A total of 178 bread wheat germplasms (including 164 recombinant inbred lines received from the International Maize and Wheat Improvement Center (CIMMYT-Mexico) and 13 commercial cultivars cultivated in Ethiopia) were used in the current study ( Supplementary Table 1 ). The germplasms from CIMMYT were comprised of six genotype lines from the National Variety Trial, five from the Adaptation Trial, 34 from the High Rain Wheat Screening Nursery, 49 from the International Bread Wheat Screening Nursery, 53 from the International Septoria Observation Nursery, 14 from the High Rain Wheat Yield Trial, and the remaining three genotypes were from PVT (preliminary variety trial). The bread wheat germplasms were evaluated across three locations ( Table 1 ), namely, at Kulumsa Agricultural Research Centre (KARC) (8° 02’ N/39° 15 ‘ E), Meraro (substation of KARC) (07° 24’ 27 N/39° 14’ E), and Holeta Agricultural Research Centre (HARC) (9° 3’ N/38° 30’ E) in the 2020/21 main cropping season. The wheat genotype King-bird (G40) was used as the standard check in the field evaluation. The experiment was set up using an alpha-lattice design with two replications, six sub-blocks, and 30 entries per sub-block per replication, with two rows per entry ( Supplementary Table 2 ). Each accession was sowed manually at 1m length, 0.20 m spacing between rows, and 0.4 m between entries. Spaces between blocks and replications were maintained at a 1.50 m distance. All trials were seeded with 150 kg of seeds per hectare, fertilized with 100 and 75 kg of N and P2O5 each, and weeded three times by hand-picking. To provide enough disease pressure, the sensitive cultivar “Morocco” was sown as a diffuser row throughout the length of the blocks.

The modified Cobb scale (Peterson et al., 1948) was used to measure the disease severity (DS), with values ranging from 0% to 90%. Moreover, the genotypes’ field response (FR) to stripe rust infection was scored following the procedure described by Roelfs et al. (1992) as Immune = no uredinia or other macroscopic sign of infection; R = resistant, small uredinia surrounded by necrosis; MR = moderately resistant, medium to large uredinia surrounded by necrosis; MRMS=moderately resistant to moderately susceptible; MS = moderately susceptible, medium to large uredinia surrounded by chlorosis; or S = susceptible, large uredinia without necrosis or chlorosis. For statistical analysis, each replicate’s genotype responses for immune, resistant, moderately resistant, moderately resistant to moderately susceptible, moderately susceptible, and susceptible were translated into the values 0, 0.2, 0.4, 0.6, 0.8, and 1, respectively (Elbasyoni et al., 2019). When the disease severity was greater than or equal to 50% and a susceptible reaction (S) was seen on the spreader rows, the plant resistance response was scored, and this resulted in a combined value of 50S (Alemu et al., 2021). In addition to the disease data, important agronomic traits such as days to 50% heading (HD), days to 50% flowering (FD), days to maturity (MD), thousand kernel weight (TKW), grain yield per plot (GYPP), plant height (PH), spike length (SL), flag leaf area (LA), number of spikelets per spike (NSs/S), number of kernels per spike (NK/S), number of kernels per spike (NK/S), and spike weight (SW) were recorded. Days to heading, flowering, and maturity were recorded for whole plots when 50% of the plants reached the corresponding Zadock’s growth stages. Plant height and spike length were measured at the physiological maturity stage in five randomly selected and tagged plants from the middle rows of each entry. Data on the number of spikelets per spike, the number of kernels per spikelet, and the total number of kernels per spike were determined from the five randomly tagged plants per entry per replication by collecting their spikes separately. These yield and yield-related data were taken from the two rows of each plot and converted to kilograms per hectare (kg ha⁻¹) at 12.5% moisture content using plot size as a factor.

