CRISPR/Cas9-mediated multiplex gene editing of gamma and omega gliadins: paving the way for gliadin-free wheat

Abstract

Wheat is a staple cereal in the human diet. Despite its significance, an increasing percentage of the population suffers adverse reactions to wheat, which are triggered by wheat gluten, particularly the gliadin fractions. In this study, we employed CRISPR/Cas [clustered regularly interspaced palindromic repeats (CRISPR)/CRISPR-associated protein] multiplexing to introduce targeted mutations into γ- and ω-gliadin genes of wheat, to produce lines deficient in one or both immunogenic gliadin fractions simultaneously. For this study, eight single guide RNAs (sgRNAs) were designed and combined into four plasmids to produce 59 modified wheat lines, of which 20 exhibited mutations in the target genes. Characterization of these lines through Sanger sequencing or next-generation sequencing revealed a complex pattern of InDels, including deletions spanning multiple sgRNAs. The mutations were transmitted to the offspring, and the analysis of homozygous derived lines by reverse-phase HPLC and monoclonal antibodies showed a 97.7% reduction in gluten content. Crossing these lines with other CRISPR/Cas lines deficient in the α-gliadins allowed multiple mutations to be combined. This work represents an important step forward in the use of CRISPR/Cas to develop gluten-free wheat.

This study demonstrates a significant advance in developing gluten-free wheat using CRISPR/Cas to target γ- and ω-gliadin genes, followed-by crosses with α-gliadin mutants, to achieve a 97.7% reduction in gluten content.

Introduction

Wheat is one of the most widely consumed cereals in the world in the form of a wide range of products such as noodles, biscuits, bread, etc. Wheat gluten is the major determinant of the unique viscoelastic properties of wheat dough. However, wheat consumption, particularly gluten proteins, also triggers certain human pathologies in an increasing fraction of the population. The adverse reactions to wheat include celiac disease (CD), non-celiac wheat sensitivity (NCWS), and IgE-mediated food allergies (Larré et al., 2011; Catassi et al., 2013; Ludvigsson et al., 2013). Altogether, these pathologies could affect up to 13% of the population (Aziz et al., 2016). Gluten is a complex group of proteins that includes the α/β-, γ-, and ω-gliadins, and the high molecular weight (HMW) and low molecular weight (LMW) glutenin subunits (GSs). They are also known as prolamins and possess a particular structure with large regions of repetitive sequences with high proportions of the amino acids glutamine and proline (Shewry, 2009). Gluten proteins represent ~80% of the total grain proteins and they are encoded by multiple genes located at complex loci found on chromosomes 1 and 6. Specifically, α-gliadins are encoded by genes at the Gli-2 loci located on the short arm of group 6 chromosomes, while ω- and γ-gliadins are encoded by genes at the Gli-1 loci on the short arm of group 1 chromosomes. Typical LMW-GSs are encoded by genes at the Glu-3 loci, genetically linked to the Gli-1 loci. Finally, HMW-GSs are encoded by genes at Glu-1 loci located on the long arm of group 1 chromosomes.

CD is the most studied of these pathologies. In CD, the unit responsible for triggering an immune response is a core of nine amino acids (epitope) present in some gluten peptides, which form stable complexes with HLA-DQ molecules (Jabri et al., 2014). From the list of the main CD-related epitopes described in Sollid et al. (2020), most are located in the gliadin fraction of gluten, with those present among the α-gliadins being the most immunogenic, followed by the ω-gliadin types (Tye-Din et al., 2010; Vriz et al., 2021). In addition, food allergy-related IgE-binding epitopes, distinct from those in CD, have also been identified in all major groups of gluten proteins, mainly in the ω5-gliadins (Battais et al., 2005).

