Exploring GWAS and genomic prediction to improve Septoria tritici blotch resistance in wheat

Septoria tritici blotch (STB) is a destructive foliar diseases threatening wheat grain yield. Wheat breeding for STB disease resistance has been identified as the most sustainable and environment-friendly approach. In this work, a panel of 316 winter wheat breeding lines from a commercial breeding program were evaluated for STB resistance at the seedling stage under controlled conditions followed by genome-wide association study (GWAS) and genomic prediction (GP). The study revealed a significant genotypic variation for STB seedling resistance, while disease severity scores exhibited a normal frequency distribution. Moreover, we calculated a broad-sense heritability of 0.62 for the trait. Nine single- and multi-locus GWAS models identified 24 marker-trait associations grouped into 20 quantitative trait loci (QTLs) for STB seedling-stage resistance. The seven QTLs located on chromosomes 1B, 2A, 2B, 5B (two), 7A, and 7D are reported for the first time and could potentially be novel. The GP cross-validation analysis in the RR-BLUP model estimated the genomic-estimated breeding values (GEBVs) of STB resistance with a prediction accuracy of 0.49. Meanwhile, the GWAS assisted wRR-BLUP model improved the accuracy to 0.58. The identified QTLs can be used for marker-assisted backcrossing against STB in winter wheat. Moreover, the higher prediction accuracy recorded from the GWAS-assisted GP analysis implies its power to successfully select superior candidate lines based on their GEBVs for STB resistance.

Septoria tritici blotch (STB) is a destructive foliar diseases threatening wheat grain yield.Wheat breeding for STB disease resistance has been identified as the most sustainable and environmentfriendly approach.In this work, a panel of 316 winter wheat breeding lines from a commercial breeding program were evaluated for STB resistance at the seedling stage under controlled conditions followed by genome-wide association study (GWAS) and genomic prediction (GP).The study revealed a significant genotypic variation for STB seedling resistance, while disease severity scores exhibited a normal frequency distribution.Moreover, we calculated a broad-sense heritability of 0.62 for the trait.Nine single-and multi-locus GWAS models identified 24 marker-trait associations grouped into 20 quantitative trait loci (QTLs) for STB seedling-stage resistance.The seven QTLs located on chromosomes 1B, 2A, 2B, 5B (two), 7A, and 7D are reported for the first time and could potentially be novel.The GP cross-validation analysis in the RR-BLUP model estimated the genomic-estimated breeding values (GEBVs) of STB resistance with a prediction accuracy of 0.49.Meanwhile, the GWAS assisted wRR-BLUP model improved the accuracy to 0.58.The identified QTLs can be used for markerassisted backcrossing against STB in winter wheat.Moreover, the higher prediction accuracy recorded from the GWAS-assisted GP analysis implies its power to successfully select superior candidate lines based on their GEBVs for STB resistance.
Hexaploid winter wheat (Triticum aestivum, 2n = 6× = 42, AABBDD) occupies the largest arable land in Northwest Europe 1 .An estimated 21% of world's wheat production is lost due to diseases caused by pests and pathogens 2 .Septoria tritici blotch (STB) caused by the fungal species Zymoseptoria tritici (teleomorph Mycosphaerella graminicola) 3 is the second leading disease for yield loss after stripe rust in Northwest Europe 4 .The disease causes an estimated yield loss of 5.51% in Northwest Europe that is two folds higher than its global average with 2.44% 2 .Yield losses up to 50% have been reported in epidemic years of the STB disease 5,6 .Most applied management strategies may still be as effective to overcome the disease outbreaks where wheat fields are subject to STB epidemics as a result of the airborne Z. tritici ascospores 7 .Moreover, STB management practices negatively affect the environment due to the intensive application of fungicides to control the disease accounting 70% of all fungicides used in wheat fungal disease management 8 .Besides, fungicide application costs up to $1.2 billion USD annually to manage the disease incidences in Europe 9 .
An integrated approach consisting of breeding for host resistance combined with other management practices is a sustainable strategy to mitigate STB impact.Wheat resistance breeding plays a major role in developing varieties with enhanced resistance lowering the environmental impact of fungicides application.So far, 22 qualitatively inherited major genes (Stb genes) have been identified for STB resistance in wheat 10,11 .However, the rapidly evolving Z. tritici populations as a result of sexual reproduction under field conditions caused the selection to be in favor of emerging new virulent strains that can overcome the identified major gene resistance 12 .In contrast, the higher number of minor to moderate effect minor genes inherited quantitatively are advantageous to that of major qualitatively inherited genes.Compared to qualitative resistance, quantitative resistance to STB is long acting against the range of diverse Z. tritici isolates.That is due to the cumulative effect of many genes