The panel was divided into resistant, intermediate, and susceptible groups based on DS, FR, and CI data that were pooled from all growth stages and environments (Liu et al., 2017). The panel showed variable reaction groups for DS, FR, and CI. For DS, 65.73% of the genotypes were resistant (0≤DS ≤ 10) at Hol (Holeta), 10.11% at Kul (Kulumsa), and 1.69% at Mer (Meraro) ( Figure 1 ) with regard to FR, 96.6% of the genotypes joined the resistant group at Holeta, 50.56% at Kulumsa, and 6.74% at Meraro ( Figure 2 ). Likewise, for CI, 60.11%, 3.37%, and 1.12% tested of the wheat genotypes were resistant at the Holetta, Kulumsa, and Meraro locations, respectively ( Figure 3 ). Since CI is a combined representation of DS and FR (Liu et al., 2017), clustering based on this parameter resulted in 5 stable resistant genotypes (2.81%), 119 intermediate genotypes (66.85%), and 54 susceptible genotypes (32.02%) out of the 178 tested genotypes ( Figure 3 ).

DArTSeq genotyping of the 178 bread wheat germplasms resulted in a total of 35,672 SNPs ( Figure 5 ). Among the three wheat sub-genomes (A, B, and D), 10,317, 10,979, and 9,756 SNPs were distributed, respectively ( Figure 5 ). Among the 21 wheat chromosomes, chromosome 4D contained the lowest (833) number of SNPs, while the highest (2065) number of SNPs was found on chromosome 7D. After filtering with call rates > 70% and minor allele frequencies > 0.05, a total of 6,788 SNP markers were obtained that were distributed as 2,410 SNPs on the A sub-genome and 2,872 on the B sub-genome, and the filtered 6,788 SNPs were used in successive analyses.

Three sub-populations were inferred based on the output of the STRUCTURE software ( Figure 6A ). The three clusters ( Figure 6B ) showed a significant degree of genetic mixing, indicating that the studied wheat populations are closely related. All of the individual genotypes shared genes inherited from all three subpopulations. The scatter plot ( Figure 7A ) and 3D plot of the first three principal components ( Figure 7B ) also confirmed the existence of three clusters in the association with greater admixture, where the first two PCs (PC1 and PC2) coordinates explained the majority of the variation in the association panel. The presence of cryptic familiar relatedness was also verified by kinship analysis ( Figure 7C ), demonstrating the significance of including both population structure (Q) and kinship (K) as covariates in marker-trait association analyses.

For the 6,788 SNP markers across 178 genotypes, LD was calculated. LDs’ alleles differ between chromosomes and within sub-genomes ( Table 6 ). The scatter plot of the genome-wised pairwise LD decay plot is shown in Figure 8 . Overall, 97,723 (27.61%) of the 338,125 marker pairings with average LD values of r² = 0.11 showed significant p ≤ 0.01 LD. The B sub-genome had the highest number of (143,600, or 42.47%) marker pairs, while the D sub-genome had the fewest (75,300, or 22.27%) marker pairs. Comparatively, SNPs on the B sub-genome showed the greatest LD with a mean value of r² = 0.1187. The LD between SNPs declined across all chromosomes at the LD cut-off r² = 0.2 within a physical distance of 31.44 Mbp. The marker pairs on chromosome 4D and 2D showed the poorest (r² = 0.03, and strongest r² = 0.21) ( Table 6 .

Linkage-disequilibrium and Bayesian information were used in the MTA analysis. At a nominal p-value of 0.001, or –log10 (0.001) =3 the BLINK statistical model-based association analysis discovered 148 SNPs significantly associated with the coefficient of infection. Supplementary Table 4 reports MTAs that exceeded the nominal p-value of 0.001 or -log10 (0.001) = 3 significance thresholds for yellow rust resistance at pre-heading, heading, flowering, mid-maturity, and maturity growth stages. In addition, allele identity, marker position, p-values, additive effects, and r² of the identified MTAs were computed. Among 148 identified MTAs, 17 (11.49%) MTAs conferred yellow rust resistance at pre-heading, 24 (16.22%) at heading, 34 (24.32%) at flowering, 21 (18.24%) at mid-maturity, 25 (16.89%) at the level of maturity, and 28 (18.92%) MTAs for all stages were effective. The percentage of phenotypic variation explained by the markers varied significantly from 1.3% for yellow rust resistance measured at heading at Holeta to 14.14% for the resistance measured in the pre-heading stage at Meraro.

