, three of which were found to be polymorphic on the basis of the validation carried out on a panel of 46 independent wheat lines (Table S3). Three key KASP markers AX94381808, AX94874313, and AX94807942 were validated to be associated with their respective QTNs in this panel and clearly show the resistance and susceptible allele discrimination (Figure 16). These markers are located on distinct chromosomes 3A, 3D, and 3D, respectively. Further validation identified specific allelic variants of these markers that correlate with the rust resistance phenotype. Moreover, allelic variants of markers AX‐95018072, AX‐94946941, and AX‐95232570 exhibiting a statistical significance (p‐value < 0.001) emerged prominently associated with MRR (Figure 17). The identification of these markers and their strong associations with the trait suggests their potential utility in marker‐assisted selection (MAS) to improve rust resistance in wheat.

FIGURE 16.

Genotypic evaluation of the testing panel using Kompetitive Allele Specific Polymerase Chain Reaction (KASP) markers (a–c). The red and blue colors represent homozygous alleles. The black color represents non‐template control (NTC).

FIGURE 17.

Kruskal–Wallis test confirms significant multiple rust resistance differences between alleles of three KASP validated markers: AX‐94946941, AX‐95018072, and AX‐95232570.

4. DISCUSSION

Bread wheat (T. aestivum) plays a key role in the world by providing over one‐fifth of calorie and protein rich food to people (Goel et al., 2021). Wheat is a globally significant and staple crop, its yield is frequently hindered by various biotic stresses, and the need for disease resistance is a crucial target for sustainable agriculture (Karri & Nalluri, 2024). Among all the biotic stresses in wheat, the risk of rust diseases LR, SR, and YR caused by P. triticina, P. graminis f. sp. tritici, and P. striiformis f. sp. tritici, respectively, is high, and their susceptibility level increases (Rehman et al., 2024). These diseases can cause yield losses of more than 70% in susceptible wheat under favorable conditions, degrading both quality and yield and reducing market value and farmer profits (Kashyap et al., 2025). Large‐scale fungicide use creates sustainability issues because it can harm the environment, cause fungal pathogens to develop resistance, and is expensive. Currently, genetic resistance is crucial, and both R genes and QTLs must be deployed for long‐term protection (Abdul et al., 2024). Vertical resistance is race specific in nature and horizontal resistance is generally more durable for effective protection of diverse pathogen races. Achieving durable resistance requires the strategic deployment of desired resistance allele combinations that respond to biotic stresses. Limited germplasm diversity is needed for the evaluation of landraces and gene bank accessions for crop improvement, with GWAS and haplotype‐based breeding (HBB) technologies facilitating the rapid development of resistant cultivars (Ayiecho & Nyabundi, 2025).

4.1. Phenotypic screening and cluster analysis

To develop durable resistance in crops, we focused on identifying wheat genotypes that can provide resistance to all three rust diseases. Additionally, at the global and national levels, various efforts have been made to screen wheat germplasm accessions from gene banks for disease resistance (Gurung et al., 2014; Sehgal et al., 2015; S. Kumar et al., 2016). Additionally, owing to new genotyping technologies, these germplasm lines can be clearly identified and used in efficient breeding strategies (Burridge et al., 2024; Chachar et al., 2024). Under controlled glasshouse conditions, we studied 346 wheat accessions for resistance to LR, SR, and YR, which is challenging in MRR breeding programs (Azizinia et al., 2020; Bhavani et al., 2021). According to the LR scoring data, most of the accessions were susceptible, so the distribution was right skewed, and only a few accessions, namely, IC290173 and IC416141 were resistant. The bimodal distribution of SR disease score data across the accessions clearly provides two distinctions: moderately resistant and highly susceptible accessions rather than highly resistant accessions (Gessese, 2019; Zenebe, 2022). YR displays a pattern similar to that of LR and is characterized by moderately susceptible and highly susceptible accessions with only one highly resistant accession, IC336645 (M. Wang & Chen, 2017; Srinivas et al., 2024). National breeding priorities continue to prioritize the exploration and exploitation of these sources for developing multiple rust‐resistant genotypes. Wheat improvement initiatives that benefit from the use of phenotypic and molecular characterization techniques should focus on a variety of long‐lasting rust resistance targets.

D ² clustering analysis effectively differentiated the wheat accessions by providing genetic variability for the differences in resistance against LR, SR, and YR (Yasin et al., 2024). Clustering analysis revealed clear genetic variation in response to multiple rust diseases. The accessions in Cluster 3 presented broad‐spectrum resistance to all three diseases, making them valuable resources for resistance breeding.

