Ancestral Wheat Types Release Fewer Celiac Disease Related T Cell Epitopes than Common Wheat upon Ex Vivo Human Gastrointestinal Digestion

Celiac disease (CeD) is an autoimmune enteropathy triggered by immunogenic gluten peptides released during the gastrointestinal digestion of wheat. Our aim was to identify T cell epitope-containing peptides after ex vivo digestion of ancestral (einkorn, spelt and emmer) and common (hexaploid) wheat (Fram, Bastian, Børsum and Mirakel) using human gastrointestinal juices. Wheat porridge was digested using a static ex vivo model. Peptides released after 240 min of digestion were analyzed by liquid chromatography coupled to high-resolution mass spectrometry (HPLC-ESI MS/MS). Ex vivo digestion released fewer T cell epitope-containing peptides from the ancestral wheat varieties (einkorn (n = 38), spelt (n = 45) and emmer (n = 68)) compared to the common wheat varieties (Fram (n = 72), Børsum (n = 99), Bastian (n = 155) and Mirakel (n = 144)). Neither the immunodominant 33mer and 25mer α-gliadin peptides, nor the 26mer γ-gliadin peptide, were found in any of the digested wheat types. In conclusion, human digestive juice was able to digest the 33mer and 25mer α-gliadin, and the 26mer γ-gliadin derived peptides, while their fragments still contained naive T cell reactive epitopes. Although ancestral wheat released fewer immunogenic peptides after human digestion ex vivo, they are still highly toxic to celiac patients. More general use of these ancient wheat variants may, nevertheless, reduce CeD incidence.


Introduction
Wheat proteins can trigger hypersensitivity reactions such as allergy or intolerance. Celiac disease (CeD) is an autoimmune hypersensitivity reaction induced by wheat gliadins in genetically susceptible individuals. Population screening has revealed that the prevalence of CeD is 1-2% in Europe and the United States [1,2], although many patients remain undiagnosed [3]. In addition to wheat gluten (gliadin and glutenin), CeD patients react to structurally related gluten proteins in rye (secalins), barley (hordeins) and in extremely rare cases from oat (avenins) [4]. The gluten-induced intestinal hereafter referred to as "wheat types". The harvested wheat was dried to below 15% moisture at 30 • C for 3 days before threshing and cleaning (Perten Instruments AB, Hägersten, Sweden).

Wheat Characterization
Einkorn, emmer and spelt were hulled manually after threshing. The samples were milled to whole meal flour by Falling Number Laboratory 3100 with a 0.8 mm screen (Perten Instruments AB, Hägersten, Sweden) before further analysis. Kernel size was recorded as weight per thousand kernels (TKW). Grains were counted by an Elmor C1 seed counter (Elmor Ltd., Schwyz, Switzerland), and presented as weight in grams per thousand grains. The moisture content of the grain was determined by drying kernels for 24 h at 105 • C. Further, the nitrogen content of the wheat samples was measured by the micro Kjeldahl method (Kjeltec™ 8400, Tecator, Foss, Hillerød, Denmark), and wheat protein content was determined using 5.7 as the Kjeldahl factor. Total starch content was analyzed by using the Megazyme kit (K-TSTA-100A 08/19, Megazyme, Bray, Ireland) [20]. The porridge was prepared by mixing whole wheat flour and water (1:20 w/v), which was then heated at 100 • C in a water bath for 10-15 min, homogenized, cooled and stored at 4 • C until ex vivo digested.

