Diversification of the celiac disease α-gliadin complex in wheat: A 33-mer peptide with six overlapping epitopes, evolved following polyploidization

Diversification of the celiac disease a-gliadin complex in wheat: a 33-mer peptide with six overlapping epitopes, evolved following polyploidization Carmen V. Ozuna 1,† , Julio C. M. Iehisa 1,†,‡ , Mar ıa J. Gim enez 1 , Juan B. Alvarez 2 , Carolina Sousa 3 and Francisco Barro 1, * 1 Departamento de Mejora Gen etica, Instituto de Agricultura Sostenible (IAS), Consejo Superior de Investigaciones Cient ıficas (CSIC), C ordoba, E-14080 Spain, 2 Departamento de Gen etica, Escuela Superior de Ingenieros Agr onomos y Montes, Universidad de C ordoba, C ordoba, E-14071 Spain, and 3 Departamento de Microbiolog ıa y Parasitolog ıa, Facultad de Farmacia, Universidad de Sevilla, Sevilla, 41012 Spain Received 31 January 2015; revised 1 April 2015; accepted 2 April 2015; published online 10 April 2015. *For correspondence (email [email protected]). † These authors contributed equally to this work. ‡ Present address: Departamento de Biotecnolog ıa, Facultad de Ciencias Qu ımicas, Universidad Nacional de Asunci on, San Lorenzo, Paraguay. SUMMARY The gluten proteins from wheat, barley and rye are responsible both for celiac disease (CD) and for nonceliac gluten sensitivity, two pathologies affecting up to 6–8% of the human population worldwide. The wheat a-gliadin proteins contain three major CD immunogenic peptides: p31–43, which induces the innate immune response; the 33-mer, formed by six overlapping copies of three highly stimulatory epitopes; and an additional DQ2.5-glia-a3 epitope which partially overlaps with the 33-mer. Next-generation sequencing (NGS) and Sanger sequencing of a-gliadin genes from diploid and polyploid wheat provided six types of a-gliadins (named 1–6) with strong differences in their frequencies in diploid and polyploid wheat, and in the presence and abundance of these CD immunogenic peptides. Immunogenic variants of the p31–43 peptide were found in most of the a-gliadins. Variants of the DQ2.5-glia-a3 epitope were associated with specific types of a-gliadins. Remarkably, only type 1 a-gliadins contained 33-mer epitopes. Moreover, the full immunodominant 33-mer fragment was only present in hexaploid wheat at low abundance, probably as the result of allohexaploidization events from subtype 1.2 a-gliadins found only in Aegilops tauschii, the D-genome donor of hexaploid wheat. Type 3 a-gliadins seem to be the ancestral type as they are found in most of the a-gliadin-expressing Triticeae species. These findings are important for reducing the incidence of CD by the breeding/selection of wheat varieties with low stimulatory capacity of T cells. Moreover, advanced genome-editing techniques (TALENs, CRISPR) will be easier to implement on the small group of a-gliadins containing only immunogenic peptides. Keywords: alpha-gliadin, wheat, celiac disease, 33-mer peptide. INTRODUCTION Wheat is one of the most important crops in the world, with an annual production of about 715 million tons (2013; http://faostat3.fao.org/). Bread wheat (Triticum aestivum, 2n=6x=42; genomic code BBAADD) is an allohexaploid species that arose by natural hybridization between emmer wheat (Triticum turgidum ssp. dicoccum,2n=4x=28, BBAA), and the diploid Aegilops tauschii (2n=2x=14, DD) (Petersen et al., 2006). In turn, tetraploid emmer wheat is hypothesized to have originated through hybridization between the diploids T. urartu (AA) and, possibly, Ae. speltoides (SS) (Petersen et al., 2006). Despite its relatively low protein content (8–15%), wheat is the most important protein source in the human diet. Gluten, the water insoluble fraction of wheat flour protein, is responsible for the bread-making quality of wheat and is mainly composed of two prolamin fractions, called gliadins (a,cand x) and glutenins (Shewry, 2009). The ingestion of these proteins is responsible for two important pathologies: (i) celiac disease (CD), a food-sensitive enteropathy with a prevalence of about 0.7–2% in the human population, in genetically predisposed individuals (Rewers, 2005); and (ii) gluten sensitivity, a newly-recognized pathology ©2015 Society for Experimental Biology and John Wiley & Sons Ltd794 The Plant Journal (2015) 82, 794–805 doi: 10.1111/tpj.12851 with an estimated prevalence of 6% in the USA population (Sapone et al., 2011). In CD, T cells isolated from the intestinal mucosa typically recognize gluten peptides in which specific glutamine residues are converted to glutamate by tissue transglutaminase 2 (tTG2). These modified peptides are able to bind to class II human histocompatibility leukocyte antigen (HLA) molecules DQ2 and DQ8, which stimulate T cells and trigger an inflammatory response in the small intestine leading to flattening of the mucosa (Wieser and Koehler, 2008). Over 90% of CD patients possess HLADQ2, encoded by the DQA1*05 and DQB1*02 genes (Karell et al., 2003). The a-gliadin 33-mer is one of the digestion-resistant gluten peptides that is highly reactive to isolated celiac T cells and is the main immunodominant toxic peptide in celiac patients. This peptide is present in the N-terminal repetitive region of a-gliadins and contains six overlapping copies of three different DQ2-restricted T-cell epitopes with highly stimulatory properties (Shan et al., 2002). a-gliadins also contain an additional DQ2-restricted epitope which partially overlaps with 33-mer peptide (Vader et al., 2002). Moreover, the peptide p31–43 of these a-gliadins has been reported to induce the innate immune response necessary to initiate the T-cell adaptive response (Maiuri et al., 1996a, 2003). The a-gliadins are encoded by the Gli-2 loci located on the short arms of the homoeologous group 6 chromosomes of wheat (Payne, 1989). The estimated copy numbers of a-gliadins in hexaploid wheat are between 25 and 150 copies (Harberd et al., 1985; Anderson et al., 1997). Analysis of this highly variable multigene family has been performed in tetraploid wheat through RNA-amplicon sequencing applying 454’s NGS technology (Salentijn et al., 2013). In this work a comprehensive study combining NGS genomic-amplicon sequencing and Sanger sequencing of the entire fragment of a-gliadins containing immunogenic epitopes has been carried out in diploid, tetraploid and hexaploid wheats. We identify six different types of a-gliadins but only one type contains all the immunogenic peptides and epitopes, and the five other types of a-gliadins do not contain epitopes for the 33-mer peptide. RESULTS Genes and pseudogenes of a-gliadins We subjected the genome of domesticated and wild wheat and relatives, including 96 accessions of Triticum and Aegilops species (Data S1 and Figure S1), to amplicon NGS. The N-terminal repetitive region of a-gliadins, containing three highly immunogenic peptides, was amplified and sequenced (Figure 1). We obtained 200 340 cleaned reads (see Experimental Procedures) with an average of 2087 reads per accession, which were clustered at 99% identity and then a consensus sequence was extracted from each cluster. The high-confidence sequence variants (see Experimental Procedures) were grouped into 999 unique clusters which consisted of 88 736 total reads (Table S1). Multiple alignments of consensus sequences were performed. However, alignments of sequences with frequent insertions/deletions and repeat units, such as a-gliadins, are not accurate using traditional software (Loytynoja and Goldman, 2005; Jordan and Goldman, 2012). In a-gliadins, the repeat unit PFPPQQPYPQPQ or its variants can be found along the entire fragment from the p31–43 peptide to the DQ2.5-glia-a3 epitope. Considering this repeat unit, we manually aligned the consensus sequences obtained by clustering. (b) (a) Figure 1. Amplicon design. (a) In the full-length a-gliadin gene (accession number AJ133612), following parts are indicated: signal peptide (SP), transition peptide (Nter), repetitive domain and CterI, CterII and CterIII domains. (b) Amplicon segment and the main immunotoxic region in which the peptides p31–43, 33-mer and DQ2.5-glia-a3 are indicated. Primers used are indicated by green triangles: aGli900F1, MJ_R6 and MJ_R3 to amplify the amplicon, then VH_agliF1 and VH_agliR3 to amplify the complete a-gliadin gene. ©2015 Society for Experimental Biology and John Wiley & Sons Ltd, The Plant Journal, (2015), 82, 794–805 The a-gliadin immunogenic complex in wheat 795 In 612 clusters (representing 65 778 out of 88 736 reads), the consensus sequence did not present frame shift or premature stop codon (PSC). These sequences were classified as putative genes and the remaining clusters as pseudogenes. The gene-derived reads ranged in average from 93.4% in T. monococcum to 60.0% in T. polonicum (Figure S2a). T. spelta and S-genome accessions (Ae. speltoides, Ae. searsii, and Ae. longissima) showed high frequency of gene-derived reads. We also cloned and sequenced the complete sequence of a-gliadins from one accession of each species by the Sanger method (Table 1). The proportion of pseudogenes was higher in tetraploid (average of 76%) and hexaploid (average of 63%) wheats compared with their wild diploid progenitors T. urartu (49%), Ae. speltoides (36%), and Ae. tauschii (21%). The domesticated diploid wheat T. monococcum also presented a higher proportion of pseudogenes. In