The research team, led by Cathrine Kiel Skovbjerg of Nordic Seed A/S and Aarhus University together with Guillaume Ramstein of Aarhus University’s Center for Quantitative Genetics and Genomics, assembled an extraordinary dataset of 6,484 phenotyped inbred lines drawn from four distinct European barley breeding programs: a six-rowed winter population of 581 lines, a two-rowed winter population of 1,640 lines, a six-rowed spring population of 631 lines, and a two-rowed spring population of 3,632 lines. These populations differ not only in growth habit, with winter types sown in autumn and spring types sown in spring, but also in row type, a morphological distinction reflecting whether each spike carries two or six grain rows, and in the breeding goals that have shaped their genomes over decades of selection. Field trials ran from 2013 to 2024, primarily at Odder in Denmark, with additional testing in Nienstädt and Güstrow in Germany, generating between five and eleven distinct year-by-location environments per population.

Phenotyping was carried out under real epidemic conditions. Fungicides were deliberately withheld so that natural leaf rust pressure could reveal genetic differences in susceptibility, and each plot was scored on a one-to-nine scale, from fully resistant to severely infected, at disease onset and again roughly a week later, corresponding to stem elongation through pre-flowering growth stages. The researchers then applied linear mixed models to partition the observed variation into genetic, environmental, and residual components. Genotype-by-environment interaction accounted for 6.9 to 23.4 percent of plot-level variation, while main genetic effects explained between 37.1 and 48.3 percent, encouraging the team to focus on quantitative trait loci with consistent effects across environments. Broad-sense heritability estimates confirmed that a substantial and usable fraction of the variation in rust scores was genetic rather than noise.

On the genotyping side, the lines were profiled with Illumina iSelect SNP arrays, imputed with Beagle, and filtered down to a final dataset of 13,045 SNPs across 6,484 individuals. Principal component analysis confirmed what the breeding histories implied: the first principal component cleanly separated winter from spring types, the second reflected row-type differentiation, and the third fully isolated the six-rowed winter population, which carries a unique genetic signature. Private alleles, variants unique to a single population, were surprisingly few, ranging from just 10 in the six-rowed spring population to 95 in the two-rowed spring population, many of the latter concentrated on chromosome 1H near the Mla powdery mildew resistance locus, a hallmark of historical introgression from wild germplasm. Critically, when variants rare within any single population were pooled across all four populations, only 2.2 percent remained rare, demonstrating that combining populations dramatically increases allele frequencies and, with them, the statistical power to detect associations.

When the researchers ran conventional single-population GWAS, each population yielded a small number of major signals: nine marker-trait associations in the six-rowed winter population, sixteen in the two-rowed winter population, nine in the six-rowed spring population, and thirty-eight in the two-rowed spring population. In three of the four populations, all associations collapsed into a single quantitative trait locus, while the large two-rowed spring population revealed three. The effect sizes were striking. A QTL on chromosome 6H explained 46.2 percent of the additive genetic variance in the two-rowed winter population, and a QTL on chromosome 2H explained 35.3 percent in the six-rowed spring population. The latter mapped to a narrow interval physically distinct from all previously described Rph loci, with its lead SNP sitting inside a protein kinase gene flanked by three NBS-LRR class disease resistance genes, both canonical components of plant pathogen recognition machinery. The six-rowed winter signal on chromosome 5H, explaining 14.2 percent of genetic variance, was likewise anchored near a TIR-NBS-LRR resistance protein gene. In the two-rowed spring population, one QTL colocalized with the known Rph5 and Rph7 genes on chromosome 3H, another overlapped the six-rowed winter signal on 5H, and a third coincided with Rph20, a well-characterized adult plant resistance gene.

A key subtlety emerged when the team examined how these marker effects behaved across individual environments. Effect magnitudes varied significantly between years and locations, yet the direction of effect, whether an allele increased or decreased susceptibility, remained consistent across environments for at least five of the six lead SNPs tested. Some markers showed dramatic temporal shifts: the lead SNP of the second chromosome 5H peak in the two-rowed spring population had essentially no effect from 2013 to 2017 but exerted effects of minus 0.33 to minus 0.77 disease-scale units per minor allele copy from 2018 onward, a pattern the authors suggest may reflect shifting virulence dynamics in the pathogen population. Other markers showed isolated spikes in single environments, such as an outsized effect in the drought-heavy 2018 Odder season. Importantly, changes in allele frequency over time did not explain these shifts, pointing to genuine genotype-by-environment and genotype-by-pathogen interactions rather than statistical artifacts.