The only available treatment for wheat-related disorders is a strict gluten-free diet (GFD), which has important health and economic issues (Jnawali et al., 2016; Vici et al., 2016; Cardo et al., 2021). Therefore, breeding wheat varieties without immunogenic peptides could be of great value, from which all people suffering from wheat-related pathologies would benefit. However, it is an extremely difficult goal to achieve through conventional plant breeding due to the high complexity of the genes encoding gliadin proteins, since each gliadin group presents multiple and highly homologous copies arranged in tandem on different chromosomes of the A, B, and D bread wheat subgenomes (Huo et al., 2018a, b). Despite this genomic complexity, RNAi wheat varieties with some, or even all three gliadin fractions strongly down-regulated have been reported (Gil-Humanes et al., 2010; Altenbach and Allen, 2011; Barro et al., 2016; Altenbach et al., 2019). These RNAi wheat lines showed very low reactivity, suggesting that they could be tolerated by patients suffering from NCWS (Haro et al., 2018) or CD (Guzmán-López et al., 2021b). However, one potential drawback associated with these RNAi lines is that to trigger the post-transcriptional silencing mechanism, it is essential to express a transgene that forms dsRNA with a high degree of homology to the target gliadin mRNAs, resulting in the interference with gliadin synthesis in the developing kernel. Consequently, these lines are classified as genetically modified organisms (GMOs) in numerous countries. This classification poses challenges, primarily in terms of the extensive and expensive regulatory processes they must undergo, particularly in the European Union (EU), and secondarily in gaining acceptance from consumers and the general public.

Gene editing technologies such as the CRISPR/Cas [clustered regularly interspaced palindromic repeats (CRISPR)/CRISPR-associated protein] system provide a revolutionary tool that overcomes the limitations of traditional plant breeding by enabling precise edits to crop genomes. Unlike RNAi technology, where target genes remain functional and the presence of a transgene is necessary, CRISPR/Cas operates with non-functional genes and does not require a transgene in the final crop variety. Moreover, CRISPR/Cas technology is subject to more open and flexible regulations, allowing for more affordable development of downstream products. This technology has been successfully applied for knocking out gliadin genes responsible for triggering the CD immune response in bread and durum wheat (Sánchez-León et al., 2018; Jouanin et al., 2019), in which the most immunogenic α-gliadins were targeted, resulting in the editing of up to 35 out of the 45 identified α-gliadin genes (Sánchez-León et al., 2018). Mutations observed in these lines included deletions and insertions (InDels), with frequencies of up to 75% in the next-generation sequencing (NGS) amplicon reads. Subsequent segregation allowed the identification of wheat lines carrying mutations but lacking transgenes, displaying phenotypes indistinguishable from the wild type (WT) (Sánchez-León et al., 2018). The multiplex gene editing approach led to a significant reduction in α-gliadin proteins, with reductions of up to 85% in gluten content detected by monoclonal antibodies (MoAbs).

The γ- and ω-gliadins are also implicated in playing a role in triggering immune responses associated with adverse reactions to wheat. They possess epitopes that can activate the immune system in individuals with CD, leading to an inflammatory response in the small intestine (Arentz-Hansen et al., 2002). Additionally, ω5-gliadins have been recognized for their significant role in wheat-dependent exercise-induced anaphylaxis (WDEIA), an uncommon form of anaphylaxis triggered when an individual exercises shortly after consuming wheat or wheat-containing products (Morita et al., 2009). Therefore, the precise editing of the genes encoding the γ- and ω-gliadins in bread wheat using CRISPR/Cas targeted mutagenesis is also crucial. Notably, targeting ω5-gliadins would be particularly valuable for individuals with WDEIA. Moreover, combining this approach with the previously described α-gliadin mutants would enhance the low-gliadin profile, further reducing reactivity for individuals with CD or other adverse reactions to wheat. In this work, we aimed to take a step forward in the targeting of wheat gliadins by precisely editing the genes encoding the γ- and ω-gliadins in bread wheat, developing a set of wheat lines deficient in either the ω-and γ-gliadins, or both gliadin fractions simultaneously. These, together with the previously reported lines deficient in α-gliadins, provide a set of low-gliadin lines that can be incorporated into breeding programs or for the development of tailor-made low-gliadin wheat varieties.

Materials and methods

Plant material and growth conditions

Wheat plants (Triticum aestivum cv. Bobwhite, designated as BW208) were cultivated in a greenhouse under a day/night cycle of 12 h each at temperatures of 24 °C during the day and 16 °C during the night. The relative air humidity was maintained at 60%.

