**4. Wheat digestion: comparison between two different models**

To perform immunological assays on gluten peptides, it is necessary to simulate the human gastrointestinal digestion on gliadin/gluten/wheat samples. In literature, *in vitro* digestion models were amply used to study gluten peptides. But the weakness of these approaches is that they are all different and there was no standard method. Previously used models are not consistent with each other for the type of enzyme used (peptic/tryptic digests, peptic/chymo‐ tryptic, pancreatin, eventual exoproteases), for the digestion times of the gastric and intestinal phase (from 20 min to several hours), for the buffering agents used (hydrochloric acid, formic acid, bicarbonate, or phosphate buffer), for the protein:enzyme ratio used and so on. All these factors could have a strong influence on the outcome of the digestion, both in terms of pep‐ tide sequence and amount, which can be reflected also on subsequent analysis on the gluten digest. In view of this, a qualitative and quantitative comparison of the peptides generated was performed [32] applying two extremely different digestion models: a very simple peptic/tryp‐ tic‐chymotryptic digestion of a gliadin ethanol extract and a more complex and more physi‐ ological method involving the use of artificial digestive juices [33]. These juices strictly reflect the physiological composition of salts and enzymes, involving the use of α‐amylase, pepsin and pancreatin (a mixture of α‐amylase, lipase, trypsin, chymotrypsin, elasase and carboxy‐ peptidase). Peptides generated were identified using LC‐MS/MS techniques both at high and low resolution and quantified using the isotopically labelled internal standard method.

A qualitative comparison of the peptides generated with the two models is reported in **Table 2**, together with the protein of origin of the peptide and the retention time. Results clearly showed that the peptide composition obtained is completely different. While with the simplified diges‐ tion model, quite all the peptides derive from α‐gliadins; using the physiological digestion model, they are equally distributed among α‐gliadins, γ‐gliadin and low‐molecular‐weight glutenins. This fact can be ascribed to the different solubilisation power of the two methods. Ethanol extraction of the prolamin fraction probably leads to a better extractability of α‐glia‐ dins; on the opposite, in the physiological digestion model, the presence of additional enzymes other than proteases (such as α‐amylases and lipases), together with the bile salts, contrib‐ utes to matrix degradation, improving the extractability and digestibility of higher molecular weight proteins such as γ‐gliadins and low‐molecular‐weight glutenins. Another important difference among the two models is the presence of specific cleavage sites for the enzymes used. In the simplified digestion model, all the peptides show specific cleavage sites for the three enzymes used (pepsin, chymotrypsin and trypsin), for example, tyrosine, phenylalanine and leucine. In the physiological model indeed, in most cases, there are no specific cleavage sites, due to the action of the exoproteases present in pancreatin. Thus, changing the *in vitro* digestion model, the peptide profile is completely different. Using the isotopically labelled internal standard method, the peptides were quantified for both digestion models, and their total amount was plotted to obtain a quantitative comparison. Despite the different identified peptide sequences, the two *in vitro* digestion methods showed a good correlation in terms of immunotoxic sequences. This means that, despite the different amino acid sequence of the peptides generated, the immunotoxicity of a wheat variety is an own intrinsic characteristic of its gluten. In **Figure 4**, the total amount of immunogenic and toxic peptides obtained with the physiological digestion is plotted with those arising from the simplified digestion model.

**Figure 3.** Total content of toxic peptides (expressed in μg of peptide for gram of sample) in 25 samples from a Durum Panel collection (upper panel). Samples were grouped on the bases of phylogenetic affinity on dendrogram (lower

**Figure 2.** Total content of toxic peptides (expressed in μg of peptide for gram of sample) in 25 samples from a Durum Panel collection (upper panel). Samples were grouped on the bases of phylogenetic affinity on dendrogram (lower

panel). \*\* Statistically different group. Adapted with permission from Ref. [31].

panel). Adapted with permission from Ref. [31].

318 Wheat Improvement, Management and Utilization


**Table 2.** Pathogenic peptides identified in the digested samples obtained with the two different models, with the protein of origin.

**Figure 4.** Correlation between the amount of immunogenic and toxic peptides generated with the two digestive models. Peptide amounts are expressed as parts per million (ppm).

A good correlation was found using the Pearson test (*p* < 0.05, coefficient of correlation = 0.73 for immunogenic peptides and 0.61 for toxic peptides), as shown in **Figure 4**. This means that for biological experiment, the physiological systems should be suitable, because the peptides generated are similar to those that really come in contact with intestinal cells. Of course, the limit of these models is the lacking of brush border membrane enzymes (that can further proteolyse the peptides) and of the intestinal microflora. A possible interesting continuation of the work could be the use of a physiological model taking into account also these latter variables and studying the effects in terms of peptides produced. However, the good correlation between the total amount of immunogenic and toxic peptides, would suggest to use the simplified method for varietal screening or comparison purposes, where a high throughput, low cost and simple analysis is required.