Ex vivo Digestion of Wheat Porridge
Human gastric and duodenal juices were collected according to Ulleberg et al. [21] by aspiration of self-reported healthy volunteers (n = 20) at Lovisenberg Diaconal Hospital, Norway. All subjects reported no CeD symptoms and gave their informed consent for inclusion before participation. The aspiration was approved by the Regional Committees for Medical and Health Research Ethics (REK 2012/2230 and 2012/2210) in Norway. In short, a flexible three-lumen silicone tube was placed through the nose or mouth into the gastric antrum and duodenum, using gastroscopic guidance. An isotonic stimulatory solution (17.5 g/L sucrose, 450 mg/L NaCl, 800 mg/L L-phenylalanine and 575 mg/L L-valine in H 2 O) was continuously infused (100 mL/h) simultaneously as the gastric and duodenal fluids were aspirated. The aspirates were pooled and stored at −20 • C, then at −80 • C [21].
The enzymatic activity of pepsin and trypsin was assayed according to Minekus et al. [22]. Digestion with human GI enzymes was performed according to the standardized INFOGEST consensus model [22] with some modifications. A porridge aliquot (1 g with approximately 5 mg/mL protein) was mixed 1:1 (w/v) with salivary fluid (SSF) containing α-amylase (75 U/mL, Sigma Aldrich) and incubated for 2 min, simulating the oral phase. The gastric digestion phase was performed by adding simulated gastric fluid (SGF) with human gastric juices (HGJ) (2000 U/mL pepsin activity) to the oral phase (1:1, v/v) and adjusting the pH to 3.0 by the addition of 1 M HCl. The samples were incubated in a water bath at 37 • C with gentle magnetic stirring for 120 min. The duodenal digestion phase was done by adding simulated intestinal fluid (SIF) containing human duodenal juice (HDJ) (100 U/mL trypsin activity) to the gastric sample (1:1 v/v). The pH was adjusted to 7.0 by the addition of 1 M NaOH and the samples were incubated in a water bath at 37 • C for another 120 min with magnetic stirring, then terminated by adding 5 mM Pefabloc ® (Sigma Aldrich, St. Louis, MO, US). The digestion was performed in parallel and all samples were immediately stored at −20 • C until further analysis.

Peptide Profile by HPLC-ESI MS/MS
Prior to HPLC-ESI MS/MS analysis, digests (100 µL) were desalted using a C18 spin column (Thermo Scientific, San Jose, CA, USA), according to the manufacturer's instructions, eluting with 70% acetonitrile (v/v)/0.1% trifluoroacetic acid (TFA). MS analysis was performed using a Q Exactive Orbitrap mass spectrometer (Thermo Scientific, San Jose, CA, USA), online coupled with an Ultimate 3000 ultra-high-performance liquid chromatography instrument (Thermo Scientific, San Jose, CA, USA). Purified peptides were diluted in 50 µL of 0.1% (v/v) formic acid solution, loaded through a 5 mm long, 300 mm internal diameter pre-column (LC Packings, San Jose, CA, USA) and separated by an EASY-Spray™ PepMap C18 column (2 µm, 15 cm-75 µm; 3 mm particles; 100 Å pore size (Thermo Scientific, San Jose, CA, USA)). Eluent A was 0.1% formic acid (v/v) in Milli-Q water and eluent B was 0.1% formic acid (v/v) in acetonitrile. The column was equilibrated with 5% eluent B. Peptides were separated by a 4-40% eluent B gradient over 60 min (300 nL/min). The mass spectrometer operated in data-dependent mode and all MS1 spectra were acquired in the positive ionization mode by scanning the 1800-350 m/z range. A maximum of 10 of the most intense MS1 ions were fragmented in MS/MS mode. The resolving power was set at 70,000 full width at half maximum (FWHM), using automatic gain control (AGC) target of 1 × 10 6 ions and 100 ms as a maximum ion injection time (IT) to generate precursor spectra. MS/MS fragmentation spectra were obtained at a resolving power of 17,500 FWHM and 10 s dynamic exclusion was used to prevent repeated fragmentation of the most abundant ions. Ions with one or more than six charges were excluded from fragmentation. Spectra were elaborated using the Xcalibur Software 3.1 version (Thermo Scientific, San Jose, CA, USA).