Materials and methods
Plant material and experimental design.A set of 316 winter wheat advanced breeding lines provided by Lantmännen, Svalöv, Sweden, were evaluated for their resistance for STB at seedling-stage under controlled growth conditions.Four known varieties having three levels of resistance against the applied STB isolates were included as checks on the current experiment.These checks were Julius (resistant), KWS Kerrin and Stigg (moderate resistant) and Nimbus (susceptible).
The experiment was conducted in a randomized augmented block design with two replicates.Each replicate included 15 blocks and a single block comprised 23 breeding lines and 4 checks.Randomization of genotypes was done using the design.daufunction in Agricolae package 36 using R environment 37 .

Plant growth condition and inoculation.
Winter wheat seeds were stratified for 48 h in dark conditions at 3 °C on wetted filter paper followed by germination for 24 h at room temperature.Six to eight healthy germinated seeds were transferred to 8 × 8 × 9 cm plastic pots filled with potting peat soil (Gröna linjen, SW HORTO AB, Sweden) and kept for 4 days followed by thinning to two seedlings per pot.One gram of KH 2 PO 4 per block was applied to promote root development and enhance seedlings recovery after thinning.The plants continued growing at 23/22 °C with 16/8 h of day/night cycle with relative atmospheric humidity (RH) of 60%.Plants were weekly fertilized with nitrogen and potassium fertilizer (SW-BOUYANT 7-1-5+MIKRO) added with equal amounts to individual blocks.
The inoculum was prepared by growing two single spore isolates collected from Alnarp and Svalöv, Sweden, following the protocol described by 18 .The inoculum concentration was adjusted to 10 × 10 5 conidial spores/ml followed by adding the surfactant Tween ® 20 with 0.002% v/v.Three-leaf-stage 19-days-old winter wheat seedlings were spray-inoculated three times after marking the second and the third leaves.The leaves were left to dry after each spray for 20-30 min.On the third time spraying, plants were moved into a high humidity chamber with 90% RH at 23 °C for 48 h (16 and 8 h of light and dark conditions).After incubation, RH was lowered to 65% and continued until completion of the experiment.
In a small separate test, the virulence of the two isolates used in this study was initially examined by inoculating the four varieties with known resistance background to STB at seedling-stage including Kranich and Julius (resistant), Stigg (moderately resistant), and Nimbus (susceptible).Previous studies have identified cultivar Stigg as resistant to STB in different field conditions by extended latent phase before switching to the necrotrophic phase 14,38,39 .Odilbekov et al. 18 identified cultivars Kranich and Nimbus as STB resistant and susceptible, respectively.Other studies showed that the cultivar Julius has high level of resistance to several diseases of wheat including STB 28,40 .For this purpose, 25 advanced breeding lines and 41 official trial lines were tested in an unreplicated augmented block design.The four cultivars (Kranich, Julius, Stigg and Nimbus) were used as checks dispersed across the 66 genotypes and replicated for a total of 19 times.

STB disease evaluation.
Unlike natural infection (Fig. 1A), plants in greenhouse condition starts with general chlorosis that continues to spread across the whole leaf or partially from the leaf tip (Fig. 1B).Subsequently, reddish necrosis develops in place of chlorosis leading to collapse of the tissues in the infected area (Fig. 1C-E).
In the current study, the visual assessment of the disease was carried out 15 days post inoculation (DPI) on the second and third leaves of plants at seedling-stage under greenhouse conditions.Genotypes reaction to the disease was visually assessed every third day for four consecutive time points.An adjusted visual scaling scheme for disease severity was followed where reddish necrotic areas were estimated as percentage of disease severity ratio relative to the total leaf area following the procedure recently applied by Odilbekov et al. 18 .