It is important to note that the name Bobwhite refers to a collection of 129 accessions (Pellegrineschi et al., 2002) with diverse characteristics and agronomic behaviors, leading to potential confusion across different research laboratories. The entire group of 129 accessions originated from the cross CM 33203, with the pedigree ‘Aurora’//’Kalyan’/’Bluebird3’/’Woodpecker’, made by the CIMMYT bread wheat program in the early 1970s. Notably, the parent cultivar Aurora carries the 1BL.1RS (T1BL.1RS) translocation from rye, and ~85% of the sister lines inherited this translocation (Warburton et al., 2002). In the case of line BW208 used in this study, it is derived from line SH 98 26 (Pellegrineschi et al., 2002), which does not have rye translocation and was chosen for its high regeneration and transformation capabilities. The T1BL.1RS translocation is easily recognizable on acidic PAGE (A-PAGE) gels, where rye secalins replace wheat ω-gliadins, creating a distinct profile (Supplementary Fig. S1).

CRISPR/Cas constructs and wheat transformation

For the design of CRISPR/Cas constructs, 14 γ-gliadin genes and 19 ω-gliadin genes from hexaploid Chinese Spring (CS) wheat, along with 57 clones containing γ-gliadin genes and 43 clones with ω-gliadin genes (including 24 ω1,2-gliadins and 19 ω5-gliadins) from a previous study (Sánchez-León et al., 2018), were aligned to identify highly homologous regions. These regions were analyzed for PAM (protospacer adjacent motif) locations and protospacer on-target scoring using Geneious prime v2020.1.1 (Biomatters Ltd, Auckland, New Zealand). Additionally, possible off-targets across the genome were considered during protospacer selection. To that end, protospacer sequences were searched in the wheat genome (Appels et al., 2018), discarding those that showed off-targets.

CRISPR/Cas final expression vectors were constructed based on the Golden Gate assembly protocol incorporating three intermediate modules (A, B, and C) into the final backbone (Čermák et al., 2017). Each protospacer was cloned into pMOD-B2518 (Addgene #91075) or pMOD-C2518 (Addgene #91087) modules via synthetic oligo annealing with complementary overhangs to the Esp3I restriction enzyme present in both entry modules upstream of the single guide RNA (sgRNA) scaffold. Since the U6 promoter requires an initial G to initiate transcription, an extra G was added at the beginning of the 20 nt protospacers in sgGamma2, sgGamma3, and sgGamma9 (Table 1). Moreover, the first base of the target sequence in the sgRNAs targeting ω-gliadins was substituted for a G. An intermediate module that contains a wheat codon-optimized spCas9 under the control of a maize ubiquitin promoter (ZmUbi), a nuclear localization signal (NLS), and the octopine synthase terminator (OCSt) (Supplementary Fig. S2) was subcloned into a final backbone. Sanger sequencing was carried out in each cloning step to confirm the correct sequences of plasmid components.

Table 1.

Expression vectors and characteristics of the protospacer for γ- and ω-gliadins

Vector ID	sgRNA ID	Protospacer sequence (5'→3')	Length (bp)	PAM	Target gliadins
pSSLGamma16	sgGamma8	GGCTGGGGAAAAGGTTGTTG	20	TGG	γ
	sgGamma2	GTGATGGGGGAATGTTTGTTG	21	GGG, AGG	γ
pSSLGamma17	sgGamma3	GATTGTTGTTGTGGTTGATGG	21	GGG	γ
	sgGamma9	GTGTTGGGGGAATGATTGTTG	21	CGG, TGG	γ; ω1,2
pSSLOmega8	sgOmega4	GATGGTTGTTGGGGTTGCTG	20	GGG	ω1,2; γ
	sgOmega1	GGTTCATCGCCATGGCAAGG	20	AGG	ω1,2; ω5
pSSLOmega9	sgOmega2	GGCTGGGGGAATGGTTGTTG	20	TGG, GGG, CGG	ω1,2; γ
	sgOmega5	GTTATAACGTCGCTCCCAGA	20	TGG	ω1,2; ω5

For the delivery of CRISPR/Cas reagents, we have used immature scutella of wheat as a target tissue for transformation by particle bombardment, and plant selection as previously described (Pistón et al., 2008). For plant selection, the bar gene was used under the control of the ubiquitin promoter from Panicum virgatum (PvUbi2). Plants that survived the selection were transferred to the greenhouse and screened by PCR to identify the Cas9 gene using primers listed in Supplementary Table S1.

