Absolute and relative quantitation of ATI in modified wheat

Amylase/trypsin-inhibitors (ATIs) are wheat grain proteins triggering health-related disorders such as Bakers’ asthma and non-celiac wheat sensitivity (NCWS) in up to 6% of the population. Gene-silencing tools aim to knock out genes responsible for expression of ATIs to reduce the immunogenic potential of wheat used for bread and pasta (common and durum wheat). These efforts are facing challenges due to the high complexity of wheat's genetic background, as common wheat is hexaploid and durum wheat is tetraploid. Thus, sensitive, accurate and precise methods are required to confirm the success of silencing of ATIs in wheat and to monitor the effects on other wheat proteins. The aim of the study was to compare different methods for the absolute and relative quantitation of ATIs.
Three ATIs of Bobwhite (common wheat) and two ATIs of Svevo (durum wheat) were silenced by RNAi and CRISPR-Cas9, respectively. Twelve different ATIs were quantitated using heavy isotopic labeled peptides and a LC-MS/MS-SRM method on a triple quadrupole system. For relative quantitation, data-dependent acquisition (DDA) was performed on a Q-Exactive+ and on a Bruker amaZon Speed Ion Trap without labeling, and iTRAQ experiments on a Q-Exactive-HF. DDA data analyzed at the ion trap were evaluated using MASCOT MS/MS Ions search and manual quantification on MS level, DDA data from the Q-Exactive+ using MaxQuant, intensity based absolute quantitation (iBAQ) and Skyline, and iTRAQ data using PEAKS. Additionally, RP-HPLC, SDS-PAGE and measurement of relative gene expression helped to verify obtained results.
Generally, high accordance of peptides for quantitation of most abundant ATIs (e.g., CM2, CM3 and CM16) was observed among used methods with different MS setups. Contrary, for less abundant ATIS like CM1, CM17 or WCI partially divergent peptides were assigned. Similar findings were observed in other studies, which applied LC-MS/MS for detection of ATIs. Absolute quantitation verified silencing of the three ATIs in the mutants of Bobwhite and that of the two ATIs in the mutants of Svevo. In Bobwhite, along with target genes three other non-target ATIs were silenced in the transgenic plants. In contrast, this was not observed in the tetraploid mutants, in which all ATIs were expressed in lower levels compared to Svevo, confirming that CRISPR-Cas9 is a more precise and efficient technology compared to RNAi and other mutagenesis techniques. The relative quantitation methods demonstrated the same results for Svevo and the tetraploid mutants. In contrast, these methods detected peptides of the silenced ATIs in the hexaploid transgenic lines at a maximum level of 5% compared to Bobwhite. Manual curation of the identified peptides using Skyline confirmed their presence in the hexaploid transgenic lines. The largest discrepancy between absolute and relative quantitation was observed for the most abundant ATI called 0.19, which is only expressed in hexaploids, but not in tetraploid wheat. The relative quantitation of 0.19 using MaxQuant and iBAQ showed that 0.19 was not present in Bobwhite and the hexaploid mutants. Further, ProteinPilot did not use the peptides of 0.19 for relative quantitation. This uncovered a major problem of wheat proteomics. The number of ATI peptides with the attribute ‘unique’ is very limited, because there are several entries for all ATIs in the UniProtKB database, e.g., seven entries for 0.19 (P01085) with 100% identity. Furthermore, the role of 0.28 inhibitor in tetraploids is unclear. On the one hand, all MS approaches detected unique peptides for this protein, but relative gene expression reveled absence of 0.28 in the durum wild type. Further analyses aimed at clarifying this puzzling result are being carried out.
In summary, this study shows the high necessity either for selective absolute methods or for the manual curation of software-based protein identification. Further, comprehensive and correct databases are required to identify and quantify peptides and proteins. Wheat proteomics is facing challenges due to the high genetic complexity, the close relationship to other cereals and the incomplete and partially inaccurate protein database requiring sensitive, precise and accurate LC-MS/MS methods.