Quantitation of the immunodominant 33-mer peptide from α-gliadin in wheat flours by liquid chromatography tandem mass spectrometry

Coeliac disease (CD) is triggered by the ingestion of gluten proteins from wheat, rye, and barley. The 33-mer peptide from α2-gliadin has frequently been described as the most important CD-immunogenic sequence within gluten. However, from more than 890 published amino acid sequences of α-gliadins, only 19 sequences contain the 33-mer. In order to make a precise assessment of the importance of the 33-mer, it is necessary to elucidate which wheat species and cultivars contain the peptide and at which concentrations. This paper presents the development of a stable isotope dilution assay followed by liquid chromatography tandem mass spectrometry to quantitate the 33-mer in flours of 23 hexaploid modern and 15 old common (bread) wheat as well as two spelt cultivars. All flours contained the 33-mer peptide at levels ranging from 91–603 μg/g flour. In contrast, the 33-mer was absent (<limit of detection) from tetra- and diploid species (durum wheat, emmer, einkorn), most likely because of the absence of the D-genome, which encodes α2-gliadins. Due to the presence of the 33-mer in all common wheat and spelt flours analysed here, the special focus in the literature on this most immunodominant peptide seems to be justified.


Results
) following in intensity were used for qualification (qualifiers) as well as the MRM transitions of the 3+ charge state. The collision energy was optimised for each MRM transition to achieve the highest possible product ion intensity 22 .
The Norwegian wheat cv. MJO was used as a positive control to develop a SIDA, because it is known to contain the 33-mer 13 . Two different approaches were taken: quantitation of the 33-mer directly in chymotryptically hydrolysed wheat flour and quantitation in hydrolysed gliadins which had been extracted from the flour. The results showed that it was not possible to quantitate the 33-mer directly in hydrolysed flour, because the peptide signal was overlaid by signals originating from the flour matrix. The 33-mer showed signals with high intensity in hydrolysed gliadins and interfering matrix effects were reduced, because only one protein fraction was taken for analysis instead of the entire wheat protein present in flour. To balance out the loss of analyte during sample preparation, the isotopically labelled standard was added prior to chymotryptic digestion of the gliadins.
Resistance of the 33-mer to enzymatic hydrolysis. Preliminary experiments using a combination of pepsin and trypsin/chymotrypsin (PTC) for enzymatic hydrolysis were performed with the 33-mer. This PTC hydrolysate of the 33-mer was analysed by untargeted LC-MS/MS followed by data evaluation using the MS/MS ions search module of the Mascot software based on the NCBI database (National Library of Medicine, Bethesda, MD, USA). In addition to the original 33-mer, the truncated forms after N-terminal removal of leucine resulting in a 32-mer (m/z 950.0, 4+ and m/z 1266.3, 3+ ) and pyro-32-mer (m/z 945.7, 4+ and m/z 1260.6, 3+ ) were identified, but no 30-mer (m/z 889.7, 4+ and m/z 1185.9, 3+ ) after potential removal of N-terminal LQL. The ratio 33-mer/32-mer/pyro-32-mer was about 12/76/12. To see whether cleavage of the N-terminal leucine could be minimised, only chymotrypsin was used for further experiments. Therefore, the 33-mer peptide was incubated with chymotrypsin using exactly the same conditions as described for the samples to check its resistance towards cleavage, because it contains two potential chymotryptic cleavage sites at positions L1 and L3. In this case, untargeted LC-MS/MS revealed no truncated forms. To ascertain this, full MS/MS scans looking only for the above m/z values were acquired and these showed no detectable amounts of 32-mer, pyro-32-mer or 30-mer, only the intact 33-mer. Corresponding experiments, specifically looking for the potential isotopically labelled *32-mer (m/z 954.0, 4+ and m/z 1271.3, 3+ ), pyro-*32-mer (m/z 949.7, 4+ and m/z 1265.6, 3+ ) and *30-mer (m/z 893.4, 4+ and m/z 1190.9, 3+ ), confirmed that the *33-mer standard was also stable under the conditions applied. Full MS/MS scans were also done for the chymotryptic hydrolysates of the albumin/globulin and gliadin fractions obtained from wheat cv. MJO. Again, only the intact 33-mer was detected in the gliadin hydrolysate, but no detectable traces of 33-mer in the albumin/globulin hydrolysate. These experiments confirmed that no truncated forms of the 33-mer were generated during chymotryptic hydrolysis using the applied conditions. This is consistent with the original report by Shan et al. 4 who only observed pepsin-catalysed cleavage of the N-terminal leucine. Additionally, no analyte loss is expected to take place during sample preparation, which involves the removal of albumins/globulins.