contrast, Ae. searsii (S genome) showed the lowest proportion of pseudogenes (12%). In general, the number of pseudogenes per genome is lower in diploids while increased in polyploids. In some pseudogenes, frame shift and/or PSC only appeared downstream of the PIS motif (Figure 1 and Table 1). Thus, the NGS amplicons include sequences without any mutation (real genes) and those containing mutations only downstream of PIS, which we cannot distinguish between them as mutations are out of the amplicon. Considering this observation and based on the proportion of pseudogenes in the Sanger sequencing, we estimated the proportion of genes in amplicons and obtained a result similar to the Sanger sequences (Figure S2b) except in T. spelta (higher than Sanger) and T. durum (lower than Sanger). Types of a-gliadins From the alignment of consensus sequences obtained by clustering, six types of a-gliadin sequences (named 1–6) were identified which varied mainly in the pattern and number of repeats in the region corresponding to the 33mer (Figure 2a), although some variants differed in regions other than the 33-mer. In comparison with type 2 a-gliadins, type 1 contained a deletion of PFPPQ, and type 3 a deletion of PYPQPQ. Type 4 sequences had one repeat unit (PFPPQQPYPQPQ or its variant) fewer than type 2. Type 5 also lacked one repeat unit compared with type 3, with an additional deletion of PFPPQQ or its variant. In type 6 a-gliadins, deletion of one repeat unit was observed as compared with type 3. The a-gliadin genes of S-genome diploids (Ae. speltoides,Ae. longissima and Ae. searsii) were mainly composed of type 3 sequences (Figure 2b). In contrast, type 1 a-gliadins predominated in diploids with A (T. urartu and T. monococcum) and D genomes (Ae. tauschii). Type 1 a-gliadins were rare or absent in Ae. longissima (Data S2), type 2 were not found in Ae. speltoides and Ae. tauschii and type 3 was absent in A-genome diploids. Type 4 a-gliadins were found only in one accession of Ae. tauschii. Type 1 a-gliadins also predominated in tetraploid and hexaploid wheats, followed by type 6 sequences. These trends were also observed in pseudogenes, except in Ae. longissima (higher proportion of type 2 a-gliadins), Ae. tauschii (higher proportion of type 3), T. monococcum (higher proportion of type 2) and T. spelta (dominance of type 6 and very low proportion of type 1). A lower proportion of type 2 and type 3 sequences was also notable in both genes and pseudogenes of tetraploid and hexaploid wheats (Figure 2b). In general, similar results were obtained from Sanger sequencing (Table S2), with the differences in that type 2 was not found in T. monococcum, type 4 was also found in Ae. speltoides,Ae. searsii, and T. aestivum, and type 5 in T. dicoccum and T. durum. Type 6 was found only in T. macha and T. spelta in a lower proportion than expected. Other types not found in NGS amplicon sequencing were also found at lower frequencies. Table 1 Description of complete a-gliadins genes and pseudogenes sequenced by Sanger Genotypes Genome Genes Pseudogenes Total Genes/ genome Pseudogenes/ genome Pseudogenes with mutation after PIS motif Maximum length (bp) Minimum length (bp) T. macha BBAADD 38 78 116 12.7 26.0 40 970 846 T. spelta BBAADD 42 73 115 14.0 24.3 27 957 657 T. aestivum BBAADD 43 57 100 14.3 19.0 34 924 837 Average 41.0 69.3 110.3 13.7 23.1 950.3 780.0 T. dicoccum BBAA 14 42 56 7.0 21.0 13 939 838 T. durum BBAA 19 60 79 9.5 30.0 39 939 845 Average 16.5 51.0 67.5 8.3 25.5 939.0 841.5 T. urartu AA 19 18 37 19.0 18.0 13 912 831 T. monococcum A m A m 6 18 24 6.0 18.0 7 885 852 Ae. searsii S s S s 22 3 25 22.0 3.0 3 933 858 Ae. speltoides SS 14 8 22 14.0 8.0 7 936 864 Ae. tauschii DD 22 6 28 22.0 6.0 3 906 840 Average 16.6 10.6 27.2 16.6 10.6 914.4 849.0 ©2015 Society for Experimental Biology and John Wiley & Sons Ltd, The Plant Journal, (2015), 82, 794–805 796 Carmen V. Ozuna et al. Subtypes of type 1 a-gliadins The type 1 a-gliadins can be divided in subtypes according to the number of P(F/Y)PQPQL repeat units present in the region of 33-mer (Figure 2a), ranging from one (subtype 1.1) to four (subtype 1.4). Subtype 1.1 was found in almost all accessions analyzed and was the predominant class in most. This subtype can contain up to two canonical CD epitopes in the region corresponding to the 33-mer, depending on the presence of amino acid substitutions that affect one or both epitopes (Figure 2a). In hexaploids, tetraploids, and diploids with the A genome, subtype 1.1 with one epitope (subtype1.1-1) was in the majority with a lower proportion of the subtype without epitopes (subtype 1.1-0, Table 2). Hexaploids also contained subtype 1.1-2 with