Phenotypic analysis.
The phenotypic data analysis from the STB disease scoring was conducted in two steps.First, the four checks repeated in each block were used to adjust the means recorded from four consecutive scoring time points within each replicate using the Agricolae package 41 following the model: where Y ij is the adjusted mean of the ith genotype in the jth block, u is the overall mean, G ij is the effect of the ith genotype in the jth block, B j is the effect form j th block and ɛ ij the overall residual.An area under diseases progression curve (AUDPC) was estimated from the adjusted means using the following model: where y i = disease score at time t i ; t (i+1) − t i = time (days) between two STB scoring time points; n = total number of scoring time points.
The second step was estimating the best linear unbiased prediction (BLUPs) across the two replicates from the adjusted mean AUDPC values of genotypes using META-R 6.04 42 following the model: where Y ik is the BLUPs of ith genotype in the kth replicate, u is the overall mean, G ik is the ith genotype effect in the kth replicate, R K is the effect of Kth replicate and ɛ ik the overall residual.Analysis of variance (ANOVA) and broad-sense heritability (H 2 ) was retrieved in this step along with the adjusted mean values of AUDPC scores.Frequency distribution of AUDPC BLUPs was performed in the Minitab software package (Version 18).
Genome-wide association analysis.The current winter wheat panel was previously genotyped using 25K SNP array by TraitGenetics GmbH, Germany that produced 24,145 SNP markers 43 .Finally, 10,120 SNP markers were selected for the GWAS analysis after excluding markers with minor allele frequency (MAF) ≤ 0.05 and with ≥ 0.2 missing values per individual.Nine GWAS models including 2 single and 7 multi-locus based models available in the Genome Association and Integrated Prediction Tool (GAPIT) 3.0 44 and multi-locus random-SNP-effect mixed linear model (mrMLM) v4.0.2 45 were employed to spot marker-trait associations.
The fixed and random model circulating probability unification (FarmCPU) 46 , Bayesian-information and linkage-disequilibrium iteratively nested keyway (BLINK) 47 , multi-locus random SNP-effect mixed linear model (mrMLM) 48 , fast multi-locus random-SNP-effect efficient mixed model association (FASTmrEMMA) 49 , FAST multi-locus random-SNP-effect Mixed Linear Model (FASTmrMLM) 50 , polygene-background-control-based least angle regression plus empirical Bayes (pLARmEB) 51 and integration of Kruskal-Wallis test with empirical Bayes under polygenic background control (pKWmEB) 52 were the seven multi-locus models applied in the current analysis.Settlement of MLM under progressively exclusive relationship (SUPER) 53 and general linear model (GLM) 54 were the two single-locus models included in the GWAS analysis.Population structure as principal components (PCs) and pair-wise kinship similarity matrix were included with some of the statistical models required to overcome the false-positive marker-trait associations.The Quantile-quantile (Q-Q) plots generated from the observed against expected − log 10 p-values were used to evaluate the performance of the statistical models.Furthermore, the Bonferroni corrected threshold, applied to report major-QTLs, was calculated as, Bonferroni threshold = − log 10 (0.5/n) , where n = number of total SNP markers applied to explore markers linked to STB resistance.Hence, Bonferroni threshold = − log 10 (0.5/10,120) = 4.3.Several previous GWAS studied with Bonferroni corrected threshold (P ≤ 0.05) have reported as major QTLs 20,55,56 .
Depending on the model, either − log 10 p-value ≥ 4 (P ≤ 0.0001) or the logarithm of odds (LOD) scoring ≥ 4 were used as the exploratory significance thresholds to report the nominal-QTLs identified from the current marker-trait associations.Manhattan plots were generated for figurative visualization of the associated SNP markers across the wheat chromosomes.The SNP markers genetic positions were retrieved from the 90K SNPs consensus map 57 while their physical position from the International Wheat Genome Sequence Consortium (IWGSC) v1.1 58 and markers within 5 cM were considered as a single QTL.