MS Analysis Spectra Identification
Peptides were identified from the MS/MS spectra using the Proteome Discoverer 2.1 software (Thermo Scientific, San Jose, CA, USA), based on the Sequest searching algorithm. Searches were taxonomically restricted to the Triticum database extracted from UniProtKB (downloaded in February 2018). Search parameters were: Met oxidation and pyroglutamic acid for N-terminus Gln as variable protein modifications; a mass tolerance value of 10 ppm for precursor ions and 0.01 Da for MS/MS fragments; no proteolytic enzyme selected. The false discovery rate and protein probabilities were calculated by a target decoy peptide spectrum match (PSM) validator working between 0.01 and 0.05 for strict and relaxed searches, respectively. Data from three replicate LC-MS/MS analyses were merged. Peptide amount was inferred by the number of PSMs and the relevant ion count.
The T cell epitopes were determined by their native gliadin sequences as they appear prior to deamidation because tTG2 treatment was not included in the current study. These peptides are expected to be modified in vivo by lamina propria tTG2 and become immunogenic as tTG2 deamidate glutamine (Q) to glutamate (E). Only deamidated peptides fit in the HLA-DQ2.5/8 peptide binding groove and stimulate CeD promoting CD4+ T cells [23,24].

Results
Protein content varied from approximately 8.2% to 11%, and all ancestral wheat types showed values above 10% protein ( Table 2). The starch content was in the standard range (55% to 66%) for all wheat types and the thousand-kernel weight (TKW) varied from 30 to 41 g between the different wheat types.
Ex vivo digestion of porridge samples with human GI juices produced a complex variety of gluten protein fragments, which were identified by HPLC-ESI MS/MS and software-based matching. Overall, 1051-2689 peptides were identified for each sample. An assorted list of non-redundant unique peptide sequences was generated and used to identify T cell reactive epitopes in the reference list [19]. Whereas spelt digestion released few peptides, it gradually increased in einkorn (diploid) and emmer (tetraploid), and further in the common hexaploid wheat varieties Fram, Mirakel and Bastian. Thus, the latter two varieties had the highest number of unique peptides, of which 144 and 155 were T cell epitope-containing immunogenic peptides (IPs), respectively (Figure 1). This is in contrast to einkorn, spelt and emmer, which released only 38, 45 and 68 IPs, respectively ( Figure 1). Thus, digesting einkorn released considerably less IPs (18% of the peptides) than Bastian (37% of the peptides). Interestingly, none of the digested wheat types contained intact gliadin-33mer, -26mer or -25mer ( Figure 2).
γ-Gliadins released the highest number of T cell epitope-containing peptides in all wheat types, as illustrated in Figure 4. Low molecular weight glutenins were the second largest contributor of IPs in einkorn and emmer; ω-gliadins ranked as the second largest contributor, followed by α-gliadins in the common hexaploid wheat types. Only Fram and Mirakel released some IPs from high molecular weight glutenins. In addition, we also observed T cell epitope-containing peptides from wheat secalins in all wheat types, which are proteins commonly found in rye and most likely identified here by homology.  Ex vivo digestion of porridge samples with human GI juices produced a complex variety of gluten protein fragments, which were identified by HPLC-ESI MS/MS and software-based matching. Overall, 1051-2689 peptides were identified for each sample. An assorted list of non-redundant unique peptide sequences was generated and used to identify T cell reactive epitopes in the reference list [19]. Whereas spelt digestion released few peptides, it gradually increased in einkorn (diploid) and emmer (tetraploid), and further in the common hexaploid wheat varieties Fram, Mirakel and Bastian. Thus, the latter two varieties had the highest number of unique peptides, of which 144 and 155 were T cell epitope-containing immunogenic peptides (IPs), respectively (Figure 1). This is in contrast to einkorn, spelt and emmer, which released only 38, 45 and 68 IPs, respectively (Figure 1). Thus, digesting einkorn released considerably less IPs (18% of the peptides) than Bastian (37% of the peptides). Interestingly, none of the digested wheat types contained intact gliadin-33mer, -26mer or -25mer (Figure 2).     The number of epitopes is retrieved by counting the given epitope sequences in the assorted peptide list. Each bar represents a naïve T cell epitope sequence, e.g., the PFPQPQLPY sequence is specific to the DQ2.5-glia-α1a epitope [19]. Abbreviations are used to denote which gliadin fraction the epitopes are derived from: glia-α, α-gliadin; glia-γ, γ-gliadin; gliaω, ω-gliadin; glut-L, low molecular weight glutenin; glut-H, high molecular weight glutenin; hor, hordein.  . Graphical representation of the number of peptides with the indicated T cell epitope in each wheat type represented by different colors. The number of epitopes is retrieved by counting the given epitope sequences in the assorted peptide list. Each bar represents a naïve T cell epitope sequence, e.g., the PFPQPQLPY sequence is specific to the DQ2.5-glia-α1a epitope [19]. Abbreviations are used to denote which gliadin fraction the epitopes are derived from: glia-α, α-gliadin; glia-γ, γ-gliadin; glia-ω, ω-gliadin; glut-L, low molecular weight glutenin; glut-H, high molecular weight glutenin; hor, hordein.