DNA extraction and PCR analysis

For DNA isolation, young leaf tissue was harvested, ground in liquid nitrogen, and stored at –80 °C. Genomic DNA was isolated using the CTAB (cetyltrimethylammonium bromide) method (Murray and Thompson, 1980). PCR was performed to detect CRISPR/Cas plasmids using primers listed in Supplementary Table S1. For PCR, 150 ng of DNA was used in a 25 μl volume reaction, consisting of 400 nM forward and reverse primers, 320 µM DNTP mix, and 0.650 U of Taq DNA polymerase (Biotools, Madrid, Spain). PCR conditions were as follows: 95 °C for 3 min followed by 35 cycles at 95 °C for 15 s, 58 °C for 30 s, and 72 °C for 1 min, with final extension at 72 °C for 7 min. All PCR products were checked by 1% agarose gel electrophoresis.

DNA sequencing and characterization of InDels

The DNAs from the WT and putative edited lines were subjected to Sanger sequencing using the primers listed in Supplementary Table S1 and ligated into the pGEMT Easy vector (Promega, Madison, WI, USA). Full-length genes were cloned into Escherichia coli DH5α cells and sequenced by the STABVIDA sequencing service (Caparica, Portugal). Geneious was used to assemble the sequences from the Sanger sequencing.

To characterize the mutations, references were created using 14 contigs for γ-gliadins and 18 contigs for ω-gliadins. These contigs were assembled from 124 Sanger clones for γ-gliadins and 91 Sanger clones for ω-gliadins obtained from the WT BW208, previously published in Sánchez-León et al. (2018). The assemblies were performed with the default Geneious algorithm. The resulting contigs were then compared with the γ- and ω-gliadin sequences from the RefSeq v1.0 assembly of wheat (Appels et al., 2018). Sanger clones from each transformed line were aligned and compared with the γ- and ω-gliadin contigs, allowing the identification of on-target InDels.

In addition, deep amplicon next-generation sequencing was used to characterize mutations on selected lines. The sequencing was conducted by Fundación Parque Científico de Madrid (Cantoblanco, Madrid, Spain) using the MiSeq system with 2 × 280 bp reads. The length of the amplicons was verified using the Agilent 2100 Bioanalyzer system (Agilent Technologies, Santa Clara, CA, USA). For the γ-gliadins, the amplicon covered the 26-mer region in the first repetitive domain. The amplification was carried out with primers described in Marín-Sanz et al. (2023) and listed in Supplementary Table S1.

Acid PAGE analysis

Gliadins were characterized in each line using the half-seed technique, where the embryo was preserved for line propagation, and the endosperm was used to extract grain proteins. For each line, 10 mature seeds were used when possible. Subsequently, the endosperms of the half-seeds were individually ground into a fine powder, and this powder was used for sequential extraction of storage proteins, followed by separation by A-PAGE as outlined (Gil-Humanes et al., 2012b).

Prolamin quantification by reverse-phase HPLC

Gliadins and glutenins were extracted from 100 mg of flour or from the half-seed endosperm and quantified by reverse-phase HPLC (RP-HPLC) following the protocol reported (Pistón et al., 2011). For this purpose, the volumes of protein extracts were applied to a LiChrospher® 300SB-C8 column (Merck, Darmstadt, Germany) using a 1200 Series Quaternary LC System liquid chromatography system (Agilent Technologies, Santa Clara, CA, USA) coupled to a DAD UV-V detector. Protein content was expressed as μg protein mg^–1 flour.