Calibration and quantitation. The response factor of the 33-mer peptide was determined using the peak area ratio A (*33-mer)/A (33-mer) at different values of n (*33-mer)/n (33-mer) between 0.02 and 9.2, that lay within the linear range. Quantitative 1 H nuclear magnetic resonance spectroscopy ( 1 H qNMR) was used to determine the exact concentrations of the methanolic solutions of the 33-mer (1.90 μ mol/mL) and the *33-mer standard (1.81 μ mol/mL). The characteristic signals located in the aromatic field (δ /ppm: 6.5-8) of the two phenylalanine residues (5 protons each) and the three tyrosine residues (4 protons each) were integrated (22 protons in total) and compared to a reference solution containing L-tyrosine 23 . The area ratio A (*33-mer)/A (33-mer)    Optimisation of sample preparation. Studies by Fiedler et al. 25 demonstrated that protein digestion was improved with reduction/alkylation in the first step followed by digestion in the second step, because this approach resulted in a higher number of identified peptides. Therefore, reduction of disulphide bonds with tris-(2-carboxyethyl)phosphine (TCEP) followed by iodoacetamide (IDAM) alkylation of liberated free cysteine residues of the gliadin fraction from wheat cv. Akteur (A13) were performed to see if the quantitated amount of 33-mer was influenced compared to native hydrolysed gliadins. After reduction and alkylation, the amount of 33-mer was 6.3 ± 0.4 mg/g gliadin compared to 6.2 ± 0.4 mg/g of native gliadin. The values showed no significant difference (p = 0.705). α -Gliadins typically contain six cysteine residues and form three intrachain disulphide bonds located in the C-terminal domain consisting of sections III, IV, and V 26 . The 33-mer is located within section I and represents a part of the N-terminal domain 27 . After cleavage of disulphide bonds following reduction, the N-terminal domain containing the 33-mer was apparently not affected regarding accessibility to enzymatic attack, which resulted in no significant change of 33-mer contents. To simplify sample preparation, the reduction/ alkylation step was omitted. After chymotryptic hydrolysis of the gliadins extracted from wheat cv. A13, the obtained peptide mixture was purified by solid phase extraction (SPE) using C 18 -cartridges in order to reduce matrix effects. The 33-mer content was 6.3 ± 0.3 mg/g gliadin after purification in comparison to 6.2 ± 0.4 mg/g gliadin without purification of the hydrolysate. The quantitative values were not significantly different (p = 0.506). The only impact was a higher signal intensity (by a factor of 2) of both 33-mer and *33-mer after SPE purification, which did not influence the quantitated amount of 33-mer, because the ratio of analyte to standard did not change. Having ascertained that purification did not influence the content of the 33-mer, the gliadin hydrolysates were analysed by targeted LC-MS/MS without SPE to speed up sample preparation.

Limit of detection (LOD) and limit of quantitation (LOQ). The LOD and LOQ of the MS method
to quantitate the 33-mer were determined according to Vogelgesang and Haedrich 28 . The analyte was spiked in seven different concentrations between 0.1 and 200 μ g/g to rye prolamins as matrix, which did not contain the 33-mer peptide. The absence of the 33-mer in rye flour (cv. Visello) had been confirmed by LC-MS/MS of the hydrolysed flour and prolamin fraction. The 33-mer was identified with high sensitivity resulting in an LOD of 13.1 μ g/g rye flour and an LOQ of 47.0 μ g/g rye flour.
Analysis of the 33-mer content in different common wheat and spelt cultivars. The quantitative determination of the 33-mer was performed in flours of 23 hexaploid modern and 15 old common wheat cultivars from different harvest years and two spelt cultivars harvested in 2014 (Table 1). In this context, old common wheat is defined as a cultivar from T. aestivum with its year of first registration prior to 1950. All flours were characterised including determination of crude protein contents according to ICC Standard No. 167 29 and quantitation of α -gliadins, gliadins and glutenins after modified Osborne fractionation combined with RP-HPLC as reported by Wieser et al. 30 . Total gluten contents were calculated as sum of gliadin and glutenin contents (see Supplementary Table S1).
