Improving wheat grain composition for human health by constructing a QTL atlas for essential minerals

Mineral malnutrition, particularly deficiencies of iron (Fe) and zinc (Zn), remains widespread globally and is reflected in national dietary surveys (e.g., UK NDNS). Wheat is the most widely consumed crop and provides a substantial portion of dietary minerals; therefore, enhancing wheat grain mineral content can benefit public health. Mineral content in wheat grain depends on soil availability and plant uptake/partitioning. Two strategies exist: agronomic biofortification (fertilisation) and genetic biofortification (breeding). While agronomic approaches can be effective, they add cost and may be inaccessible in low-resource settings. Genetic biofortification exploits natural genetic variation; however, modern cultivars have reduced diversity compared to landraces. The A.E. Watkins landrace collection captures broad diversity and was used here to map quantitative trait loci (QTLs) affecting essential minerals. The study focuses on Fe, Zn, calcium (Ca), magnesium (Mg), potassium (K), and copper (Cu), aiming to identify consistent QTLs with marker support and candidate genes to enable breeding of nutrient-enriched wheat, with traits assessed both as concentration and as content per grain to ensure suitability for high-yield backgrounds.

Prior studies report substantial genetic and environmental variation for Fe and Zn in wheat grain, with approximately twofold differences among genotypes (e.g., Fe ~26–56 mg/kg, Zn ~13–35 mg/kg across various panels and environments). CIMMYT successfully developed high-Zn wheat with 8–10 mg/kg (25–40%) Zn increases while retaining yield, illustrating feasibility of genetic biofortification. Numerous QTLs and GWAS signals for Fe and Zn have been detected across all wheat chromosomes, with notable clusters previously reported on group 2 and 7 chromosomes; however, correspondence between studies requires detailed genomic comparison. Reports on Mg highlight broad variability and strong genotype effects, suggesting potential for genetic improvement, though major Mg QTLs have been less documented. For Ca, previous GWAS identified a strong locus on 5A among others. The background literature also recognizes yield dilution effects (negative correlation between yield/TGW and mineral concentrations) and positive correlations between Fe and Zn due to shared transport and remobilization processes.

Genetic materials: Three biparental RIL populations (94 F4 RILs each) were developed from crosses between the spring wheat cultivar Paragon and A.E. Watkins landraces W160, W239, and W292. W292 originated from Cyprus (ancestral group C7), W160 and W239 from Spain (group C6). Field experiments: Trials were conducted at Rothamsted Farm, Harpenden, UK (51.80 N, 0.40 W) between 2012 and 2020. Each population was grown for three years in three replicate randomized plots (1×1 m), under one (N2 = 200 kg N/ha) or two N rates (N1 = 50 or 100 kg/ha; N2 = 200 kg N/ha). Split-plot design with blocks split for N treatment; seed rate 350 seeds/m²; rainfed. Soil mineral N (0–90 cm) measured annually. N applied as ammonium sulphate/nitrate per protocol; no P or K fertiliser applied; soil P and K maintained non-limiting. Genotyping and maps: Par × W292 was genotyped with the Axiom 35K Wheat Breeder array; Par × W160 and Par × W239 with the Axiom 44K TaNG array. Genetic maps were constructed using ASMap in R following established procedures. Phenotyping: Grain and straw yields per plot were measured; dry matter determined after oven drying. Thousand grain weight (TGW) was recorded. Grain and straw mineral contents were quantified after nitric/perchloric acid digestion by ICP-OES (Agilent 5900 SVDV), with certified standards and QC (Shewhart control charts). Primary traits for QTL analysis were mineral concentration in grain (mg/kg DW) and mineral content per grain (µg/grain). Additional derived traits included off-take (total mineral per m²), grain mineral deviation (concentration adjusted for yield via regression to account for yield dilution), biomass mineral content, and harvest index. Statistics: Within-population genotype effects evaluated by one-way or two-way ANOVA (when N varied) with n = 3 replicates per genotype per environment. Pearson correlations among mineral and yield-related traits were computed (R Hmisc and corrplot). Broad-sense heritability calculated with the heritability package. QTL mapping: Conducted in R/qtl. Genome-wide LOD thresholds were established via permutation tests (alpha = 0.05), typically ~3.0–4.0. To prioritize robust loci for breeding, only QTLs detected in at least two sample sets with LOD >5.0 in at least one set were advanced (one noted exception). Peak markers, confidence intervals, additive effects, and variance explained were recorded. Co-location with straw/biomass mineral QTLs was noted to infer uptake vs partitioning effects. QTLs were anchored to IWGSC RefSeq v1.0. Candidate gene analysis: For selected QTLs (highest LOD per mineral), genes within ±5 Mb of the peak marker were catalogued using Ensembl BioMarts and annotated via WheatOmics and KnetMiner. Whole-genome sequence comparisons between Paragon and landraces assessed SNPs and structural variation. Functional validation: For the major Ca 5A QTL, TILLING mutants in cultivar Cadenza with non-synonymous or truncation variants in candidate genes TraesCS5A02G543300 (cation transporter/plasma membrane ATPase) and TraesCS5A02G542600 (MFS transporter) were grown in the glasshouse. Only homozygous mutants (validated by KASP) were analysed. Grain Ca concentration was measured by XRF calibrated to ICP-OES (R²=0.823, Pearson R=0.907, p<0.001) and compared to WT by ANOVA. Grain traits (GN, GY/plant, grain dimensions) were recorded.