At the pre-heading stage, the proportion of phenotypic variation (r²) explained by the detected significant markers extended from 2.19% for the SNP 1214020|F|0-68:A>C-68:A>C on chromosomes 1D to 14.14% for the SNP 5969589|F|0-49:C>G-49:C>G on 1D. Likewise, the proportion of phenotypic variation (r²) explained by the significant SNP markers at the heading stage ranged from 1.3% for 3020463|F|0-7:A>G-7:A>G and 1000215|F|0-22:G>T-22:G>T A on chromosomes 1A to 11.09% for 4910323|F|0-27:C>A-27:C>A on 1D. Similarly, the r² for MTAs for yellow rust resistance at flowering ranged from 9.74 - 14.05% for the allele 1062638|F|0-63:G>A-63:G>A and 1000215|F|0-22:G>T-22:G>T A on chromosomes 1A and 2A respectively. The r² of MTAs for yellow rust resistance at mid-maturity ranged from 8.71 - 13.63%, for the SNP markers 1115117|F|0-21:T>C-21:T>C and 1094246|F|0-48:A>G-48:A>G on chromosomes 1A and 2B respectively. Moreover, the phenotypic variance explained by the MTAs at maturity for yellow rust resistance ranged from 7.1% for the SNP 1245237|F|0-24:G>A-24:G>A on chromosomes 2A to 14.05% for the SNP marker 2264044|F|0-68:T>C-68:T>C on chromosomes 3B. Furthermore, the proportion of yellow rust resistance explained by the significant alleles at the maturity stage ranged from 7.94% for the allele whereas the allele 4008773|F|0-17:G>A-17:G>A on chromosomes 2B to 13.63% for the SNP 1094246|F|0-48:A>G-48:A>G on chromosomes ( Supplementary Table 4 ).

A genome-wide scan for the coefficient of infection in individual environments identified significant markers associated with yellow rust resistance at different locations. For instance, MTA analysis at Holeta identified six MTAs significantly associated with yellow rust resistance at the pre-heading stage on chromosomes 1D, 2B, 3A, 5A, 3D, 4A, and 6B; 12 MTAs at the heading stage on chromosomes 1D, 2B, 3B, 6B, 7A, and 7D; seven significant MTAs conferring coefficient of infection at the flowering stage on chromosomes 2A, 2B, 3A, 5A, 7A, and 7D; nine MTAs at the mid-maturity stage on chromosomes 2B, 3A, 5A, 7A, and 7D; eight significant MTAs at the maturity stage on chromosomes 2B, 3A, and 6A; and seven MTAs for yellow rust resistance across all stages on chromosomes 2B, 3A, 5A, 7A, and 7D. The analysis revealed that MTAs on chromosome 2B conferred stable yellow rust resistance across all growth stages at Holeta. Likewise, a GWA scan for yellow rust resistance at Kulumsa identified 30 MTAs conferring resistance to the coefficient of infection at various growth stages: two MTAs at the pre-heading stage on chromosome 1B; seven MTAs at the heading stage on chromosomes 1D, 4D, 6B, and 7A; four MTAs at the flowering stage on chromosomes 1A, 1B, 1D and 7A; six MTAs at the mid-maturity stage on chromosomes 1A, 1B, 1D, 6B and 7A; five MTAs at the maturity stage on chromosomes 1D, 1B and 7A; and six MTAs for yellow rust resistance across all stages on chromosomes 1A, 1B, 1D, 6B and 7A. Moreover, a GWA scan for the yellow rust resistance analysis at Meraro discovered numerous (46) MTAs that confer resistance to yellow rust at various growth stages. The analysis identified four MTAs at the pre-heading stage on chromosomes 1A, 1D, 2B, and 5B; three MTAs at the heading stage on chromosomes 2B, 3A, and 7D; 15 MTAs at the flowering stage on chromosomes 1A, 2A, 2B, 3B, 4B, 5A, 5D, 6B, and 7A; five MTAs at the mid-maturity stage on chromosomes 2B, 3A, 4B and 6B; eight MTAs at the maturity stage on chromosomes 2D, 3A, 4A, 4B, 5A, 6B and 7A; and 11 MTAs for yellow rust resistance at all stages of growth on chromosomes 1A, 1B, 2D, 3A, 3B, 3D,4B, and 5A. The combined measure of yellow rust resistance across locations at the heading, mid-maturity, and maturity stages provided significant associations. It found 20 MTAs, including one at mid-maturity on chromosomes 1A, two at the maturity stage on chromosomes 1A and 1B, and two for yellow rust resistance across locations and all stages on chromosomes 3A and 3B ( Supplementary Table 4 ; Figure 9 ; Supplementary Figure 1 ). Eight MTAs were found at the flowering stage on chromosomes 1A, 2D, 3D, 5A, and 7A; one MTA at the mid-maturity stage on chromosomes 1A; two MTAs at the maturity stage on chromosomes 1A and 1B; and two MTAs for yellow rust resistance across locations and all stages on chromosomes 3A and 3B. It was observed that MTAs on chromosome 7A were detected to contribute to stable yellow rust resistance across all environments. The allele contributes to yellow rust resistance at the flowering stage across all environments. Moreover, the allele on chromosome 2B is also found to confer yellow rust resistance at Holeta and Meraro.