Genomic prediction analysis.
The genomic prediction analysis was conducted with the ridge regression BLUP (RR-BLUP) model using the "rrBLUP" package 59 in R environment following the mixed model formula: where y is the vector of adjusted AUDPC mean score for STB resistance; x and β represents the designed matrix and vector of fixed effects, respectively; Z is the designed matrix for random effect SNP markers and μ is a vector of estimated random effect with μ ~ N(0, Gσ 2 μ), where G is a genomic relationship matrix calculated from all SNPs; and ε is the residual error.
The weighted RR-BLUP (wRR-BLUP) model was tested after fitting the five topmost significant GWAS-SNP markers (based on their P values) and fitted as fixed effects 60 .To avoid model overfitting, GWAS-SNPs were discovered from the training population (four-folds) excluding the validation population (one-fold) following the five-fold cross-validation scheme using FarmCPU model.The random sampling of genotypes into folds was done using the sample() function in R environment.
The prediction accuracy of models was assessed through cross-validation analysis.For the RR-BLUP model, individual genotypes were randomly split into training and validation sets with 80 and 20% ratio, respectively, and repeated for 500 iterations.However, in the wRR-BLUP model, the panel was first randomly divided in to five folds followed by GWAS analysis with only the four folds, which later used as training set.Then, the five topmost significant SNPs were identified and fitted as fixed effects in the prediction model.The genomic estimated breeding values of the remaining one fold was estimated with the trained model.The GWAS followed by genomic prediction analysis was repeated for 20 times.
Predictive ability was estimated as the correlation coefficient between the observed AUDPC adjusted mean values of genotypes and genomic estimated breeding values predicted for the test set based on the effect estimates of genotypes in the training set.Prediction accuracy was then calculated from prediction ability divided by the square root of traits' broad sense heritability 61,62 .

Ethics approval.
All plant experiments were conducted according to relevant institutional, national, and international guidelines and legislation.

Results
Phenotypic analysis for STB resistance.The four known varieties, Kranich, Julius, Stigg and Nimbus having a varied degree of STB resistance was initially tested for any possible genotype-isolate/strain specific resistance via inoculating with the two isolates.The inoculation with the two mixed isolates elicited STB responses corresponding to the degree of resistance relative to their reported resistance backgrounds (Fig. 2). 3 falling between the resistant and the susceptible cultivars (Fig. 2).Analysis of variance indicated a highly significant phenotypic differences among evaluated winter wheat lines for STB resistance (Supplementary Table S1).The AUDPC-BLUPs scores of the 316 breeding lines followed a normal distribution (Fig. 3) ranging from 193.6 to 666.2 with average and standard deviation values of 434.2 and 91.58, respectively.High broad-sense heritability (0.62) was recorded from the evaluated breeding lines.

GWAS results.
The multi-locus GWAS models discovered 24 marker-trait associations (MTA) significantly linked to STB resistance (Supplementary Table S2).From these MTAs, 14 were detected with − logP ≥ 4 (P ≤ 0.0001) while the remaining 10 with − logP ≥ 3 (P ≤ 0.001) with atleast one tested model.These markers were located within 20 different QTLs identified on 14 chromosomes including 1B, 2A, 2B, 2D, 3A, 3B, 4A, 4B, 5A, 5B, 6D, 7A, 7B and 7D (Table 1).The AUDPC_BLUP distribution of genotypes based on the alleles of nine significantly linked markers with MAF > 0.35 is given in Supplementary Fig. 1.The SNP marker AX-158596603, under the major-QTL SLUSTB_4, on chromosome 2A (64 cM) had a highly significant association (P < 0.00001) and LOD score > 5 for STB resistance.Several models including Blink, pLARmEB and FASTmrEMMA identified this SNP marker highly significantly linked to host resistance to the pathogen.The other nearby SNP marker AX-158573239 (63.8 cM) also exhibited a multi-model based highly significant association with the trait.The SLUSTB_7 comprising two co-localized SNPs Kukri_c17_1246 and BS00083329_51 (135.5 cM) on chromosome 2B was the other major-QTL with a highly significant multi-model based association with STB resistance.The  www.nature.com/scientificreports/SNP marker wsnp_Ex_c12220_19528388 (SLUSTB_10) on chromosome 3B (60.5 cM) has shown a highly significant association with the trait via several models including GLM, SUPER, BLINK, mrMLM, pLARmEB and pKWmEB (Table 1).The SLUSTB_12 QTL comprised the marker wsnp_CAP12_c1101_569783 on chromosome 4B (66.4 cM) which was detected with several models such as FarmCPU, mrMLM and pKWmEB with a highly significant threshold P < 0.00001 (Fig. 4).Four different models including FarmCPU, Blink, FASTmrEMMA and FASTmrMLM identified the SNP marker Kukri_c51101_3510 (SLUSTB_19) on chromosome 7B (79.8 cM) highly significantly (P < 0.00001) linked to STB resistance.Four multi-locus models, pLARmEB, mrMLM, pKW-mEB and FASTmrMLM identified the SNP AX-158537280 (SLUSTB_17) on chromosome 7A (51.2 cM) significantly linked to STB resistance with the current panel.Moreover, several QTLs were identified with either at least with two multi-locus or both with the single-and multi-locus models.For instance, the SNPs on QTL SLUSTB_5 (Tdurum_contig54634_815, 51.9 cM) and SLUSTB_6 (AX_158557660, 81.4 cM) on chromosome 2B were detected by pLARmEB and pKWmEB models with a high significance threshold.