The 33-mer was present in all common wheat and spelt flours in a range from 90.9 to 602.6 μ g/g of flour (Fig. 2a). The modern wheat Y14 had the highest amount of 33-mer (602.6 μ g/g flour), that was significantly different to all analysed cultivars with the exception of Z14 (see Supplementary Table S2). The old wheat ABD with the lowest 33-mer content of 90.9 μ g/g flour differed significantly to all wheat and spelt cultivars. In contrast, the old wheat SLD (528.0 μ g/g flour) contained one of the highest 33-mer amounts of the analysed flours and showed no significant difference to the modern wheats A13, A14, D05, Z14, V12, and MJO and spelt OBE. Special attention was directed to MJO, because the 33-mer was first identified in this cultivar 13  19.0 488.9 (y4) 3 26 713.5 (y6) 3 14 973.5 (y8) 3  19.0 510.3 (y4) 3 26 735.2 (y6) 3 14 996.0 (y8) 3 12 Table 2. Multiple reaction monitoring (MRM) parameters of the 33-mer and the isotopically labelled *33mer peptides. 1 Charge state: 1+ . 2 Precursor to product ion transitions were used as quantifier. 3 Precursor to product ion transitions were used as qualifier.
of 368 ± 109 μ g/g flour. As a result, only some differences in 33-mer contents between these wheat cultivars were significant. A certain trend, e.g., that modern wheat cultivars generally contain higher amounts of 33-mer than old cultivars could not be derived from the data. Considering the amounts of 33-mer in the two spelt cultivars, it was noticeable that OBE contained one of the highest amounts of the 33-mer peptide (523.4 μ g/g flour). The content of 33-mer in FRK (353.9 μ g/g flour) was in the range of 200-400 μ g/g flour, and did not differ significantly from the common wheat cultivars. The 33-mer contents of all analysed flours were also calculated based on the amount of α -gliadins ( Fig. 2b) determined after modified Osborne fractionation by RP-HPLC 30 . MJO had the highest content of 33-mer in α -gliadin (23.2 mg/g α -gliadin). It was significantly different to all other cultivars (see Supplementary Table S3) and was caused by the high 33-mer content and the low amount of α -gliadins (2.2%) in flour. SLD and M14 had similar 33-mer contents (17-18 mg/g α -gliadin) and differed significantly to all other varieties. ABD had the lowest amount of 33-mer in α -gliadin (4.1 mg/g α -gliadin) and did not show significant differences to CAY, CAR, GPS, GSW, and GEF, but differed statistically to the other cultivars. The overall average content was 11.7 ± 3.1 mg/g α -gliadin.
Many studies in the literature have focused on peptide quantitation in enzymatically hydrolysed prolamin extracts 25,31 , hydrolysed gluten extracts 32 or hydrolysed wheat flours 22 , but the putative immunodominant 33-mer was not quantified. Because of the missing data for 33-mer contents, it was difficult to compare the peptide contents to existing data. Only studies by van den Broeck et al. reported the quantitation of the 33-mer using LC-MS with external calibration, but not SIDA. Peptide concentrations were converted into the corresponding contents of α -gliadin per microgram digested gluten protein extract using the average mass of 32,285.5 of α -gliadins. The 33-mer contents determined for two wheat cultivars corresponded to 10.3 and 5.8 mg/g of α -gliadin 33 , which agreed well with the data in Fig. 2b. Correlations and principal component analysis. The 33-mer contents of the 51 modern and old common wheat and spelt cultivars (based on flour) were correlated to the contents of α -gliadin, total gliadin and total gluten analysed by RP-HPLC after modified Osborne fractionation and to crude protein contents (see Supplementary Table S1). A weak correlation (r = 0.568, p < 0.001) was observed between 33-mer and α -gliadin contents, but there was no correlation to gliadin contents (r = 0.469, p < 0.001), gluten contents (r = 0.526, p < 0.001) or crude protein contents (r = 0.481, p < 0.001) (Fig. 3).