Sandwich R5 and G12 ELISA

Gluten content was analyzed by R5 and G12 MoAbs: the RIDASCREEN Gliadin R7001 (R-Biopharm, Darmstadt, Germany) and GlutenTox ELISA Rapid G12 (Hygiena, CA, USA) kits were used, respectively. To provide accurate results, the gliadin extraction and ELISAs were repeated on three different days and each assay was performed in duplicate. Total gluten content (ppm or mg kg^–1) was estimated by multiplying the obtained gliadin concentration by a factor of 2. These analyses were carried out by the Laboratorio de Análisis de Gluten UPV/EHU—GLUTEN3S from the University of the Basque Country (Victoria-Gasteiz, Spain).

Statistical analysis

Significant differences between two-sample group comparisons were determined using Welch’s t-test for parametric data or the Mann–Whitney–Wilcoxon test for non-parametric data. The one-way ANOVA and Tukey’s test analyses were performed to compare multiple groups for parametric data. The statistical analyses were conducted using R (R Core Team, 2020). For the calculation of Pearson correlation coefficients, the pandas library was used (McKinney, 2010).

Results

Multiplex genome editing of γ- and ω-gliadins by CRISPR/Cas9

The structure of wheat γ- and ω-gliadin proteins is illustrated in Fig. 1A, indicating the positions of CD DQ2.5 epitopes and IgE-binding sites, which predominantly reside in the repetitive region of both protein families, rich in proline and glutamine residues. It should be noted that, for ω-gliadins, the two DQ2.5 epitopes map on ω1,2-gliadins but not on ω5-gliadins (Fig. 1B). In addition, IgE-binding sites are mostly on ω5-gliadins. Moreover, both γ- and ω-gliadins contain R5 and G12 MoAb-binding sites that are used to quantify gluten content in foods. In this study, we designed eight sgRNAs targeting γ- and ω-gliadin genes (Fig. 1A; Table 1). The design was based on (i) the sequence annotated in CS, which comprises 14 and 19 genes for γ- and ω-gliadins, respectively (Huo et al., 2018a), and (ii) the γ- and ω-gliadin sequences from the bread wheat cultivar BW208 obtained through Sanger sequencing (Sánchez-León et al., 2018). The BW208 contigs and the CS γ/ω-gliadin genes used as reference sequences for designing and identifying the InDels in edited plants are presented in Fig. 1C and D.

Fig. 1.

The γ- and ω-gliadin genes. (A) Schematic diagram of a typical γ- and ω-gliadin gene indicating the different protein domains according to Qi et al. (2006). Target sequences for the sgRNAs are represented by colored arrows. The position of the DQ2.5 epitopes and the 26-mer peptide is indicated for the γ-gliadin. (B) The presence of hits and sequences containing the specified CD epitopes in the ω-gliadins of both Chinese Spring (CS) and BW208 genotypes. One mismatch was allowed in relevant positions. * Indicates pseudogenes in CS. Maximum-likelihood tree of γ- (C) and ω-gliadins (D). BW208 contigs and the CS γ/ω genes were used as reference sequences for InDel identification. (E) Number of sgRNAs (without mismatches) mapping on γ/ω contigs/genes.