Genomic prediction analysis.
The RR-BLUP model with 80 and 20% of genotypes as training and validation sets, respectively, estimated the GEBVs for STB resistance with a prediction accuracy of 0.49 averaged from 500 iterations.The prediction accuracy with this model ranged from 0.21 to 0.84 (Fig. 5A).Genomic prediction of STB resistance with the current panel was further evaluated with the wRR-BLUP model.In this model, the five top-most significantly linked SNP markers identified only from the training set were fitted as fixed effects.Following the five-fold cross-validation scheme, GWAS was conducted with the training set (four-folds) with FarmCPU model masking a fold that was used as validation set and analysis was repeated for 20 times.The five-top-most GWAS-identified SNP markers can be find in the supplementary file (Supplementary Table S3).
The wRR-BLUP model estimated the GEBVs of validation individuals for STB with a mean genomic prediction accuracy of 0.58 averaged from 20 runs (Fig. 5B).Similarly, the 20 training-validation set iterations were conducted with the RR-BLUP model excluding the five-fixed effect SNPs and the average prediction accuracy was 0.53.The genomic prediction accuracy of the wRR-BLUP and RR-BLUP models with the 20 iterations was in a range of 0.31-0.85and 0.26-0.70,respectively.

Discussion
Phenotypic characterization and GWAS analysis for STB.Genetic improvement of winter wheat for resistance against STB is an indispensable approach to minimize the disease impact on wheat grain yield.The deployment of breeding lines with a wide genetic base can accelerate the development of wheat varieties resistant to wheat diseases, while keeping other important traits, such as yield optimized to a particular environment.The potential source of variation observed in the current studied breeding lines enabled the identification of genetic factors underpinning resistance to STB in juvenile plants under controlled conditions.Selection at seedlingstage is an essential step before testing in multi-environment trials hence an increased number of seedlingstage markers has been spotted that overlaps with known APR genes to STB resistance.This method minimizes the sheer number of susceptible individual lines from being tested in field conditions saving high amount of resources while increasing the selection accuracy.Previous studies reported Stigg as a resistant variety for STB.However, in the current study, the variety fell below the most resistant variety Kranich and followed by Julius, while Nimbus was susceptible.Variations in seedling-stage resistance attributed to the spatio-temporal appearance in the latent phase leading to necrotic symptoms can aid in explaining the quantitative nature of resistance to STB of the tested genotypes.By using an AUDPC approach in association with the genomic markers of the tested genotypes, seedling-stage markers were detected that can potentially serve as a source of resistant novel QTLs.Additionally, the analysis showed a number of identified seedling-stage markers that co-localized with QTL identified as either seedling or adultstage resistance.
In this study, 14 SNPs identified as significantly linked to STB-seedling-stage resistance were from six chromosomes located on the B sub-genome while the other 7 and 3 SNPs were on the chromosomes from the A and D sub-genomes, respectively.Several studies have similarly reported the B and A sub-genomes with highest number of STB major resistance genes where 16 out of 21 Stb genes were also identified from these two sub-genomes 10 .
The current GWAS analysis discovered twenty STB resistance QTLs out of which seven could potentially be novel.The identified novel markers were located on seven chromosomes including 1B, 2A, 2B, 5B (two), 7A, and 7D.The remaining 13 QTLs overlapped with previous reports either as seedling-stage (three), adult-stage (four) or all plant stage (six) QTLs for STB resistance.The majority of the identified MTAs were discovered from the multi-locus GWAS models even though few were exclusively discovered with single locus models such as the markers on 5B and 7B (Table 1).This could be due to the strong statistical power of the multi-locus models considering that they could manage detecting associations utilizing information from multiple markers simultaneously.Even though being able to detect lower number of MATs 63 , single-locus models are still an invaluable tool for marker discovery mainly for the discovery of markers with strong effect 64,65 and can indicate to the robustness of the associations when MTAs are detected together with multi-locus models such as that on 3B (Table 1).