The coefficient of infection putative QTLs were discovered by combining the MTAs based on their physical distance in Mbp. MTAs on the same linkage group within the physical distance for LD decay specific to that chromosome were assigned to the same putative QTL ( Supplementary Table 5 ; Figure 10 ; Supplementary Figure 2 ). Accordingly, the 148 MTA markers were assigned to 54 putative coefficients of infection QTLs based on our LD criteria. The association analysis for yellow rust resistance at pre-heading in individual environments found 12 putative QTLs that were on chromosomes 1A (qYrS.02), 1B (qYrS.06 and qYrS.08), 1D (qYrS.09, qYrS.10, and qYrS.11), 2A (qYrS.12), 2B (qYrS.17 and qYrS18). It was observed that some of the detected putative QTLs were effective for the coefficient of infection at several growth stages while others contributed to the coefficient of infection at a particular growth stage (eg., qYrS.56 contributed to yellow rust resistance at the maturity stage). For instance, qYrS.10 contributed to a stable coefficient of infection at the pre-heading, heading, mid-maturity, and maturity stages. Similarly, qYrS.28 was found to be effective for yellow rust resistance at the heading, flowering, mid-maturity, and maturity stages. qYrS.02 also contributed to the coefficient of infection at the pre-heading, mid-maturity, and maturity stages. qYrS.54 was effective for yellow rust resistance at the heading, flowering, and maturity stages. Likewise, qYrS.13 contributed to resistance in later stages such as the mid-maturity, and maturity stages. Putative QTLs that contributed to stable yellow rust resistance from the flowering to maturity stages include qYrS.20, qYrS.27, qYrS.28, and qYrS.43. Supplementary Table 5 presents more information on the detected putative QTLs.

By annotating genes detected in the QTL regions using the recently released IWGSC RefSeq Annotation v2.1, the functional relationship between the identified QTLs and yellow rust resistance was further studied. Annotation identified several resistance-associated genes that are involved in the plant defense system ( Supplementary Table 7 ). Some of the identified high-confidence candidate genes involved in defense response to fungus include TraesCS1A02G325400 on chromosome 1A, TraesCS1B02G067000 on 1B, TraesCS1D02G018500 on 1D, TraesCS2B02G548900 on 2B, TraesCS2D02G497400 on 2D, TraesCS3A02G395300, on 3A, TraesCS3B02G041900 on 3B, TraesCS3D02G461200 on 3D, TraesCS4A02G312500 on 4A, TraesCS4B02G074400, on 4B, TraesCS4D02G286100 on 4D, TraesCS5D02G483900 on 5D, TraesCS6A02G083200 on 6A, TraesCS6B02G328600 on 6B, and TraesCS7A02G028300 on 7A. Furthermore, regarding high-confidence candidate genes important for systemic acquired resistance (SAR), it has been revealed that wheat has a long-lasting, broad-spectrum resistance to pathogen infections. These include TraesCS1A02G18340 on 1A, TraesCS1B02G324300 and TraesCS1B02G480300 on 1B, TraesCS1D02G018500 on 1D, TraesCS2A02G540000 on 2A, TraesCS2B02G533800, TraesCS2B02G535000 and TraesCS2B02G583600 on 2B, TraesCS3B02G041900 on 3A, TraesCS5D02G480600, TraesCS5D02G501100 and TraesCS5D02G501800 on 5D, TraesCS7A02G026100 and TraesCS7A02G175200 on 7A, and TraesCS7D02G016800 on 7D. Focusing on the significant QTL regions identified that high-confidence genes TraesCS1B02G324300 on 1B, TraesCS6B02G017900 on 6B, TraesCS7A02G175200 on 7A and TraesCS7D02G016800 on 7D regulate mitogen-activated protein kinase (MAPK) cascades, which are involved in signaling a variety of plant defense responses against pathogen infections (Meng and Zhang, 2013) ( Supplementary Table 6 ).