Principal component analysis (PCA) with 33-mer, α -gliadin, gliadin, gluten, and crude protein contents of the 49 common wheat and 2 spelt flours was performed to assess whether these variables could be used to differentiate between spelt, modern common wheat, and old common wheat cultivars (Fig. 4). Both principal components together accounted for 94.9% of data variability. Component 1 was positively correlated with 33-mer  Table 1)) and two spelt cultivars were analysed. Wheat cultivars registered prior to 1950 were designated as old. For abbreviations of the cultivars, see Table 1.
contents (r = 0.643), but even more so with α -gliadin, gliadin, gluten, and crude protein contents (r ≥ 0.920). In contrast, component 2 was only positively correlated with 33-mer contents (r = 0.765), but negatively associated  Table 1)) and two spelt cultivars were analysed. Wheat cultivars registered prior to 1950 were designated as old.  Table 1)) and two spelt cultivars were analysed. Wheat cultivars registered prior to 1950 were designated as old.
with α -gliadin, gliadin, gluten, and crude protein contents (r ≤ − 0.061). The vector indicating the contribution of the content of 33-mer was downsized for visibility reasons, but it pointed to 3.0/8.7 as x-and y-coordinates. The contents of α -gliadin, gliadin, gluten, and crude protein all had strong positive correlations in all possible pairwise combinations (p ≥ 0.850). PCA essentially confirmed the results of the correlation analyses that had already shown the 33-mer contents to be mostly unrelated to α -gliadin, gliadin, gluten, and crude protein contents. Cv. MJO was placed in the top left corner, because of its high content of 33-mer, but comparatively low contents of gluten proteins, especially α -gliadins. In comparison, cv. Y14 appeared on the far right, because of high contents of all five variables whereas cv. CAR was located in the bottom right corner, because of high (gluten) protein contents, but a comparatively low 33-mer content (229.4 μ g/g flour). In addition, PCA revealed that these five variables were unsuitable to differentiate between spelt, modern common wheat, and old common wheat cultivars. The five old common wheat cv. ABD, BED, RFB, RPD, and STD were placed on the far left, but the other ten old cultivars were located right in the middle at similar coordinates as the modern common wheat cultivars. The two spelt cv. FRK and OBE were situated next to the common wheat cultivars. Therefore, the hypothesis that spelt may be less CD-immunoreactive than modern common wheat cultivars could not be confirmed. This finding is in accordance with Ribeiro et al. 34 who compared modern common wheat to spelt cultivars and showed that spelt cultivars had a higher amount of toxic epitopes than common wheats. Analysis of durum wheat, emmer and einkorn. The 33-mer peptide was also analysed in two durum wheat and two emmer cultivars (genome AABB) as well as two diploid einkorn cultivars (genome AA) ( Table 1). In each of these wheat species, the 33-mer was not detected (< LOD). In comparison to hexaploid common wheat, durum wheat, emmer, and einkorn do not contain the D-genome, which originated from hybridisation of T. turgidum dicoccum (genome AABB) with Aegilops tauschii (genome DD) 36 . The absence of the 33-mer peptide can be explained by the fact that this peptide is encoded by genes located in the Gli-2 locus on chromosome 6D, which is missing in durum wheat, emmer, and einkorn. Studies by Molberg et al. showed clear variations in intestinal T-cell responses between common wheat and tetra-or diploid species due to different degrees of T-cell immunoreactivity between the gluten proteins encoded on the A-, B-, and D-genome. Einkorn cultivars were only recognized by DQ2.5-glia-α 1a-specific T-cell clones, but not by DQ2.5-glia-α 1b-and DQ2.5-glia-α 2-specific T-cell clones. Emmer and durum wheat cultivars were all recognized by DQ2.5-glia-α 1a-specific T-cell clones, but only two out of four emmer cultivars and three out of ten durum wheat cultivars activated DQ2.5-glia-α 1band DQ2.5-glia-α 2-specific T-cell clones 37 . Consistent with our results, Prandi et al. 38 found that the 33-mer was not present in durum wheat. As a consequence, this peptide was used as a marker peptide to identify the presence of common wheat in durum wheat flours. One durum wheat cultivar was also analysed by van den Broeck et al. 33 and the 33-mer peptide was not detected either.

Discussion
The present study is the first to establish a SIDA combined with targeted LC-MS/MS for the quantitative determination of the immunodominant 33-mer peptide in wheat flours. Due to the use of a stable-isotope-labelled *33-mer standard, sample preparation could be simplified without reduction/alkylation and SPE purification.