- Surveyed mineral ranges across populations/environments: Ca 273–1532 mg/kg; Mg 834–1532 mg/kg; K 3541–5935 mg/kg; Fe ~30–64 mg/kg; Zn 23.6–49.1 mg/kg; Cu 3.61–7.24 mg/kg (Table 1). Negative correlations of mineral concentrations with yield and TGW were generally weak (r > −0.5), but Zn concentration commonly showed negative correlation with yield. Positive Fe–Zn correlations were consistent (e.g., r = 0.591–0.779 in Par × W160); Ca–Mg correlations were moderate in Par × W239 and Par × W292 (r ≈ 0.29–0.66 depending on environment). - Robust QTL set: 23 increasing alleles for grain minerals met the criteria (detected in ≥2 sample sets, LOD >5 in ≥1), with 16 increasing alleles from Watkins landraces and 7 from Paragon. QTLs spanned 14 of 21 chromosomes, with clusters on 5A (4 QTLs), 6A (3), and 7A (3). Some loci co-located across minerals (e.g., Mg/Zn on 6A; Ca/Zn/K on 5A) and with straw/biomass mineral QTLs, implicating uptake vs partitioning. - Calcium (Ca): Strongest locus across the study on 5A from W239/W292, affecting Ca concentration and Ca/grain across eight sample sets (LOD 6.1–12.2). Candidate region (±5 Mb) contained 127 genes; functional validation via TILLING showed five independent mutants in TraesCS5A02G543300 significantly increased grain Ca (>10%) versus WT, with three lines showing no change in grain weight, supporting causality. No effect was detected for TraesCS5A02G542600 mutants. Additional Ca QTLs: 4A (W292), 5D (W160), and 2B (Paragon). Example statistics (peak set): 5A Ca concentration additive effect 31.81, variance explained 40.4%; Ca/grain variance explained 40.0%. - Copper (Cu): Five QTLs identified. Increasing alleles from Paragon on 4B (Cu concentration; LOD 5.1–8.8 across four W239 sample sets), 5B (Cu concentration and Cu/grain; LOD up to 6.7), and 7D; from W292 on 7A (Cu concentration and Cu/grain); from W239 on 7B (Cu concentration and Cu/grain). Candidate genes in 4B region included three ZIFL-like genes (TraesCS4B02G131400/131500/131700) and a MATE transporter (TraesCS4B02G128600). On 5B, a SNP in the 3′UTR of an ABC transporter C subfamily gene (TraesCS5B02G479900) was detected in W239. Homoeology analysis showed 7A/7B/7D Cu QTLs are not homoeologous. Example variance explained: 4B Cu conc 20.1%; 7A Cu/grain 23.9%; 7D Cu/grain 23.5%. - Iron (Fe): Four QTLs with increasing alleles from Watkins landraces. Strongest on 2D (W239) for Fe concentration and Fe/grain (LOD up to 6.3). Additional loci: 3A (W239/W292), 5D (W160/W292), 6A (W239 for conc; W160 for Fe/grain). The 2D ±5 Mb region contained 120 genes with high allelic diversity and large deletions in Paragon, suggesting divergent haplotypes/introgression. Example variance explained: Fe/grain (2D) 27.8%; Fe conc (5D) 27.5%. - Magnesium (Mg): QTLs detected on 5A and 6A (increasing alleles from W239 and W160), 6D (Paragon; Mg concentration LOD up to 8.3), and a consistent Mg QTL on 7A across populations (increasing alleles from all three landraces; Mg/grain LOD up to 7.8). The 7A candidate region (±5 Mb) contained 128 genes with multiple predicted functional variants, including stop-gain SNPs in serine/threonine receptor kinase (TraesCS7A02G126100) and MYB-related TFs (TraesCS7A02G135200/135400), a stop-loss in a leucine-rich repeat RLK (TraesCS7A02G130400), and splice-region SNPs in an ATP-dependent zinc metalloprotease (TraesCS7A02G128400) and a MATE transporter (TraesCS7A02G131500). Example variance explained: 6D Mg conc 28.2%; 6A Mg/grain 29.8%; 7A Mg conc 22.7%. - Potassium (K): Three QTLs in Par × W239: K concentration on 3D (Paragon increasing allele) and 5A (Paragon), and K/grain on 4B (W239; strongest with LOD 5.6–8.2 across four sample sets). The 4B region was gene-sparse (29 genes within ±5 Mb), with low SNP density but evidence of copy number variation between parents. Example variance explained: 4B K/grain 34.4%; 5A K conc 21.7%. - Zinc (Zn): Three QTLs, all with increasing alleles from landraces. On 7A, Zn concentration and Zn/grain QTLs from all three landraces (highest LOD ≈5). On 5A, Zn/grain from W292 (LOD 5) and W239 (LOD 3.6). On 6A, Zn concentration and Zn/grain from W239 (LOD 4.1–4.9) and W160 (Zn/grain LOD 3.6–3.7). The 7A region (±5 Mb; 91 genes) showed structural variation between Paragon and W239; predicted functional SNPs in a calmodulin gene (TraesCS7A02G435500) and absence of a bHLH TF (TraesCS7A02G435800) in Paragon. Example variance explained: 6A Zn conc 24.3%; 6A Zn/grain 30.0%; 7A Zn conc 18.7%. - Overall, increasing alleles for Cu, Ca, Mg, and K were also found in Paragon, suggesting exploitable variation within elite germplasm, while Fe and Zn increases were predominantly from landraces.