The centerpiece of the study, however, is its multi-population GWAS framework. Rather than either analyzing populations in isolation or naively pooling them into a single model, which would mask population-specific effects and can be biased by imbalanced sample sizes, the team fitted a multivariate mixed model that estimates correlated, population-specific marker effects simultaneously. Three complementary tests were then applied to each SNP. The average effect test asks whether the scaled effects summed across two populations deviate from zero, detecting loci with a shared effect. The differential effect test asks whether the effects differ between populations. Finally, a sign change test, adapted from recent statistical methodology, determines whether effects actually point in opposite directions, the most dangerous scenario for marker-assisted selection because breeding for the favorable allele in one population would select the wrong allele in another.

The results vindicated the approach. Pairwise multi-population analyses uncovered 21 peaks with significant average effects across population pairs, seven of which were completely invisible in the single-population analyses. Among these hidden signals were broad regions of chromosome 6H colocalizing with Rph24, a known adult plant resistance gene, demonstrating that joint modeling can recover genuine shared loci that low within-population allele frequencies or limited statistical power had concealed. The differential effect test flagged 14 peaks, twelve of which overlapped average-effect signals, indicating loci significant overall but variable in magnitude between populations. Most consequentially, five signals showed significant sign changes across populations, most likely representing three distinct QTLs whose effects reverse direction depending on genetic background, linkage phase, or epistatic context. A final four-population analysis testing for row-type and growth-type specificity found only a single SNP, within a two-component response regulator gene, whose effect differed significantly between two-rowed and six-rowed barley, and none differentiating winter from spring types, suggesting that row type plays a modest role in shaping resistance architecture while growth habit plays little detectable role at the marker level.

From these patterns the authors distilled a practical decision framework for breeders, sorting every detected QTL into three categories: loci where the same marker allele should be selected across all populations, loci where different alleles must be tracked in different populations, and loci relevant only within a single population. This taxonomy directly addresses the central anxiety of applied marker-assisted selection, namely whether a marker validated in one breeding program can be trusted in another. The study’s answer is nuanced but actionable: many resistance loci are indeed transferable, and multi-population GWAS increases power enough to reveal them, but a meaningful minority carry effects that flip sign across genetic backgrounds and would actively mislead breeders if applied indiscriminately. The authors recommend that all multi-population GWAS hits be subjected to the full battery of average effect, differential effect, and sign change tests before any breeding decision, and they caution that newly identified QTLs generally require validation and fine-mapping before conversion into diagnostic markers.

Beyond its immediate utility for barley improvement, the work carries a broader message for quantitative genetics. The finding that apparent absence of overlap between independent GWAS results often reflects insufficient power rather than genuinely distinct genetics echoes debates in human genetics, where effect sizes have proven harder to transfer across ancestries than linkage and allele frequency differences alone can explain. By explicitly modeling correlated effects across populations and testing for both shared and divergent signals, the framework offers plant and animal breeders, and potentially human geneticists, a rigorous middle path between the fragmentation of separate analyses and the homogenization of naive pooling. As rust pathogens continue their evolutionary arms race against cereal crops, tools that distinguish durable, transferable resistance from population-bound and environment-sensitive loci will only grow in importance, and this study provides both the evidence and the statistical machinery to make that distinction with confidence.

Subject of Research: Identification of common and population-specific QTLs for leaf rust resistance in barley breeding populations using multi-population genome-wide association studies.

Article Title: Discovering common and population-specific QTLs for leaf rust resistance in different Barley populations

Article References: Skovbjerg, C. K., Mahmood, K., Sarup, P., Orabi, J., Wahlström, E. M., Jensen, J. D., Olesen, L., Jensen, J., Jahoor, A., & Ramstein, G. (2026). Discovering common and population-specific QTLs for leaf rust resistance in different Barley populations. Theoretical and Applied Genetics, 139(10), Article 267. https://doi.org/10.1007/s00122-026-05375-7

Image Credits: AI Generated

DOI: 10.1007/s00122-026-05375-7

Keywords: barley, leaf rust, Puccinia hordei, GWAS, QTL, Rph genes, marker-assisted selection, quantitative genetics, plant breeding, disease resistance, multi-population GWAS, genetic architecture