As to our knowledge, the seven identified QTLs in this study including SLUSTB_2, SLUSTB_3, SLUSTB_6, SLUSTB_14, and SLUSTB_20 on chromosomes 1B, 2A, 2B, 7A and 7D, respectively, and SLUSTB_14 & SLUSTB_15 on chromosome 5B have not been reported in previous studies and could potentially be novel QTL sources to seedling-stage STB resistance in wheat.
Using the same isolates and under similar growth conditions, Odilbekov, et al. 18 identified 10 MTAs associated with seedling-stage resistance located within 5 QTLs using plant materials comprised landraces and cultivars widely grown in the Nordic region for the last 100 years.Out of these, the two overlapped exactly with the currently identified QTLs.The marker BS00066305_51 in SLUSTB_2 QTL was identified on the short arm of 1B distally located from the centromere at 110.1 cM.Generally, 1BS is a major source for Stb resistance genes including the Stb2 66 , and Stb11 10,67 .An earlier study has shown the presence of a QTL associated with several markers spanning a distance of 1.71 cM between 97.36 and 99.07 cM on 1BS identified from a panel of 175 winter wheat landraces and old cultivars at seedling-stage 18 .However, no previous QTL has been reported in close proximity to the current identified marker on the chromosome 1BS that leads us to report as a possible novel QTL for seedling-stage STB resistance in wheat.The SLUSTB_3 associated with the marker AX-158572447 at 20.1 cM on chromosome 2A appears as another potential novel marker associated with seedling-stage STB resistance.The marker Tdurum_contig54634_815 from the SLUSTB_5 QTL identified on the short arm of chromosome 2B (51.9 cM) was exactly overlapped with a previous report by Gerard et al. (2017) with a DArT marker wPt2106 located at 51.86 cM.Another reported QTL on this chromosome region associated with adult-plant stage STB resistance was located on 65 cM 20 .On chromosome 2B, the marker AX_158557660 (81.4 cM) from SLUSTB_6 QTL could possibly be the other potential novel marker identified in the current study for seedling-stage STB resistance.Odilbekov, et al. 18 identified a marker on chromosome 2B linked to seedling-stage STB resistance located at 96.99 cM, which is far from the currently detected marker by more than 15 cM.The SLUSTB_14 and SLUSTB_15 were the other two possibly novel QTL identified in the current analysis on chromosome 5B.The two significantly linked SNPs on SLUSTB_14 (73.6 cM) identified in the current study were 13 cM far from the previously reported marker Excalibur_c17489_804 identified with major effect QTL explaining 28% of the total phenotypic variation 23 .The markers AX_158537280 on chromosomes 7A (51.2 cM) and IAAV4542 on chromosome 7D on SLUSTB_17 and SLUSTB_20, respectively, are the other two candidate novel QTLs spotted on the current study.Both Stb4 and Stb5 have been mapped on the short arm of 7DS chromosome arm 68,69 while the currently identified marker IAAV4542 was found on the long arm of chromosome 7DL located at 142 cM 70 .Therefore, it is not likely that this marker is associated with either of these genes.
The SNP marker AX_89326139 (40.3 cM) on chromosome 1B appeared to overlap with the previously identified marker IAAV3905 (41.3 cM) as APR QTL for STB 29 .Kidane et al. 71 29 .The SLUSTB_9 QTL with the SNP marker wsnp_Ex_c8517_14315660 (82.4 cM) identified on chromosome arm 3AL was nearby to the marker wsnp_Ex_c5929_10402147 (86.2 cM) previously spotted MTA for seedlingstage STB resistance 18 .Nearby to this QTL, a meta-QTL analysis for several biotic stresses reported four QTL situated between 80.4-87.1 cM 74 .This chromosome region has been identified as a source of several MTAs for STB resistance in previous investigations 71,[75][76][77] .The QTL SLUSTB_10 on chromosome 3B comprised the marker wsnp_Ex_c12220_19528388 (60.5 cM) was detected by five different single-and multi-locus models.Other previous studies reported multiple MTAs for STB resistance with different Z. tritici isolate adjacent to this marker 72 .Alemu et al. 20 reported a multi-environment stable MTA for APR to STB on chromosome 3B but far by 10 cM from the MTA identified in this study.These findings could lead us to speculate a potential QTL possibly existing on this chromosome region linked to all-stage STB resistance in wheat.Hence, further investigations on the validation of this candidate QTL region could enhance marker-assisted selection against the pathogen.