The study demonstrates that substantial, reproducible genetic variation for essential mineral accumulation in wheat grain exists within crosses between elite Paragon and diverse Watkins landraces. By focusing on QTLs detected across multiple environments and expressing traits as both concentration and content per grain, the identified loci should be more resilient to yield dilution and thus suitable for deployment in modern breeding programs. Positive Fe–Zn correlations and instances of QTL co-location (e.g., on 6A and 5A) suggest shared transport or remobilization mechanisms; accordingly, selection for one mineral may co-select for others. The validated Ca 5A locus, with TraesCS5A02G543300 as the causal gene, provides a concrete target for marker-assisted selection or allele mining to enhance Ca without penalizing grain size. The discovery of multiple Cu QTLs, including strong Paragon alleles, and non-homoeologous positions on 7A/7B/7D highlights complex genetic architectures and the potential for stacking. Similarly, the robust Mg 7A locus and high-variance 6D locus (Paragon) broaden opportunities for improving Mg, a mineral linked to multiple health outcomes. While Fe and Zn enhancement remains a priority for human nutrition, their bioavailability is constrained by localization and phytate binding; nonetheless, loci affecting grain content constitute important components of broader strategies that may include processing or bioavailability traits. Collectively, the QTL atlas with associated SNP markers and candidate genes equips breeders to improve grain mineral quality alongside agronomic performance.

This work provides a robust QTL atlas for six essential minerals (Fe, Zn, Ca, Mg, K, Cu) in bread wheat by leveraging replicated multi-environment trials of three Paragon × Watkins populations. Twenty-three consistent QTLs were identified across 14 chromosomes, with clusters on 5A, 6A, and 7A. A major Ca locus on 5A was functionally validated, pinpointing TraesCS5A02G543300 as the causal gene. Candidate gene sets were defined within ±5 Mb windows for key loci, highlighting transporters and regulatory genes and revealing extensive allelic and structural variation between elite and landrace haplotypes. The results provide immediately usable markers and biological targets to improve grain mineral content in breeding programs. Future directions include: validating additional candidate genes via functional genomics; fine-mapping to narrow confidence intervals and resolve causal variants; pyramiding favorable alleles across minerals while monitoring trade-offs with yield and quality; testing QTL effects across diverse environments and soils to assess stability; and integrating bioavailability traits (e.g., phytate, grain tissue localization) and processing strategies to translate content gains into nutritional impact.

- Environmental scope: Trials were conducted at a single research site (Rothamsted) across multiple years; genotype-by-environment interactions across broader edaphic and climatic conditions were not comprehensively assessed. - Trait scope: Primary focus was on mineral concentration and content per grain; mineral bioavailability and sub-grain localization were not directly measured in these populations. - Mapping resolution: Some QTL confidence intervals are broad; candidate gene surveys were restricted to ±5 Mb around peak markers, potentially excluding causal genes outside this window. - Population scope: Three biparental populations (94 F4 RILs each) capture limited allelic diversity relative to broader panels; some loci with LOD just below thresholds were excluded by stringent selection criteria. - Analytical constraints: Correlations suggest yield dilution effects, which may complicate translation of concentration-based QTLs into high-yield backgrounds without evaluating content per grain (addressed here but warrants further validation).