4.2. Marker coverage and population structure

Genetic diversity is important for successful breeding programs. To use plant genetic resources effectively, understanding their extent of genetic diversity and population structure are essential steps (Atwell et al., 2010). We determined the marker density of 11,910 selected high‐quality SNPs on different chromosomes and reported that the B sub‐genome (39.77%) presented the highest marker density as compared to A and D sub‐genomes (D. Kumar et al., 2020).

Population structure is a critical factor that significantly influences LD (Flint‐Garcia et al., 2003). The AM panel used in the study includes the accessions from different regions across the globe and India. Among the 346 accessions, 51 accessions were from different parts of the world, such as Mexico, Nepal, and Australia. The major part of the AM panel (59.2%) is from North India (Delhi, Himachal Pradesh, Uttarakhand, Jammu and Kashmir, Punjab, Haryana, Uttar Pradesh, and Bihar); 58 accessions were collected from parts of western and central parts of India, such as Rajasthan, Gujarat, Maharashtra, and Madhya Pradesh; 16 accessions were from South India (Tamil Nadu, Karnataka, and Andhra Pradesh); and the remaining 16 accessions were from Jharkhand, Assam, West Bengal, and Sikkim.

The population structure determined by the STRUCTURE program suggested five SPs. A considerable degree of admixture can be attributed to the sharing and exchange of germplasms between different breeding programs (D. Kumar et al., 2020). The observed whole‐genome LD decay of 3.49 Mb suggests moderate LD across the wheat panel and the possibility of recombination events (Service et al., 2006). We observed more rapid LD decay in the A sub‐genome than in the B and D sub‐genomes, which aligns with findings from previous studies (Chao et al., 2010; Voss‐Fels et al., 2015).

4.3. Multi‐trait GWAS

The method of multi‐trait GWAS with StatgenQTLxT enhances the ability to identify MTAs via trait correlations to facilitate the detection of pleiotropic loci with enhanced accuracy and very few false positives, and the package can handle large amounts of data with mixed models (Zhou & Stephens, 2014). Note that 11 MTAs with an LOD threshold of 5.37 were identified using a multi‐trait mixed model that effectively captures pleiotropic loci and reveals novel associations (Korte et al., 2012). In our current study seven MTAs were found to be novel and remaining four MTAs were consistent with previously conducted GWASs in wheat for rust resistance (Table S7). Chromosomes 3A, 3B, and 3D have been repeatedly implicated in conferring MRR in wheat. Among them, chromosome 3A harbors several confirmed QTLs contributing to rust resistance and has demonstrated consistent breeding value (Yang et al., 2020). Chromosome 3B, particularly the short arm (3BS), carries the well‐known gene Yr30, which is closely linked with Sr2 and confers durable APR to both YR and SR. Similarly, chromosome 3D contains multiple QTLs such as QRYr3D.1 and QRYr3D.2, which provide resistance at both the seedling and adult plant stages (Rosewarne et al., 2013). Additional genes associated with rust resistance have also been identified on chromosome 3D (Bariana et al., 2016; Huang et al., 2021; Khan et al., 2024; Q. Li et al., 2011). In the present study, the MTA AX‐94681641 identified at 739.95 Mb on chromosome 3A corresponds closely with previously reported loci for rust resistance. Notably, Pradhan et al. (2023) reported a marker at the same position associated with seedling stage SR resistance. Gardner et al. (2025) also reported a significant MTA (MT25Mb_5) for LR resistance located between 736.3 and 741.24 Mb, encompassing the region identified in our study. Furthermore, Jambuthenne et al. (2022) also mapped QNV.Yr3A.5 at 738.90 Mb for YR resistance, which is in close proximity to our identified MTA, suggesting that this genomic region on chromosome 3A may harbor a conserved locus conferring broad‐spectrum rust resistance. MTA_AX94805563 identified at 742.13Mb found in our study was also very close to QLr_3AL found at 742.80 Mb for LR resistance by P. Zhang et al. (2021), suggesting that both may represent the same genomic region contributing to rust resistance on chromosome 3A. Similarly, the MTA_AX‐94873612 identified on chromosome 3B at 810.53 Mb in the present study aligns closely with the locus BR_3B826, reported between 810.33 and 826.1 Mb on the same chromosome for LR resistance by Gardner et al. (2025), indicating consistency across studies for this genomic region. Additionally, Khan et al. (2024) reported the same MTA (MTA_AX‐94723806) at 607.1 Mb for SR and YR resistance, which corresponds precisely to the position identified in the present study (607.09 Mb). The close concordance in physical position suggests that both studies likely detected the same allele, potentially representing the same underlying gene conferring rust resistance in this genomic region.