two epitopes. Diploids with the S genome contained only subtype 1.1-0 with no epitopes, except in accession 406 of Ae. longissima which had a low proportion of 1.1-1. In contrast, subtype 1.1-2 predominated in Ae. tauschii with a low proportion of 1.1-1 and absence of 1.1-0. Subtypes with more than one P(F/Y)PQPQL repeat unit (1.2 to 1.4) were observed only in species with the D genome such as hexaploid wheat and Ae. tauschii (Figure 2a and Table 2). In general, the subtype 1.2 variants with four epitopes (1.2-4) were abundant in these species followed by 1.2-2 in hexaploids and 1.2-1 in Ae. tauschii. The variants 1.2-0 and 1.2-1 were absent in hexaploid wheat. The subtype 1.3 with three repeat units is equivalent to the complete 33-mer peptide and contained six epitopes. Although the variant 1.3-6, or 33-mer peptide, was found only in hexaploid wheat at lower frequency, and it was not detected in 10 hexaploid lines. In one accession of Ae. tauschii, subtype 1.4-6 was found but in low proportion. DQ2.5-glia-a3 variants We analyzed the variants of DQ2.5-glia-a3 epitope, located downstream of 33-mer in amplicons classified as genes. Three major variants of this epitope were identified and named FR-, FP-, and FS-type according to the first two amino acids of their sequences (Figure 3a and Table S3). Almost all type 1 a-gliadins were associated with FR-type variants, with the canonical DQ2.5-glia-a3 (FRPQQPYPQ) epitope itself the most abundant in all but diploids with the S genome. In these diploids, the most abundant was the variant FRPQQPQPQ which originated from a partial deletion of PYPQ or its variants in some FR-type sequence. Most of the type 2 and type 3 a-gliadin sequences were associated with FP-type variants, the vast majority being the type FPPQQPYPQ. The FS-type variant FSPQQPYPQ was abundant in type 2 a-gliadins of T. urartu, and type 3 of Ae. speltoides but was also found in other species. A (a) (b) – T. aestivum T. compactum T. macha T. spelta T. turgidum T. durum T. dicoccum T. polonicum T. urartu T. monococcum Ae. tauschii Ae. speltoides Ae.lo Ae.se %Reads Genes Pseudogenes %Reads Figure 2. Types of a-gliadin. (a) Alignment of six types of a-gliadin found in NGS amplicon sequencing (top panel), from the first PFPPQQ motif at the p31–43 to PIS motif. Region corresponding to the 33-mer is shaded. Alignment of type 1 subtypes (bottom panel) with different number of CD epitopes. The number after the dot indicates the subtype which represents the number of P(F/Y)PQPQL repeat unit, and that after the hyphen the number of CD epitopes (present in the region indicated in bold). DQ2.5-glia-a1a epitopes are indicated by blue underlines, DQ2.5-glia-a1b by orange and DQ2.5-glia-a2 by red. Amino acid substitutions affecting these three epitopes are indicated in red. (b) Percentage of different a-gliadin types in reads corresponding to genes and pseudogenes in different species. Ae.lo–Ae.se: Ae. longissima and Ae. searsii. ©2015 Society for Experimental Biology and John Wiley & Sons Ltd, The Plant Journal, (2015), 82, 794–805 The a-gliadin immunogenic complex in wheat 797 higher abundance of other variants was observed in type 3 sequences, mainly in T. polonicum,T. dicoccum and T. turgidum, all tetraploid wheats. The lack of type 2 a-gliadins in Ae. speltoides and Ae. tauschii and type 3 in A-genome diploids explain the absence of DQ2.5-glia-a3 variants in their respective sequences. Type 6 a-gliadins, relatively abundant in polyploid wheat, were associated with the FPtype variant FPPQQSYPQ (Table S3). p31–43 peptide variants Two major variants of the peptide p31–43 (LG-, and LPtype), associated with the innate immune response induced by gluten, were found in all diploid and polyploidy species analyzed (Figure 3b and Table S4). Almost all of the type 1 a-gliadins in diploids with the A-genome and tetraploid wheats were found to be associated with the LGtype p31–43 peptide, and in Aegilops species with the PGtype. However, in hexaploids around 60% were associated with the LG-type, and 30% with the PG-type. In both cases, the two canonical peptides predominated in all species analyzed. In contrast, type 2 a-gliadin sequences were associated mainly with the PG-type where variants with one mismatch were seen at high frequency. In Ae. searsii and Ae. longissima, deletion of (L/P)GQQQP produced the variant LVQQQQFPPQQPY and accounted for around 90% of type 2 sequences. The PG-type also predominated in type 3 a-gliadins of many species but with a higher proportion of LG-type and other variants compared with type 2a-gliadins. However, the LG-type was not found in type 3 a-gliadins from T. macha and Aegilops species. As in type 1a-gliadin