Discussion
This study aimed to characterize CeD-associated T cell epitopes in peptides released during ex vivo gastrointestinal digestion of three ancestral wheat types (einkorn, emmer and spelt), and four common Norwegian wheat varieties (Fram, Børsum, Bastian and Mirakel). Similar to Shan et al. [10], several studies have established that the digestive-resistant 33mer (LQLQPFPQPQLPYPQPQLPYPQPQLPYPQPQPF), and 25mer (LGQQQPFPPQQPYPQPQPFPSQQPY) fragments from α-gliadin [25], and the 26mer (FLQPQQPFPQQPQQPYPQQPQQPFPQ) from γ-gliadin [26] contain most of the T cell reactive epitopes involved in the CeD immune reaction [27]. Surprisingly, these large immunodominant peptides were cleaved into smaller, but still T cell epitope-containing peptides by the current ex vivo digestive systems. This is in contrast to previous in vitro digestion experiments showing that the 33mer, 26mer and 25mer peptides were digestion resistant. Whereas previous digestive models used

Discussion
This study aimed to characterize CeD-associated T cell epitopes in peptides released during ex vivo gastrointestinal digestion of three ancestral wheat types (einkorn, emmer and spelt), and four common Norwegian wheat varieties (Fram, Børsum, Bastian and Mirakel). Similar to Shan et al. [10], several studies have established that the digestive-resistant 33mer (LQLQPFPQPQLPYPQPQLPYPQPQLPYPQPQPF), and 25mer (LGQQQPFPPQQPYPQPQPFPSQQPY) fragments from α-gliadin [25], and the 26mer (FLQPQQPFPQQPQQPYPQQPQQPFPQ) from γ-gliadin [26] contain most of the T cell reactive epitopes involved in the CeD immune reaction [27]. Surprisingly, these large immunodominant peptides were cleaved into smaller, but still T cell epitope-containing peptides by the current ex vivo digestive systems. This is in contrast to previous in vitro digestion experiments showing that the 33mer, 26mer and 25mer peptides were digestion resistant. Whereas previous digestive models used bovine or porcine digestive enzymes, the current experiments were performed with human gastroduodenal aspirates that increase the physiological relevancy of the model, and therefore more closely mimic in vivo gastrointestinal digestion [28]. However, as we used gastroduodenal aspirates from healthy, apparently non-celiac controls, we cannot rule out the possibility that celiac patients digest gluten differently due to either genetic variations in intestinal digestive enzymes or to differences in microbial-assisted gluten digestion.
The digested wheat peptides contained both single and multiple overlapping immunogenic core sequences. Whereas einkorn released the lowest number of possible Ips, followed by spelt, Bastian released the highest number of IPs. Although the percentage of IPs to the total number of peptides did not vary significantly between wheat types with different genomes, the ancestral varieties generally had a lower percentage of IPs. This is in contrast to Prandi et al. [29], who reported that the in vitro digestion of old wheat varieties produced more IPs compared to modern varieties. However, their modern wheat samples included both einkorn and spelt, which in our study were classified as ancestral, and their old varieties included only tetraploid and hexaploid wheat varieties (T. aestivum L., T. turgidum var. durum Desf., T. turgidum var. dicoccum L (emmer)), which makes it difficult to compare results. Our findings are also in contrast to Malalgoda et al. [30], who observed no differences between historical and modern wheat cultivars in T cell epitope-containing peptides released after in-gel gliadin digestion with porcine chymotrypsin. However, this in-gel digestion system with commercial enzymes does not mimic human gastrointestinal digestion as well as the current ex vivo digestive system does using human gastroduodenal juices.