For the design of the sgRNAs, the binding sites of the R5 MoAb were not considered as a primary target, since this antibody was raised to detect and quantify the gluten content (Valdés et al., 2003) but provides limited information on immunogenicity. This is because only 22.5% of the DQ epitopes recognized by CD4⁺ T cells (Sollid et al., 2020) present binding sites for R5. Specifically, only 40% of the DQ2.5 epitopes described in γ-gliadins present R5-binding sites and, of the two DQ2.5 epitopes for ω-gliadins, only one presents binding sites for this MoAb. Most interestingly, none of the DQ2.5 epitopes described in the α-gliadins presents R5-binding sites, but the α-gliadin epitopes are considered the most immunogenic and play a pivotal role in the mechanism of pathogenesis in CD (Koning et al., 2005; Tye-Din et al., 2010). The G12 monoclonal antibody is also used to quantify the gluten content in foods (Morón et al., 2008). In contrast to R5, G12 was raised against the highly immunogenic 33-mer peptide present in the wheat α-gliadins, but neither the 33-mer peptide or any of its variants (Tye-Din et al., 2010; Marín-Sanz et al., 2023) are present in the γ- or ω-gliadins. Therefore, the catalog of DQ2.5, DQ8, and DQ8.5 restricted epitopes, and their position in the γ- or ω-gliadin genes, was considered as the primary target for designing the RNA guides instead (Sollid et al., 2020). In addition, since both DQ2.5 epitopes of the ω-gliadins are present in ω1,2- but not in the ω5-gliadins, the sgRNAs were intended to target this ω1,2 fraction of gliadins as much as possible (Table 1). The sgRNAs were strategically designed in conserved regions containing the DQ2.5 CD restricted epitopes within the repetitive domain to disrupt reading frames and hinder protein production (Fig. 1A). For this, four sgRNAs were designed for the γ-gliadin genes, two for the ω1,2-gliadins, and two for the ω5-gliadins. Notably, the sgRNAs targeting ω1,2-gliadins also matched some γ-gliadin genes, and the sgGamma9 targeted both γ-gliadins and ω1,2-gliadins (Table 1). Most of the sgRNA target sequences were present in more than one hit in the γ/ω-gliadin genes (Fig. 1E). For example, sgOmega4 occurs in five, four, and two instances in the ω1,2-gliadin contigs ω2, ω7, and ω10, respectively. For the ω5-gliadins, the most conserved regions were found at the 5' and 3' ends, leading to the design of two sgRNAs targeting those regions (Fig. 1A). However, this complicates the end-to-end Sanger sequencing strategy for the identification of the mutations. Next, possible off-targets across the genome were considered during protospacer selection, and all eight sgRNA sequences were searched in the wheat reference genome (Appels et al., 2018; Juhász et al., 2018), the α-gliadin genes from BW208 (Sánchez-León et al., 2021), and CS (Huo et al., 2018b). Two possible off-targets for sgOmega2 and sgOmega4 were found in the TraesCSU02G257591 gene from the reference genome, which corresponds to a putative α-gliadin, but no off-targets were found in the same gene family from BW208, albeit they were present containing 1–3 mismatches. In addition, no off-targets were found in the other HMW-GS and LMW-GS genes.

Due to the requirements of the wheat U6 promoter, modifications were made to the first nucleotide in some sgRNAs, either by changing the first base or by adding an extra ‘G’ base (Table 1). As expected in these gene families, the protospacer sequence is flanked at the 3' end by different PAM sequences due to the polymorphism in the γ- and ω-gliadin genes (Table 1). These sgRNAs were combined in four CRISPR/Cas9 expression vectors (Table 1), each containing two independent expression cassettes with an sgRNA driven by the wheat Pol III promoter U6 (TaU6), and a wheat codon-optimized Cas9 gene under the control of the Zea mays ubiquitin promoter (ZmUbi1) (Supplementary Fig. S2).

Later, CRISPR/Cas expression vectors were introduced into the immature scutellum of the hexaploid bread wheat BW208 cultivar via particle bombardment as described in Sanchez-Leon et al. (2018). After the in vitro selection process, 152 lines that were successfully regenerated were subjected to PCR analysis for Cas9 gene identification (Cas9+). A total of 59 wheat Cas9+ lines were identified, and allowed to reach maturity (Table 2) and to self-fertilize.

Table 2.

Plasmid combinations, number of wheat lines containing the Cas9 gene, and number of lines showing alteration in the γ- and ω-gliadin region

Plasmid combination	Target gliadins	Cas9 PCR+ lines	Lines edited
pSSLGamma16	γ-Gliadins	13	5
pSSLGamma17	γ-Gliadins	8	0
pSSLOmega8	ω-Gliadins; γ-gliadins	24	8
pSSLOmega9	ω-Gliadins; γ-gliadins	3	3
pSSLOmega9+pSSLGamma17	γ-Gliadins, ω-gliadins	11	4

Genome editing of γ- and ω-gliadins provides multiple distinctive protein profiles of wheat gliadins