sequences, the canonical p31–43 peptides predominated in the LGand PG-types except in diploids with the S genome. As in DQ2.5-glia-a3, the lack of type 2 a-gliadins in Ae. speltoides and Ae. tauschii and type 3 in A-genome diploids explain the absence of p31–43 variants in their respective sequences. In type 6 a-gliadins, the canonical PG-type peptide accounted for more than 98% of variants (Table S4). Abundance of total CD epitopes and their variants To estimate the potential toxicity for each accession, the gluten T-cell epitopes restricted by HLA-DQ molecules (Sollid et al., 2012) were searched for in predicted amino acid sequences allowing up to two mismatches (Figure 4 and detailed in Data S3). The abundance of each epitope was calculated by multiplying the total number of epitopes found in a given gene by the frequency of that gene in the genome. The diploids with the S genome contained very few or no canonical epitopes. The total abundance of canonical epitopes was lower in tetraploid wheats, followed by hexaploids and T. urartu, and T. monococcum and Ae. tauschii with the highest abundance. The abundance of CD epitope variants with one mismatch was very high with respect to that of the Table 2 Percent of Type 1 a-gliadins containing different number of epitopes a-Gliadin subtype a T. aestivum T. compactum T. macha T. spelta T. turgidum T. durum T. dicoccum T. polonicum T. urartu T. monococcum Ae. tauschii Ae. speltoides Ae.loAe. se 1.1-0 4.26 6.02 2.84 0.00 6.31 8.00 6.61 5.85 26.00 0.22 0.00 17.73 3.49 1.1-1 24.95 23.01 25.65 25.03 29.80 43.02 39.43 37.61 66.71 79.32 0.32 0.00 0.17 1.1-2 4.71 3.86 6.40 4.39 0.00 0.00 0.00 0.00 0.00 0.00 35.55 0.00 0.00 1.2-0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.38 0.00 0.00 1.2-1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 13.92 0.00 0.00 1.2-2 2.91 1.87 4.46 2.95 0.00 0.00 0.00 0.00 0.00 0.00 3.67 0.00 0.00 1.2-4 7.56 6.45 8.93 9.82 0.00 0.00 0.00 0.00 0.00 0.00 23.38 0.00 0.00 1.3-6 1.50 1.08 1.76 1.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.4-6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.37 0.00 0.00 Total 45.88 42.29 50.05 43.39 36.11 51.03 46.04 43.46 92.71 79.54 78.59 17.73 3.65 a Number before hyphen is type and subtype of a-gliadin and that after hyphen indicates the number of total epitopes (DQ2.5-glia-a1a, DQ2.5-glia-a1b and DQ2.5-glia-a2) in the region corresponding to the 33-mer. Percentage was calculated respect to the total number of a-gliadins classified as ‘gene’ by NGS. Ae.lo-Ae.se: Ae. longissima and Ae. searsi. ©2015 Society for Experimental Biology and John Wiley & Sons Ltd, The Plant Journal, (2015), 82, 794–805 798 Carmen V. Ozuna et al. canonical sequence, in T. monococcum, followed by Ae. speltoides,Ae. searsii and T. spelta, and was low in tetraploid wheats, T. compactum, and T. urartu. In contrast to the case of canonical epitopes, the abundance of epitopes with two mismatches was higher in diploids with S genome and lower in diploids with A and D genomes. In hexaploid wheats, the abundance was slightly higher than in these latter diploids. Although the immunogenic capacity of most of the variants with one or two mismatches has not been tested, these amino acid substitutions usually abolish or decrease the T-cell stimulation (Data S4). In addition, five out of 39 variants with one mismatch and 22 out of 39 variants with two mismatches contained proline at positions 2, 4, or 9, and/or positively charged amino acids in positions 4, 6, or 7, which may decrease toxicity (Kim et al., 2004). Our results suggest that diploids with the S genome are the least toxic group, and Ae. tauschii and T. monococcum two of the most toxic species. Tetraploid wheats might also be considered one of the least toxic. Based on the abundance of CD epitope variants and the abundance of type 1 a-gliadins with different number of epitopes, we selected putative ‘reduced toxicity’ accessions. Besides diploids with the S genome, two hexaploid and seven tetraploid wheats were selected (Data S3). Among T. aestivum genotypes, the accession THA85 presented the lowest abundance of canonical epitopes and type 1 a-gliadin sequences with more than two epitopes, and a relatively low abundance of type 1 sequences with two epitopes. Another hexaploid, the T. compactum accession C2, presented a lower abundance of canonical epitopes. Although tetraploid wheats presented a relatively low level of canonical epitopes compared with hexaploid wheats and lacked type 1 a-gliadins with two or more epitopes, the seven selected accessions contained a low level of canonical epitopes, a relatively %Gene reads Type 1 Type 2 Type 3 T. aestivum T. compactum T. macha