The association analysis for some agronomic traits, including FD, LA, PH, SL, NKS, and SW, resulted in significant MTAs at the nominal p-value. For instance, a GWA scan for days to 50% flowering for pooled dates resulted in seven MTAs on chromosomes 2B, 2D, 3A, 4B, 6B, 7A, and 7D. However, dissecting the traits in individual environments detected considerable MTAs at the nominal significance threshold. The same trait provided six MTAs that were produced by days to flowering on chromosomes 2D, 3A, 4B, 5A, 6B, and 6D. The identified SNPs explained 11.66% (1123012|F|0-16:G>C-16:G>C on 6B) to 14.70% (1090345|F|0-8:T>A-8:T>A on 6B) of the total variations in days to flowering. Likewise, the association analysis for leaf area identified six MTAs on chromosomes 1A, 2B, 3B, 4D, 6A, and 7B at Holeta; seven MTAs at Kulumsa on chromosomes 1B, 3A, 5D, 6A, 6B, and 7A; and eight MTAs at Meraro on chromosomes 1A, 2B, 2D, 3B, 4D, 6A, 7B, and 7D. The identified SNPs explained 10.04-17.29% of the total variation in leaf area for the SNP markers 1119123|F|0-11:C>G-11:C>G on chromosome 1D and 1265768|F|0-44:T>C-44:T>C on chromosome 4D, respectively. The association analysis for plant height data collected at Meraro identified five MTAs pointing to chromosomes 1D, 2B, 4A, 5D, and 7B. The identified SNPs explained 6.61% (3025953|F|0-25:G>A-25:G>A on 4A) to 12.30% (SNP 1099369|F|0-26:C>T-26:C>Ton 5D) of the total phenotypic variance for plant height ( Supplementary Table 7 ; Figure 11 ; Supplementary Figure 3 ).

The association analysis for pooled spike length identified five MTAs on chromosomes 1A, 2A, 5B, 6B, and 7A. Analysis of the pooled data for days to maturity identified significant SNPs on chromosomes 1B, 2A, 2B, 3A, 5B, and 5D. The association analysis for days to maturity in individual environments identified 9 MTAs at Holeta on chromosomes 1A, 1B, 3B, 5A, and 5B; nine MTAs on chromosomes 1B, 1D, 3A, 3B, 4B, 5B, and 7A for data collected at Kulumsa; and eight MTAs on chromosomes 1B, 2A, 2B, 4A, 5D, 6A and 6B for days of maturity collected at Meraro ( Supplementary Table 5 ). The identified SNPs explained 5.54% to 8.04% of the total variations in days to maturity for the SNPs 1104531|F|0-39:T>A-39:T>A on 2B and 4911245|F|0-46:A>G-46:A>G on 6B, respectively. Association analysis for the pooled number of kernels per spike identified eight MTAs on chromosomes 1B, 1D, 2B, 3A, 3D, 4A, 6B, and 6D. Analyses of the same trait at Meraro provided five MTAs pointing to chromosomes 1D, 2B, 4A, 5D, and 7B. The identified SNPs explained 6.62% of 3025953|F|0-25:G>A-25:G>A on 4A to 13.64% (for 992022|F|0-9:G>A-9:G>A) of the total variations of days to maturity ( Supplementary Table 7 ; Figure 11 ; Supplementary Figure 3 ). GWAS for cumulative number of kernels per spikelets identified seven MTAs pointing to the chromosomes 1B, 3B, 4A, 6D, 7A, 7B, and 7D. The same trait analysis at each individual environment revealed 20 MTAs: six at Holeta on chromosomes 1B, 2B, 3B, 7A, 7B, and 7D; seven at Kulumsa on chromosomes 1B, 2A, 3D, 5A, 5D, 6A, and 7B; and seven at Meraro on chromosomes 1B, 2B, 2D, 3A, 4A, 7A, and 7D. The obtained SNPs explained 6.75-19.15% of the total phenotypic variations in the number of kernels per spikelets by the SNPs 1081747|F|0-27:T>C-27:T>C on 1B respectively.