Due to the hexaploid nature of bread wheat and the presence of multiple copies of target genes across various chromosomes, achieving simultaneous targeted mutations in all gliadin alleles poses a considerable challenge. To address this complexity, we decided to carry out an analysis of the T₁ grains from all 59 T₀ lines, aiming to identify edited protein patterns. To that end, gliadins from 10 T₁ half-grains were extracted from each of the T₀ lines, and the gliadin pattern was analyzed using A-PAGE and compared with the WT line. In total, 590 offspring lines were analyzed. As anticipated, a segregating pattern was observed in most lines (an example is shown in Supplementary Fig. S3), facilitating the identification of 20 T₀ lines (Table 2) in which the protein pattern showed significant alterations in the T₁ seeds in the γ- and/or ω-gliadin regions. Of the different T₀ lines generated, four were edited with the pSSLGamma16 plasmid (Table 2), showing an efficient reduction of the γ-gliadins. Notably, these lines exhibited differences in protein profiles despite containing the same plasmid (Fig. 2A). One line displayed the absence of a single band, two lines exhibited the absence of several bands, and one line showed the absence of the entire γ-gliadin region (Fig. 2A). In these lines, no alterations in the protein profile in the ω-gliadin region were observed. Conversely, the eight lines identified with the pSSLGamma17 plasmid (Table 2) showed no alterations in the gliadin profile in the T₁ grains.

Fig. 2.

A-PAGE gliadin profiles of edited wheat lines. A-PAGE gliadin profiles of the offspring of (A) T₀ lines transformed with pSSLGamma16, (B) pSSLOmega8, and (C) pSSLOmega9. The red arrows and red boxes indicate the absence of bands or weaker ones, and the green arrows indicate new bands in the edited lines.

The plasmids pSSLOmega8 and pSSLOmega9 harbor sgRNAs targeting the ω-gliadins, with an additional feature: two of the sgRNAs also target the γ-gliadins. A total of 24 and three T₀ lines were identified, respectively, with each of these plasmids (Table 2). Among them, eight and three lines, respectively, exhibited alterations in the pattern of ω- and/or γ-gliadins. Moreover, wheat lines transformed with these constructs displayed distinct gliadin profiles (Fig. 2B, C; Supplementary Fig. S4). For the pSSLOmega8 construct, all lines showed a strong reduction of the ω1,2-gliadins but not of the ω5-gliadins. Additionally, a reduction in γ-gliadins was observed across all lines; particularly noteworthy in the case of Omega-2, Omega-12, and Omega-35 wheat lines (Fig. 2B). In the case of the pSSLOmega9 construct, two lines, Omega-56 and Omega-57, exhibited the absence of ω1,2-gliadins and a significant reduction in γ-gliadins (Fig. 2C). This gliadin profile closely resembled that obtained with the plasmid pSSLOmega8. The Omega-43-derived lines displayed the most substantial changes in their A-PAGE gliadin profiles compared with the WT, presenting a significant number of absent bands for ω1,2- and ω5-gliadins in most of its progeny, accompanied by a reduction in γ-gliadins (Fig. 2C). In this line, one new protein band was also evident in the offspring lines for which only γ-gliadins were targeted.

Finally, 11 lines were identified as co-transformed with plasmids pSSLOmega9 and pSSLGamma17. Among these, four lines displayed alterations in the gliadin profile (Supplementary Fig. S5). Upon T₁ analysis, these lines exhibited missing bands in the ω1,2-gliadins, ω5-gliadins, and in the γ-gliadin region, along with additional protein bands in the ω5-gliadin fraction for one line that were not present in the WT (Supplementary Fig. S5).

Alterations in the gliadin profile are consequences of InDels in γ- and ω-gliadin genes