Although the UniProtKB database had only 19 out of 897 entries for α -gliadin sequences from Triticinae containing the 33-mer with an identity of 100%, all 40 analysed modern and old common wheat and spelt cultivars contained the immunodominat 33-mer peptide (51 flour samples in total, because several flours were available from different harvest years). The focus on this peptide seems to be legitimated not only because of its unique structure containing six copies of three overlapping T-cell epitopes, but also because of its presence in all hexaploid wheat cultivars analysed in this study. PCA analysis of the data demonstrated that the contents of 33-mer were not suitable to differentiate old from modern common wheat cultivars, because cultivars with high 33-mer contents were found within both flour sets. The 33-mer was not detected in two cultivars each of tetraploid emmer and durum wheat as well as diploid einkorn, which do not contain the D-genome. This observation may be explained by the fact that the 33-mer is encoded on the Gli-2 locus on chromosome 6D, but a larger set of durum wheat, emmer and einkorn cultivars would have to be analysed to conclude whether these wheat species generally lack the 33-mer peptide. Further work will focus on correlating the content of 33-mer analysed by LC-MS/MS with the gluten content determined by ELISA using the G12 monoclonal antibody.
Grain samples. Grains of 23 modern and 15 old (year of first registration before 1950) common wheat cultivars from different harvest years grown worldwide, and one rye cultivar (cv. Visello, harvested 2013, donated by KWS Lochow, Bergen, Germany) were either obtained as flours or milled on a Quadrumat Junior mill (Brabender, Duisburg, Germany) and sieved to a particle size of 0.2 mm. Two spelt, two durum wheat, two emmer, and two einkorn cultivars, grown in Germany (harvested 2014) were milled on a Laboratory 3100 cross beater mill (Perten Instruments, Hamburg, Germany) to wholemeal flours. In total, 57 samples were analysed, because some modern wheat cultivars were available from two to three different harvest years or two cultivation regions ( Table 1).
Characterization of the flours. Crude protein content. The crude protein content (nitrogen content × 5.7) of the flours was determined by the Dumas combustion method according to ICC Standard Method 167 29 using a TruSpec Nitrogen Analyzer (Leco, Kirchheim, Germany).
LOD and LOQ of the MS method. The LOD and LOQ of the quantitation method for the 33-mer peptide were determined. Rye flour (cv. Visello, harvest year 2013) was used as blank, because it was very similar to wheat regarding the gluten protein fractions, but did not contain α -gliadins. The prolamin extraction procedure and chymotryptic hydrolysis were performed as described above. To determine the LOD and LOQ of the targeted LC-MS/MS method, the prolamin extract was spiked at 7 different concentrations (0.1-200 mg/kg) of 33-mer peptide and the samples were hydrolysed by α -chymotrypsin followed by targeted LC-MS/MS analysis. The LOD and LOQ were derived statistically from the data 28 . The LOD was calculated based on a signal-to noise-ratio (S/N) of 3, and the LOQ on an S/N of 10.
Statistics. Statistically significant differences between 33-mer contents of different modern and old wheat cultivars and two spelt cultivars were determined by one-way analysis of variance (ANOVA) with Tukey's test as all pairwise multiple comparison procedure at a significance level of p < 0.05 using SigmaPlot 12.0 (Systat Software, San José, CA, USA). The significance of differences between 33-mer contents of the cv. Mv Magvas, Mv Mazurka, Mv Verbunkos, and Yumai-34 harvested in 2011, 2012, and 2014 were analysed by two-way ANOVA accordingly with harvest year and cultivar as factors. Pearson's product moment correlations were calculated between contents of 33-mer and α -gliadins, gliadins, gluten or crude protein for all analysed wheat and spelt cultivars. Correlation coefficients (r) were defined according to Thanhaeuser et al. 43 (r > 0.78, strong correlation; 0.67-0.78, medium correlation; 0.54-0.66, weak correlation; r < 0.54, no correlation). PCA was carried out with XLStat 2016 (Addinsoft, New York, NY, USA) to determine if the contents of 33-mer, α -gliadin, gliadin, gluten, and crude protein could be used to differentiate between spelt, modern common wheat, and old common wheat cultivars.