The association analysis for spike weight found four MTAs on chromosomes 2D, 3A, 3B, and 5B at Kulumsa, and seven MTAs were found on chromosomes 1B, 2B, 2D, 3B, 4D, 5D, and 6A at Meraro. Similar to this, the association analysis for kernels per spike at Meraro identified five MTAs pointing to chromosomes 1D, 2B, 4A, 5D, and 7B. The identified SNPs explained 7.50-11.06% of the total variations in kernels per spike for the SNPs 1000905|F|0-46: A>G-46:A>G on 3B and 1126316|F|0-23:A>G-23:A>G on 2B, respectively ( Supplementary Table 7 ; Figure 11 ; Supplementary Figure 3 ). The study showed that certain QTLs discovered for yellow rust resistance overlapped with QTLs for agronomic traits. For instance, the qYrS.12 gene, which was found to be associated with yellow rust resistance at the flowering stage, is co-mapped with the putative QTL found for plant height on chromosome 2A. Similarly, the putative QTL for grain yield discovered on 1B is also co-mapped with the yellow rust resistance qYrS.07 identified at the maturity stage ( Figure 10 ).

The average disease severity and field response of 178 bread wheat genotypes to P. striiformis f. sp. tritici at Holeta, Kulumsa, and Meraro were recorded in the field. The mean DS, FR, and CI of the tested wheat genotypes at Holeta, Kulumsa, and Meraro were 8.35, 26.92, and 45.08 for DS; 0.22, 0.73, and 8.30 for FR; and 27.55, 45.05, and 47.16 for CI, respectively. Meraro showed the highest DS, FR, and CI, confirming that it is a national hotspot for P. striiformis infestation. It is one of the national and international screening sites of international enter for maize and wheat improvement (CIMMYT) wheat germplasm for yellow rust resistance. The analysis revealed the existence of significant genetic variation among the wheat genotypes examined for a coefficient of infection, suggesting that yellow resistance could be improved through selection breeding. The coefficient of infection for resistance reaction detected in this research (69.66%) is comparable with the reaction response reported by Khalil et al. (2022) (77%) for 426 Indian bread wheat materials. All the considered traits showed moderate heritability (H² = 30 – 60%) except for the coefficient of infection at the pre-heading stage which showed low heritability (H² = 27%). Narrow-sense heritability, which estimates the proportion of phenotypic variations in yellow rust resistance in a wheat panel, is due to the additive genetic effect and was high (h² = 60%), indicating the presence of a strong genetic signal in the data and hence the promising possibility of improving yellow rust resistance by exploiting the resistance source through selection breeding. Comparable high heritability (h² = 88%) for yellow rust resistance traits was reported by (Zegeye et al., 2014). Similarly, Li et al. (2020) described high heritability (h² = 81%) for yellow rust resistance analysis in Chinese endemic wheat genotypes.

Correlation analyses using CI revealed that yellow rust resistance traits were significantly negatively associated with leaf area, plant height, thousand kernel weight, number of kernels per spike, spike weight, and grain yield, indicating that high yellow rust infestations have a considerable effect on reducing yield and yield-related traits. Moreover, plant height has a considerable escaping effect from yellow rust, as the infestation can initiate from the soil. Similarly, a significant negative association between yellow rust infestation and plant height, thousand kernel weight, grain yield, and number of kernels per spike was reported by Solomon (2022) and Saleem et al. (2022). Nevertheless, days to maturity and spike length showed a moderate negative correlation (r< -0.3) with all yellow rust severity, indicating that late maturing and long spiked genotypes have an escaping mechanism for yellow rust infection. The strong negative associations of yield-related, vegetative, and phenological traits with yellow rust resistance traits indicate that these traits should be considered in indirect selection for yellow rust disease resistance.

The presence of three sub-populations (K=3) with substantial mixing was supported by analysis of the population structure and principal components. Ovenden et al. (2017) showed similar indistinct population grouping, increased admixing, and poor population sub-structuring for 312 wheat genotypes using 5k SNP markers. The existence of cryptic familiar relatedness was also verified by kinship analysis, highlighting the significance of using both population structure (Q) and kinship (K) as covariates in marker-trait association analyses. The population stratification, relatedness, and marker effects have been effectively accounted for using the BLINK model employed in the association studies, reducing the confounding effects that could lead to false-positive MTAs. Visualizing the Q-Q plots helped to confirm effective control of the confounding factors.