Notably, chromosomes 3A, 3B, 3D, and 7B emerged as key genomic regions, which is consistent with prior studies where Vikas et al. (2022) detected 51 QTNs for LR and Pradhan et al. (2023) reported 14 QTNs for SR seedling resistance (Figure 18). However, unlike these single‐trait analyses, the current study captures shared genetic control across multiple rusts, highlighting pleiotropic loci on 7B (MTA_AX94571854) at 743.41Mb—a major resistance hotspot encompassing meta‐QTLs such as MDQ7B.1 at 624.58 Mb reported to confer resistance to all three rust types (Eltaher et al., 2021; Tong et al., 2024). The discovery of pleiotropic and stable loci through this integrated framework not only reinforces the genetic basis of multi‐rust resistance but also provides high‐confidence candidate regions for deployment in marker‐assisted and genomic selection pipelines aimed at accelerating durable rust resistance breeding in wheat.

FIGURE 18.

Physical map of rust resistance loci identified on chromosome 3. Here, PI represents the previously identified quantitative trait loci (QTLs) for rusts and multiple rust resistance (MRR) represents the marker–trait association (MTAs) identified in the current study for multi‐rust.

4.4. Resistant SNPs, allele effects, and regression in wheat for MRR

Favorable allele analysis (resistant SNPs) and regression analysis revealed that rust resistance in wheat is influenced by multiple SNPs with additive effects. As more resistant SNPs accumulate, their resistance level in wheat gradually increases, thereby leading to an increased allele level of resistance (Daetwyler et al., 2014; Merrick et al., 2021; Shook, 1989; Yuan et al., 2002). A large distribution pattern of resistant SNPs exists across wheat accessions from two alleles in IC260326 to 11 alleles in HGP1‐359, which demonstrates significant diversity in allelic resistance combinations. Accessions that have nine or more resistant SNPs displayed better resistance to all three rust strains, thus validating the use of resistance‐specific SNP pyramids in the development of durable defense lines for wheat (Babu et al., 2020; R. Liu et al., 2020). The regression analysis revealed a significant negative correlation (r = −0.44) between the disease severity score and the number of favorable alleles, with an R ² score of 0.20 with a similar trend observed in our previous study by Pradhan et al. (2020). The data show that favorable SNP allele counts contribute to 20% of the phenotypic variation, which represents a significant effect on quantitative traits among complex traits. Additionally, studies have reported that the combined influence of many small‐effect QTNs leads to an overall improvement in MRR under pathogen pressure (Adhikari et al., 2023).

4.5. CG identification and expression analysis

The identified MRR QTNs are linked to the CGs that are involved in various functional roles in disease resistance in plants. These QTNs co‐localize with the genes encoding proteins with various domains, including the LRR domain (Belkhadir et al., 2004), HMS domain (X. Zhang et al., 2024), ankyrin repeats (Yan et al., 2002), and zinc finger domains (S. Gupta et al., 2012), and these domains facilitate the detection of pathogens and transduction of signals that are required for plant resistance and immune pathway regulation. Key transcriptional regulators, namely, Myb (Yu et al., 2023), NAC (Dong et al., 2024), and MADS‐box proteins (Ren et al., 2022) were also identified among the CGs linked with the QTNs. These factors are known to orchestrate the defense gene expression and modulate stress responsive pathways. Additionally, genes involved in redox regulation and primary metabolism such as NAD(P)‐binding (Pétriacq et al., 2013), G3P dehydrogenase (Shi et al., 2018), and parts of the cytochrome P450 domain catalyze the defense metabolite biosynthesis to increase pathogen resistance in plants. Overall, the co‐localization of QTNs co‐localized with these biologically significant domains highlight their potential for use in breeding programs. These potential CGs could be targeted and used to pyramid broad spectrum durable elite wheat varieties that enhance resistance to various rusts.

4.6. Identification of superior genotypes through haplotyping and MGIDI

The future of precise genomic selection and stacking of alleles can be simplified through HBB, which converts SNPs into coherent haplotype blocks. This method speeds up the breeding process by filtering desirable allele combinations with minimum linkage drag. This will increase the effectiveness of marker‐assisted and genomic selection and improve crop improvement toward global food security (Sivabharathi et al., 2024). The present study demonstrated the potential of HBB to increase MRR in bread wheat through strategic selection and stacking of superior haplotypes. This study highlights the value of HBB for improving the MRR of bread wheat. Superior haplotypes, such as H002 (IC290173, CRP‐165/3), H005 (IC535176), H006 (IC416141), H007 (IC427824), and H008 (IC252655), contribute to MRR, and by stacking these superior haplotypes, we can pyramid favorable alleles to develop genotypes with durable resistance.