The α-gliadin 33mer gene loci is located on chromosome 6D in the hexaploid wheat (AABBDD) varieties only [31]. Thus, the 33mer sequence is lacking in einkorn (AA) and emmer (AABB) but may be present in spelt, Fram, Børsum, Bastian and Mirakel (AABBDD). The α-gliadin 33mer pepride harbors six overlapping T cell epitope sequences: one copy of the DQ2.5-binding gliadin peptide, glia-α1a (PFPQPQLPY), two copies of the DQ2.5-glia-α1b (PYPQPQLPY) and three copies of the DQ2.5-glia-α2 (PQPQLPYPQ). Whereas none of these epitopes were detected in the diploid einkorn, digested emmer and spelt released one peptide with the DQ2.5-glia-α1a epitope. In contrast, these T cell epitopes were present in many peptides from the hexaploid wheat varieties. In particular, Mirakel, Bastian and Børsum released several 33mer fragments ( Figure 2). The chromosome 6D-derived gliadins were cleaved at different positions, producing peptides of different lengths with similar or multiple overlapping epitope sequences. Thus, digestion of these wheat varieties released more T cell epitope-containing peptides than the diploid and tetraploid wheat types that lack chromosome 6D.
Several studies have shown that the 13mer 31 LGQQQPFPPQQPY 43 sequence within the α-gliadin 25mer peptide and the shortened 20mer peptide detected in this study, activates the innate immune system [33] by upregulating interleukin-15, cyclooxygenase-2 (COX-2), CD25 and CD83 expression on lamina propria macrophages, monocytes, and dendritic cells prior to any CD4+ T cell stimulation [34]. More recently, Barone et al. [35] showed that the 13mer induced altered vesicular trafficking in the colonic epithelial cancer cell line Caco-2. This lead to overexpression of trans-presented IL-15/IL5R alpha complex that induced an EGFR-dependent cell proliferation that may explain the mucosal remodeling in CeD. Thus, innate immune system may also be involved in this aspect of the CeD pathogenesis.

Conclusions
The ex vivo human gastrointestinal digestion of diploid, tetraploid and hexaploid wheat types produced T cell epitope-containing peptides depending on the genomic asset. Wheat digestion with human gastrointestinal juices produced a different protein degradation pattern compared to previously reported studies that used enzymes of porcine or bovine origin. In our study, the immunodominant peptides (the 33mer and 25mer α-gliadin peptides, and the 26mer γ-gliadin peptide) were not found in their intact form, but as degraded fragments. The present digestion model did not include mucosal degradation and absorption. Therefore, the bioaccessibility and the capability of these peptides to reach the lamina propria and bind tTG2 through deamidation, and thereby acquiring HLA-DQ2.5/8 binding properties and becoming T cell epitopes, remains to be studied. The results suggested, nevertheless, that ancestral wheat types may be less CeD toxic compared to the common hexaploid varieties. Whether more general use of these ancestral wheat variants in genetically predisposed individuals could reduce CeD needs further assessment, but the incidence of diagnostic CeD in childhood has been linked to the amount of gluten in the diet [36].