Wheat is one of the world's most important crops, providing around 20% of human calorie and protein intake. Its popularity is attributed to its distinct dough and baking characteristics, which are essential for bread and pasta making. These unique properties are largely determined by the amount and composition of flour proteins. On the downside, many of those are associated with wheat-related diseases like allergies or intolerances. Breeding for higher yield and better resistances is of great importance in order to meet the increasing demand and to adapt to climate change. Improving quality is also of interest but limited by the negative correlation of yield with protein content, used as a measure for baking quality. Further methods are thus required to estimate protein quality beyond mere amounts.

To this end, we developed a high-throughput LC-MS-based bottom-up proteomics workflow for the analysis of wheat flour and analyzed 282 varieties grown under 4 environmental conditions (1200 samples). Proteins were extracted under denaturing and reducing conditions and transferred to microtiter plates. Semiautomatic sample preparation including protein assay, protein clean-up on magnetic beads and tryptic digestion were performed using a Beckman Biomek i7 liquid handling robot. Peptides were loaded onto Evotips and analyzed using an Evosep One (60 samples/day) coupled to a Bruker timsTOF HT (diaPASEF). Raw data were processed in DIA-NN (library-free mode). Results were filtered to a run-specific and global FDR of 1% at the precursor and protein level. Protein intensities were batch effect-corrected using the R package limma. Average protein intensities per variety, genetic and environmental variance components affecting the protein levels and heritabilities were modelled using the R package ASreml. Statistical analyses were performed using the R packages limma, rstatix and fgsea.

Overall, 6567 protein groups were quantified in the flours, >6000 with a data completeness >90% across all wheat varieties. For 3461 proteins high heritabilities (proportion of observed variation that can be attributed to genetic differences rather than environmental factors) h²>0.5 were detected, representing candidates whose levels can be effectively modified through selective breeding. Many showed significant and biologically meaningful correlations with conventional measures like loaf volume or falling number, indicating that proteomic data could indeed help to predict baking quality. A comparison with allergen databases identified 305 potentially immunogenic proteins. Of those, many showed h²>0.5 and no correlations with agronomic and quality parameters, indicating that a reduction of allergens through breeding seems to be possible.

With the advantage of the low input required and the prospect of faster LC-MS methods, our data show that proteomics has the potential to become a valuable tool in future wheat breeding and routine testing along the supply chain.