T. spelta T. turgidum T. durum T. dicoccum T. polonicum T. urartu T. monococcum Ae. tauschii Ae. speltoides Ae.lo (a) Other FS-type FP-type FR-type (b) %Gene reads Type 1 Type 2 Type 3 T. aestivum T. compactum T. macha T. spelta T. turgidum T. durum T. dicoccum T. polonicum T. urartu T. monococcum Ae. tauschii Ae. speltoides Ae.lo Types of -gliadin DQ2.5-glia3 variants p31-43 variants –Ae.se –Ae.se Figure 3. Frequency of the DQ2.5-glia-a3 epitope and p31–43 peptide in the three major a-gliadin types. (a) DQ2.5-glia-a3 variants were grouped into four types according to the first two amino acids: FR-type, FP-type, FS-type and other. (b) Peptide the p31–43 variants were grouped into four types: in LG-type and PG-type the first two amino acids are respectively LG and PG, variants with partial deletion and other types. Partial sequences in which the p31–43 variant could not determine are indicated as partial. Frequency was determined in reads classified as genes. ©2015 Society for Experimental Biology and John Wiley & Sons Ltd, The Plant Journal, (2015), 82, 794–805 The a-gliadin immunogenic complex in wheat 799 high level of type 1 a-gliadins without epitopes, and a low level of those with one epitope. These observations may serve to select varieties or accessions with lower toxicity. Phylogenetic analysis of a-gliadins To study the origin of different a-gliadin types in wheat, phylogenetic analysis was performed including the a-gliadin sequences available in the NCBI nucleotide database. In total, 1036 a-gliadin genes and pseudogenes were found in Triticum–Aegilops species (Table S5) and 169 genes in other Triticeae species (Figure 5a). As in amplicon sequences, the type 1 a-gliadins predominated in Triticum– Aegilops, with subtype 1.1 accounting for 594 hits. In contrast with the NGS amplicon sequencing, type 2 a-gliadins were also found in Ae. tauschii and Ae. speltoides in a small proportion. No sequences other than type 3 were found in T. turgidum, which may be explained by the small number of T. turgidum sequences in the database. Type 6 sequences were present in T. aestivum, also in Ae. speltoides, but at lower frequency than expected by NGS amplicon sequencing. We found that type 6 a-gliadins end with the GIMSTN motif resulting from three nucleotide substitutions, compared with other types (such as types 1, 2 and 3) that end with the motif GIFGTN (Figure S3). Type 3 a-gliadins predominated in many other Triticeae species, which may indicate that this is the ancestral a-gliadin type (Figure 5a). Type 2 a-gliadins seem to be the main component in Lophopyrum elongatum (24 out of 28). Type 1 was found in Secale cereale (rye) another highly toxic Perfect match T. aestivum T. compactum T. macha T. spelta T. turgidum T. durum T. dicoccum T. polonicum T. urartu T. monococcum Ae. tauschii Ae. speltoides Ae.lo Two mismatches One mismatch Total epitopes x frequency –Ae.se Figure 4. Abundance of CD epitopes in different Triticum/Aegilops species. Abundance of: canonical epitopes (top panel), variants with one mismatch (middle panel), and with two mismatches (bottom panel) in diploid, tetraploid and hexaploid wheat. Abundance of each epitope or its variants was calculated multiplying the total number of epitopes found in a given gene by its frequency on the genome. The y-axis represents the total abundance of all epitopes or their variants. (a) (b) Figure 5. Phylogenetic analysis of a-gliadins. (a) Number of a-gliadin sequences of other Triticeae species by types found in NCBI nucleotide database. (b) Phylogenetic tree of a-gliadins constructed using N-terminal region. In total, 478 sequences of the tribe Triticeae with the complete N-terminal repetitive region, from Nter to PIS (Figure 1), were used. These included 207 sequences obtained by Sanger method in this study and 271 found in NCBI database (124 of Triticum/Aegilops and 147 of other Triticeae species). Type 1 a-gliadins are indicated in blue, type 2 in red, type 3 in cyan, type 6 in gray and other types in black. Circles indicate canonical variants and other variants are indicated by triangles. In the case of type 1, the canonical subtype 1.1 is indicated by circles, variants of subtype 1.1 by triangles and subtypes 1.2 or 1.3 by squares. Number indicates branch support estimated by SH-like approach. ©2015 Society for Experimental Biology and John Wiley & Sons Ltd, The Plant Journal, (2015), 82, 794–805 800 Carmen V. Ozuna et al. cereal forbidden for celiac people. One additional type of a-gliadin (type 7) was found in other Triticeae but not in Triticum–Aegilops species and contained a deletion of PFPPQL motif (or its variants) compared with type 2. Phylogenetic analysis