MGIDI is a versatile and powerful tool for selecting superior genotypes across multiple traits in various crops. This multivariate selection index integrates data from several traits into a single composite value, ranking genotypes according to their distance from an ideotype, or ideal genotype (Olivoto & Nardino, 2021). The method groups traits into components through PCA and facilitates the identification of genotypes with optimal overall performance across multiple traits. Using MGIDI index, genotypes with stable and broad‐spectrum rust resistance were successfully identified. The 17 selected accessions demonstrated their strong phenotypic performance by consistently displaying low rust severity. Interestingly, all 10 favorable alleles were present in all of these accessions, indicating a strong correlation between resistance expression and allele accumulation.

4.7. KASP SNP markers for MRR

The validation of the identified QTNs is a critical step in confirming their usefulness for MAS. In the present study, three major SNPs—AX94381808, AX94874313, and AX94807942 were successfully validated using KASP markers in an independent panel comprising multiple rust‐resistant and susceptible wheat lines. The effectiveness of KASP markers for QTN validation and MAS in wheat has been well‐documented. For instance, Rasheed et al. (2016) developed and validated KASP markers for key genes controlling grain yield, quality, and disease resistance traits. Similarly, S. Kumar et al. (2022) and Pradhan et al. (2023) applied KASP markers to validate QTLs for diverse traits in wheat, underscoring their potential for accelerating marker‐assisted breeding (S. Kumar et al., 2022; Pradhan et al., 2023). KASP validation conducted in our study on a panel of 46 representative genotypes demonstrated consistent and reliable MTAs. Despite the moderate panel size, the inclusion of genetically and phenotypically diverse accessions ensured meaningful validation of the identified QTNs. The results highlight the robustness of the associations, and future validation across larger and independent germplasm sets will further reinforce their applicability in MAS and breeding for multi‐rust resistance.

5. CONCLUSION

This study highlights the effectiveness of integrating multi‐trait phenotyping and GWAS to determine the genetic architecture of MRR in wheat. The results of the present study highlighted the importance of tapping the genetic resources that have been conserved in the gene bank. The accessions identified in our study are valuable sources for the breeding of multiple rust resistant strains and can be used for the breeding of resistant cultivars. The MTAs and haplotypes reported in this study serve as valuable genomic resources that can be exploited for trait introgression through MAS and HBB. Future efforts should validate these haplotypes across environments and integrate them into high‐yielding lines for comprehensive crop improvement. Overall, these results contribute significantly to the development of cultivars with durable and broad‐spectrum rust resistance.

AUTHOR CONTRIBUTIONS

Thamaraikannan Sivakumar: Conceptualization; data curation; formal analysis; writing—original draft; writing—review and editing. Divya Sharma: Formal analysis; investigation; methodology; writing—original draft; writing—review and editing. V. K. Vikas: Investigation; writing—review and editing. Neeraj Budhlakoti: Conceptualization; software; writing—review and editing. O. P. Gangwar: Investigation; writing—review and editing. Pramod Prasad: Investigation; writing—review and editing. Ankita Mohapatra: Investigation; methodology. Sathishkumar R: Investigation; methodology; validation. Deepak Singh Bisht: Validation; writing—review and editing. Priyanka Jain: Validation; writing—review and editing. Ritu Sharma: Software; writing—review and editing. Bonipas Antony John: Methodology; software. Reyazul Rouf Mir: Investigation; writing—review and editing. Farkhandah Jan: Writing—review and editing. Dwijesh C. Mishra: Software; writing—review and editing. Satinder Kaur: Investigation; writing—review and editing. Amit Kumar Singh: Investigation. G. P. Singh: Investigation; writing—review and editing. Sundeep Kumar: Conceptualization; funding acquisition; investigation; project administration; resources; supervision; writing—review and editing.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Supplementary Table 1

Supplementary Table 2

Supplementary Table 3

Supplementary Table 4

Supplementary Table 5

Supplementary Table 6

Supplementary Table 7

Sivakumar, T. , Sharma, D. , Vikas, V. K. , Budhlakoti, N. , Gangwar, O. P. , Prasad, P. , Mohapatra, A. , R, S. , Bisht, D. S. , Jain, P. , Sharma, R. , John, B. A. , Mir, R. R. , Jan, F. , Mishra, D. C. , Kaur, S. , Singh, A. K. , Singh, G. P. , & Kumar, S. (2026). Dissecting multi‐rust resistance in wheat through genome‐wide association study, haplotype analysis, and marker validation. The Plant Genome, 19, e70188. 10.1002/tpg2.70188