Excalibur_c4325_1150 marker (SLUSTB_11 QTL) was identified on chromosome arm 4AL located at 120.4 cM.Two nearby SNPs on the long arm of chromosome 4A, located on 121.4 cM and 122.5 cM, were previously reported significantly linked to STB resistance inoculated by a single Z. tritici isolate at seedlingstage 72 and naturally infected winter wheat adult-plants 20 .Muqaddasi, et al. 29 reported the SNP marker wsnp_ JD_c27162_22206547 that exactly overlapped with the currently identified marker on chromosome arm 4AL at 120.4 cM associated with STB APR detected from 371 artificially field-inoculated winter wheat genotypes.
It is yet unknown whether 4B chromosome is comprised major genes for STB resistance 10 .However, the currently identified marker wsnp_CAP12_c1101_569783 (66.4 cM) on 4B had a highly significant association with STB resistance discovered by the multi-locus model FarmCPU with -log10 P > 9 (Fig. 4) and other two multilocus models with LOD score > 4.8.Louriki et al. 73 reported two SNP markers located on the same chromosome arm identified from a panel of 377 advanced breeding lines of spring wheat associated with seedling-stage STB resistance each detected by one of the two tested isolates.The marker RAC875_c24515_602 located at 76.8 cM 73 could possibly from the same QTL with the marker identified in the current study.Using eight-founder MAGIC winter wheat population, Riaz et al. (2020) discovered several QTLs for adult-plant STB resistance on chromosome 4B including the marker RAC875_c87897_333 (655.94Mbp) which is found in nearby to the currently identified MTA.Hence, this region could be another source for QTLs of all-stage STB resistance.
SLUSTB_8, SLUSTB_16 and SLUSTB_20 were the three QTLs identified from the D sub-genome on chromosomes 2D, 6D and 7D, respectively.The SNP marker AX_111036153 (2.6 cM) on chromosome 2D was identified in the current study that could be linked to a potential QTL identified for adult-plant STB resistance with strong association explaining a high portion of variation 23 .The marker IAAV64 in the QTL SLUSTB_16 discovered on chromosome 6D found at 466.55 Mbp.Similarly, Riaz et al. 23 detected a very closely SNP marker wsnp_Ex_c13188_20825019 (464.72 Mbp) linked to adult-stage STB resistance.
In general, 13 of the currently identified marker-trait associations have exactly overlapped with QTL regions previously reported by several studies.This could validate applied research protocols and procedures in the current study and increase the search and validation of valuable chromosome regions with resistance sources for the STB wheat pathogen.Furthermore, seven potentially novel QTLs identified in this study could also be used as a first brick in searching additional sources of resistance QTLs for the newly evolved pathogen strains.
Genomic prediction for STB resistance.In addition to the well characterized stb genes, various minorto major-effect QTLs are involved in host resistance against the STB pathogen in wheat 10 .Because of this, STB resistant wheat variety development through pyramiding of identified QTLs via marker-assisted selection has been a challenging task.Genomic prediction is a powerful method to accelerate the genetic gain of several quantitatively inherited traits in plant breeding 78 .Unlike the GWAS or linkage mapping methods, genomic prediction estimates the breeding values of individuals for traits of interest accounting all contributing QTLs based on their overall marker information 79 .This method is particularly an invaluable genomic tool for variety development with STB resistance and other similar traits controlled by several QTLs.The current study estimated the GEBVs of the 20% of 316 genotypes with two genomic prediction models trained with the remaining 80% of the panel.The RR-BLUP model estimated the GEBVs of STB resistance with prediction accuracy of 0.49 and 0.53 averaged from 500 and 20 iterations, respectively.A previous study reported a genomic prediction accuracy of 0.47 for STB resistance at seedling-stage with a 175 winter wheat panel comprising old cultivars and landraces 18 .With this panel, a low to moderate genomic prediction accuracy (0.15-0.35) was recorded for adult-stage STB resistance from multi-environmental field trials conducted in Denmark, Estonia, Lithuania, and Sweden 20 .Muqaddasi et al. 29 reported a genomic prediction accuracy of 0.43 for adult-stage STB in 371 European winter wheat varieties.Juliana et al. 80 reported a mean genomic prediction accuracy of 0.45 for STB adult plants resistance from CIMMYT International Bread Wheat Screening Nurseries (IBWSNs) with more than 600 lines evaluated at CIMMYT's research station, Toluca, Mexico for 3 years.However, several factors could affect the genomic prediction accuracy including the size of the training population, marker density, population structure, level of linkage disequilibrium and quality of the phenotypic data applied to train the model 78 .This RR-BLUP model