of the N-terminal region (from Nter to the PIS motif; Figure 1) of a-gliadins, using these sequences and those derived from Sanger sequencing in this study, indicated that type 3 a-gliadins are more closely related to type 2 than to type 1 (Figure 5b). We obtained a similar result from constructing a phylogenetic network using complete sequences (Figure S4), validating our classification of the a-gliadins. DISCUSSION The a-gliadin types and their toxicity The a-gliadin genes encompass a large multigene family with highly variable and highly immunogenic N-terminal repetitive regions. Two interspersed repeat motifs are readily identified (Shewry and Tatham, 1990). Amplification and NGS sequencing of this region revealed that a-gliadin sequences differed mainly in the number of repeat blocks consisting of two interspersed motifs: PFPPQQ and PYPQPQ. Through alignment in accord with the pattern of these two motifs, we found six types of a-gliadins of which types 1, 2, 3, and 6 were the most abundant. Only the type 1a-gliadins contain one or more of the canonical 33-mer CD epitopes such as DQ2.5-glia-a1a/b and DQ2.5-glia-a2. In addition, type 1 contains the canonical DQ2.5-glia-a3 epitope. Other main types contain variants of these epitopes, except the canonical DQ2.5-glia-a3, which was rarely found in types 2 and 6. This indicates that type 1 is the most immunogenic of the a-gliadins, with subtypes having a higher number of epitopes, such as 1.4-6 and 1.3-6 (33mer), being the most immunogenic of the type, followed by 1.2-4. These subtypes were found only in T. aestivum and Ae. tauschii, which may explain why these species are highly immunogenic (Molberg et al., 2005). In contrast, in diploids with the S genome, although type 1 sequences were present, immunogenic subtypes were absent or very low, explaining the inability to stimulate T cells in CD patients (Molberg et al., 2005). In the A genome, the proline-to-serine (P/S) substitution in DQ2.5-glia-a2 eliminates its toxicity (Molberg et al., 2005; Mitea et al., 2010). Despite the absence of DQ2.5-glia-a1a/b and DQ2.5-glia-a2 epitopes in type 3 a-gliadins, these seem to stimulate a-II-specific T cells (recognizing DQ2.5-glia-a2 epitope) to a lesser extent in some CD patients because type 3 a-gliadins contain peptide W09 described in (Tye-Din et al., 2010). The subtype 1.3-6 (containing six overlapping epitopes) was found only in the hexaploid wheat and at low frequency, as predicted previously (Molberg et al., 2005) and was absent in some accessions. We found that accessions with relatively low toxicity can be selected based on the abundance of total canonical epitopes and type 1 subtypes with different number of epitopes. Although T. monococcum has been found to stimulate only a-I-specific T cells (recognizing DQ2.5-glia-a1a epitope) (Molberg et al., 2005), this epitope was highly abundant and therefore we classified it as one of the most immunogenic species. However, the a-gliadin proteins present in seeds of these selected accessions, in combination with the other gluten proteins, may be still high enough to stimulate T cells in CD patients. The canonical p31–43 peptide (able to trigger innate immune response), was present in all a-gliadin types except in type 2, where mutated variants are frequent. As the toxicity of p31–43 variants has not been studied, the potential toxicity of type 2 a-gliadins cannot be predicted. Origin and evolution of a-gliadin types in wheat The a-gliadins have been found not only in the genera Triticum and Aegilops but also in many other species of the tribe Triticeae (Qi et al., 2013). In some Triticeae species such as barley and rye, a-gliadins are not present (Shewry and Tatham, 1990). Based on our results and those from the NCBI database, we hypothesize that after duplication of type 3 a-gliadin (probably the ancestral type), a duplication of the P(F/Y)PQPQ motif in the region corresponding to 33mer in one of the copies was the origin of the type 2 a-gliadins in Triticum,Aegilops and other genera (Figure 6). This latter type in turn, after duplication, gave rise to type 1.1 by a deletion of the PF(L/P)PQ motif with one DQ2.5-gliaa1a epitope in Triticum and Aegilops. Thus, the common ancestor of Triticum and Aegilops possibly had these three a-gliadins types, which is in agreement with the suggestion that the duplication of certain a-gliadins took place before diploid differentiation (Kawaura et al., 2012). After speciation, diploids with the A genome lost type 3 a-gliadins and Ae. speltoides and Ae. tauschii could have lost type 2. Recently, Marcussen et al. (2014) showed that the homoploid hybridization of diploids with A genome and B genome gave