According to the study, there is a relatively even distribution of markers among the bread wheat sub-genomes, with the A sub-genome contributing the most (10,317), the B sub-genome coming in second (10,979), and the D sub-genome harboring the least (9,756). Similar findings were published by Vikas et al. (2022), who found that 6,036 of the 18,932 mapped SNPs were mapped to the A sub-genome, 7,191 to the B sub-genome, and 5,705 to the D sub-genome. This indicates that the history of natural selection, gene conversion, mutation, and other mechanisms that result in gene-frequency evolution is reflected in each genomic region’s LD (Slatkin, 2008; Rahimi et al., 2019). The A and B sub-genomes had a disproportionately high number of SNPs, which was probably caused by the relatively recent joining of the D-genome to the hexaploid wheat genome. The LD varied amongst the three sub-genomes, and it was low for the A and B sub-genomes, perhaps as a result of their lengthy evolutionary histories in comparison to the D genome.

The discovery of significant marker-trait relationships at different growth stages revealed the presence of significant genotypic diversity among the investigated bread wheat genotypes. At a nominally significant threshold of p-value ≤ 0.001, GWAS analysis discovered 148 MTAs, pointing to 54 QTLs for the coefficient of infection and 179 MTAs for agronomic traits. (Zegeye et al., 2014) also found 38 SNPs in synthetic hexaploid wheat evaluated at Meraro and Arsi Robe. The large sample size and SNPs employed in the current investigation could be the main cause for the considerable variation in the number of significant markers. Among the 18 responsible reported by (Zegeye et al., 2014) for yellow rust resistance, four (1A, 1B, 2B, and 5A) shared the same physical position, which could be due to the difference in the number of used SNPs (2590), genotyping platform (iSelect array), and wheat populations.

In the current study, 48 of the identified yellow rust resistance SNPs were found to be location-specific (at Meraro), showing the presence of P. striiformis races at the test site that was distinct from other test sites, along with race-specific resistance genes. The present study confirmed that chromosome 2B contains more yellow rust resistance loci (24 MTAs). This finding is consistent with previous reports by Chen et al. (2012) and Mallard et al. (2005), who reported approximately 112 P. striiformis resistance genes on wheat chromosome 2B, for instance: Yr9, Yr10, Yr15, Yr24/Yr26/YrCh42 (Cheng et al., 2014), Yr29/Lr46 (Lan et al., 2014), YrExp1 (McIntosh et al., 2020), and YrH52 (Klymiuk et al., 2020). Similarly, the present study identified that chromosome 2B harbored many yellow rust resistance putative QTLs, and this finding was in line with the reports of Feng et al. (2018) and Khalil et al. (2022), who described several yellow rust resistance QTLs on the same chromosome, including Yr64, Yr65, YrTr1, YrAlp, YrH122, YrL693, YrC142, YrMY41, QYr.cau-1BS, QYrco.wpg-1BS.1, and QYrco.wpg-1BS.2. Interestingly, markers found in the current study are located at 74.20-793 Mbp.

A total of 54 QTLs were observed across all environments. Of the 54 QTLs, 17 putative QTLs were detected at Holeta, 3 at Kulumsa, 23 at Meraro, and 11 QTLs were observed in all environments ( Table 7 ). Moreover, 22 of the 54 QTLs were observed across different growth stages, including 10 QTLs at two different growth stages, seven QTLs in three growth stages, one QTL in four growth stages, and two QTLs in all growth stages ( Table 8 ). The failure to generally not repeat QTL effects across all environments could be due to site-specific QTL effects and different disease pressures. The prolonged and intense rains that characterize Meraro led to the largest natural yellow rust infestations across all growth stages. Climate factors, including lingering crop moisture and protracted periods of heavy rain, encourage disease infection and dissemination across the crop canopy (Helen and Sarah, 2015).