Figure 1 .
Figure 1.The Septoria tritici blotch (STB) disease lesions filled with pycnidia on infected winter wheat leaves under natural conditions (A), compared to STB symptoms observed in greenhouse artificially inoculated plants (B-E).Evaluating STB resistance in greenhouse-inoculated plants was carried out by scoring reddening necrotic leaf area that may be demonstrated partially (B) and (C) or fully (D) on the leaf.Scoring seedling stage infected leaves was done from the second and third leaves 15 days post inoculation (DPI), where they become the lower infected leaves (E).

Figure 2 .
Figure 2. Frequency distribution for the area under diseases progression curve (AUDPC) score with the four winter wheat varieties having three known levels of resistance with the two applied Zymoseptoria tritici isolates.

Figure 3 .
Figure 3. Distribution of the adjusted mean (BLUPs) AUDPC values of STB resistance recorded from the 316 winter wheat breeding lines.StDev, standard deviation; N, the number of tested genotypes.

Figure 4 .
Figure 4. Manhattan (left) and Q-Q (right) plots of marker-trait associations identified for seedling-stage resistance to STB. (A) Manhattan (left) and Q-Q plots generated from the three multi-locus models (mrMLM, FASTmrMLM and FASTmrEMMA) with pink dots represents the SNP markers discovered by more than one model while dark blue dots represent the markers discovered by a single model.The dashed horizontal line on the diagram represents LOD score of 2.0.The other plots are for FarmCPU (B) and Blink (C) with green solid line representing the Bonferroni corrected thresholds at P = 0.05.The red solid line and green dash line represent the exploratory and false-discovery (FDR)-based thresholds at P = 0.0001 and 0.05, respectively.

Figure 5 .
Figure 5. Genomic prediction for STB resistance using 316 winter wheat genotypes with two different statistical models.(A) Genomic prediction analysis with 80-20% training-validation sets cross-validation analysis conducted for 500 iterations with the RR-BLUP model.(B) The wRR-BLUP model with the five topmost significantly linked SNPs fitted as fixed effects.GWAS was conducted only in the training population and analysis was done for 20 times.Similarly, the 20 training-validation set splits were tested with the RR-BLUP model after excluding the fixed effect SNPs.N, number of iterations; μ, average values of iterations.

Table 1 .
List of QTLs, SNP markers and their effect, and models applied for the current GWAS analysis.
Chr chromosome, MAF minor allele frequency, QTN quantitative trait nucleotide, LOD logarithms of odds; *markers detected by the model(s) at P ≤ 0.001; ** markers detected by the model(s) at P ≤ 0.0001; *** markers detected by the model at P ≤ 0.00001; NA not applicable.QTN effect and LOD scores are estimated only in the multi-locus based mrMLM models.Vol.:(0123456789) Scientific Reports | (2023) 13:15651 | https://doi.org/10.1038/s41598-023-42856-x 73entified a QTL qSTB.04 physically located at 587.28Mbp identified in diverse panel of Ethiopian durum wheat landraces close to the currently detected marker AX_89326139 at the physical distance of 544.5 Mbp.However, it is noteworthy to mention that these studies identified the respective markers on adult plants and the region could be a potential source of allstage STB resistance in wheat.The two closely located markers AX_158573239 (63.8 cM) and AX_158596603 (64 cM) in SLUSTB_4 overlapped with the recently identified QTL linked to seedling-stage STB resistance in a diverse germplasm of 185 genotypes associated with three Z.tritici isolates72.The two SNP markers of QTL SLUSTB_7, BS00083329_51 and Kukri_c17_1246 (135.5 cM) were only distant by 7.5 cM from the previously reported seedling-stage QTL SRT_71-R3_2 on the long arm of chromosome 2B73.Similarly another nearby SNP marker AX_94734086 (145 cM) was identified for APR to STB and claimed that this QTL could be part of the Stb9 resistance gene involved in all stage STB resistance in wheat