rise to Ae. tauschii. According to this, the species specific loss of a-gliadin type might occur after the origin of Ae. tauschii or this latter inherited the Gli-2 locus of B genome. We did not find type 2 from the Sanger sequencing of T. monococcum. This might be due to the small number of sequenced clones available since type 2 is less abundant in this species. Because we found type 2 sequences in Ae. speltoides and Ae. tauschii in the NCBI nucleotide database, some accessions may have this a-gliadin or it may have been introduced by interspecific hybridization. Also at this stage, proline-to-leucine substitution occurred in diploids with the S genome leading to the loss of DQ2.5-glia-a1a and DQ2.5-glia-a2 epitopes, including the P/S substitution at DQ2.5-glia-a2 in A-genome diploids. Duplication of the a-gliadins may have occurred after a deletion event and the establishment of ©2015 Society for Experimental Biology and John Wiley & Sons Ltd, The Plant Journal, (2015), 82, 794–805 The a-gliadin immunogenic complex in wheat 801 these genome-specific variants, since some types and/or subtypes are absent in some species. In tetraploid wheat, type 1 (especially subtype 1.1-1) and type 2 a-gliadins were mainly contributed by T. urartu and type 3 by Ae. speltoides. In the D genome of Ae. tauschii, at least one duplication of P(F/Y)PQPQL has occurred resulting in a more toxic a-gliadin subtype. Although subtype 1.4 was found in Ae. tauschii, in this study and the NCBI database, no subtype 1.3 was found, suggesting that this subtype was originated after an allohexaploidization event. As type 6 a-gliadin was found only in polyploid wheat, it seems to have been originated after the hybridization between T. urartu and Ae. speltoides. However, one a-gliadin from Ae. speltoides corresponded to type 6, indicating that it could be originated before allotetraploidization. Due to its similarity to type 2 a-gliadin, type 6 might be originated from deletion of ‘PFPSQQPYLQLQPYPQPQ’ or its variant. The type 6 a-gliadin differs from other types in the presence of GIMSTN motif at the end of coding region. For this reason, the commonly used primers targeting the GIFGTN motif (van Herpen et al., 2006; Mitea et al., 2010; Xie et al., 2010; Qi et al., 2013; Li et al., 2014) can rarely amplify it, explaining the low presence of this sequence found by our Sanger sequencing and the NCBI database. Similarly, the high proportion of type 1 in pseudogenes found by Sanger sequencing in T. spelta indicates that for some reason (such as the low efficiency of amplification using our primer sets) we could not detect it in the amplicon. Pseudogenization of members of multigene families occurs relatively frequently (Kambere and Lane, 2007). In previous reports, approximately 50–87% of a-gliadins have been found to be pseudogenes (Anderson and Greene, 1997; Xie et al., 2010), even in the diploid ancestors of polyploid wheat (van Herpen et al., 2006). However, our findings suggest that after polyploidization the proportion of pseudogenes increased compared with their diploid ancestors, especially Ae. speltoides and Ae. tauschii. In hexaploid wheat, genetic redundancy created by polyploidization may allow an accelerated accumulation of mutations, leading to pseudogenization of duplicated genes (Akhunov et al., 2013). This could explain the low number of pseudogenes in wild diploids as compared with polyploid wheat. In this work six types of a-gliadins were identified in diploid and polyploid wheats but only one contains all the immunogenic peptides and epitopes, and five types of a-gliDuplication of type 3 -gliadin P(F/Y)PQPQ Duplication of type 2 -gliadin PFLPQ Most of the Triticeae species Triticum-Aegilops and some other Triticeae species Triticum-Aegilops T. urartu (AA) A. speltoides (SS) A. tauschii (DD) Deletion of type 3 Deletion of type 2 Duplication events Duplication events P(F/Y)PQPQL T. turgidum (BBAA) Origin of type 6? T. aestivum (BBAADD) P(F/Y)PQPQL Type 1.3 (33-mer) Deletion of type 2 Figure 6. Proposed model of a-gliadin evolution in wheat. Type 3 a-gliadins are indicated by blue boxes, type 2 by green boxes, subtype 1.1 in orange and 1.2 in purple. The repeat motifs P(F/Y)PQPQ and PF(L/P) PQ are indicated respectively by blue and red letters. Duplication of the a-gliadins might occurred after deletion events indicated by boxes with cross. Type 6 a-gliadin might be originated in tetraploid wheat and the subtype 1.3 after the allohexaploidization event. Broken lines indicate hybridization and polyploidization events. ©2015 Society for Experimental Biology and John Wiley & Sons Ltd, The Plant Journal, (2015), 82, 794–805 802 Carmen V. Ozuna et al.