Even though it is challenging to compare the positions of QTLs from different studies because of the different mapping methodologies, marker systems, and mapping populations employed, some QTLs found in this study coincided with the mapping positions of previously reported yellow rust resistant genes in the literature. Similar to the current study, several earlier studies reported yellow rust-resistant QTLs on 1A (Fu et al., 2019; Tehseen et al., 2020; Yao et al., 2021; Baranwal et al., 2022; Bouvet et al., 2022), on 1B (Losert et al., 2017; Alemu et al., 2020; Baranwal et al., 2022), on 1D (Long et al., 2019; Jia et al., 2020; Zhang et al., 2021; Baranwal et al., 2022; Kumar et al., 2023) on 2A (Tehseen et al., 2020; Zhang et al., 2021; Zhang et al., 2021; Baranwal et al., 2022), on 2B (Li et al., 2020), on 2D (Mu et al., 2020; Baranwal et al., 2022), on 3A (Tehseen et al., 2022), on 3B (Zegeye et al., 2014; Kumar et al., 2020), on 3D (Bouvet et al., 2022), on 4A (Mu et al., 2020), on 4B (Alemu et al., 2020), on 4D (Rollar et al., 2021), on 5A (Zhang et al., 2021), on 5B (Lu et al., 2014), on 5D (Zhang et al., 2021), on 6A (Baranwal et al., 2022), on 6B (Zhang et al., 2021) on 7A (Yel et al., 2019; Alemu et al., 2020; Tehseen et al., 2020; Baranwal et al., 2022), and on 7D (Long et al,. 2019; Rollar et al., 2021).

Out of the 54 putative QTLs detected in the current study, 12 QTLs (qYrS.11 on chromosome 1D; qYrS.16, qYrS.17, and qYrS.18 on chromosome 1B; qYrS.26 on chromosome 3A; qYrS.31 and qYrS.32 on chromosome 3B; qYrS.40 on chromosome 4B; qYrS.45 and qYrS.46 on chromosome 5D; qYrS.50 on chromosome 6B; and qYrS.58 on chromosome 7D) were not reported in previous literature on wheat and hence could potentially be novel ( Supplementary Table 8 ).

On chromosome 1A, the discovery of defense-related candidate genes like TraesCS1A02G325400 and TraesCS1D02G018500on in the significant markers suggests that the observed QTL areas are functionally associated with plant defense mechanisms against fungal infections. Furthermore, genes such as TraesCS1A02G183400, found on the qYrS.02 QTL region on chromosome 1A, and TraesCS1B02G480300, found on the qYrS.08 QTL region on chromosome 1B, are involved in plants’ systemic acquired resistance to pathogen infections. According to Lawton et al. (1995), systemic acquired resistance is a broad-spectrum resistance acquired following the initial localized infection of plants by pathogens. Genes such as TraesCS1B02G324300, found in qYrS.07, and TraesCS6B02G017900, in the qYrS.49 region, on chromosome 1B and 6B, respectively, control the MAPK cascades that signal various defense reactions, such as the production of or signaling of plant stress, gene activation, and hypersensitive response cell death. Additionally, it is assumed that the majority of the genes similar to the discovered key markers are involved in the manufacture of salicylic acid, an important plant hormone well known for mediating host responses to pathogen infection (Lefevere et al., 2020; Mekonnen et al., 2021).

Due to the lack of shared genetic effects for yellow rust resistance, the majority of the QTLs discovered for vegetative, yield-related, and phenological traits did not overlap with those discovered for yellow rust resistance traits. Numerous associations between agronomic traits and yellow rust resistance traits were non-significant, had minimal to weak negative coefficients, or were both, showing that the traits were independent. However, this study showed that some putative QTLs determined for agronomic traits and putative QTLs identified by yellow rust resistance data overlapped. For example, the potential QTL identified on 1B for grain yield was co-mapped with qYrS.07, identified for yellow rust resistance at the maturity stage. This implies that these traits are controlled by similar loci. As anticipated, grain yield may be impacted by the disease’s vertical advancement rate. The putative QTL determined for plant height on chromosome 2A is co-mapped with qYrS.12, identified for the resistance of yellow rust at flowering. That is consistent with the finding of Mekonnen et al. (2021), who stated that the putative QTL controlling grain yield is co-mapped with the Septoria tritici blotch progress coefficient on chromosome 1A. Traits that facilitate escape, such as tallness, reduce disease resistance by limiting or delaying the vertical spread of the spores up the plant, probably leading them to be scored as resistant when they are inherently not resistant (Miedaner et al., 2013).