Sanger sequencing and NGS were carried out using the primers listed in Supplementary Table S1, to assess the efficiency of the CRISPR/Cas expression vectors in inducing targeted mutations. The γ-gliadin Sanger sequences from the WT were processed and assembled, resulting in 14 contigs used as references for detecting mutations in the edited plants. The phylogenetic tree of these 14 contigs are shown in Fig. 1C alongside the 14 γ-gliadin genes from CS, enabling the identification of contig subgenomes based on the reported information from CS genes (Huo et al., 2018a). As expected, considerable variability was observed in the BW208 WT sequences due to the presence of single nucleotide polymorphisms (SNPs) and differences in the number of repetitive motifs. Concerning the ω-gliadins, challenges in cloning these genes have been previously discussed in the literature (Hsia and Anderson, 2001; Hassani et al., 2008; Anderson et al., 2009). Despite this, we successfully amplified a representative number of ω-gliadin clones from both the WT and edited lines. The sequences obtained from the WT were assembled into 18 contigs and clustered with the ω-gliadin genes from CS, as depicted in Fig. 1D. The identification of ω1,2- and ω5-gliadins was achieved by searching for specific translated peptide sequences, as reported by Altenbach et al. (2018), and these groups were well separated in the phylogenetic tree (Fig. 1D).

The sequencing analysis of both γ- and ω-gliadins corroborated the presence of CRISPR/Cas-induced InDels in the 20 lines previously identified using A-PAGE gels, employing any of the sgRNAs detailed in Table 1. Conversely, 10 T₀ lines that survived the in vitro selection process but tested negative for Cas9, and 10 T₀ lines containing Cas9 but with no alterations in the A-PAGE profile were also subjected to sequencing, but no InDels were detected in any of these lines.

The editing events in the γ- and ω-gliadins were characterized in detail in various T₀ and T₁ edited lines. Regarding the γ-gliadins, none of the lines carrying the expression vector pSSLGamma17 showed edited reads by Sanger sequencing. However, sequencing of the γ-gliadin genes in plants with pSSLGamma16, pSSLOmega8, and pSSLOmega9 detected deletions ranging from nucleotides –1 to –561, with editing events of nucleotides –1, –2, –21, –24, and –559 being the most abundant, all observed in pSSLGamma16-derived lines (Fig. 3A, B, D). Among the lines transformed with pSSLGamma16 and analyzed for γ-gliadin mutations, only one line presented insertions exceeding 40 bp, while one line from the pSSLOmega8 construct exhibited a 1 bp insertion (Fig. 3A, B, D). The distribution of mutations across contigs of γ-gliadin genes was examined. Specifically, the γ4—like γD2 from CS—exhibited aligned mutated clones in each set of lines grouped by T₀ (Supplementary Fig. S6). In the group of Gamma-39 lines, all the clones aligned to this reference contig presented InDels in the corresponding sgRNAs of pSSLGamma16 (Supplementary Fig. S6). Similarly, the γ2 contig, resembling γB6 of CS, displayed mutations in many of these line groups, with a substantial proportion in Gamma-173 (pSSLGamma16) and Omega-2 (pSSLOmega8) groups (Supplementary Fig. S6). The inheritance of mutations with the pSSLGamma16 construct was also characterized in the offspring of T₁ Cas9-negative plants (Fig. 3C). Overall, it was possible to identify lines derived from Gamma-39, Gamma-96, Gamma-173, and Gamma-229 carrying mutations in six, seven, eight, and six γ-gliadin genes, respectively, and not containing the CRISPR/Cas reagents.

Fig. 3.

CRISPR/Cas9-mediated InDels in γ-gliadin genes of wheat lines. (A) The frequency of InDels in Sanger clones is expressed as a percentage, calculated by dividing the total events within each mutation length interval by the overall number of mutations identified in γ-gliadins of pSSLGamma16, pSSLOmega8, and pSSLOmega9 lines. (B) Examples of deletions in sgGamma8, sgGamma2, sgOmega4, and sgOmega2. (C) The heritability of mutations in γ-gliadins across generations of Cas9-negative lines. (D) Deletions found in sgRNA hits with one mismatch (highlighted in blue).

The analysis of the ω-gliadin clones has allowed us to identify several types of deletions in their sequences, although the number of deletions might be underestimated due to the aforementioned ω-gliadin sequence characteristics. We characterized editing events in sgRNAs, as depicted in Fig. 4A, B, in lines transformed with pSSLOmega8 and pSSLOmega9 plasmids. Notably, the most frequent deletions, particularly in pSSLOmega8 plants, exhibited lengths exceeding 100 bp, revealing fragment deletions between pairs of sgRNAs distributed across the gene sequence (Fig. 4A; Supplementary Fig. S7). As for γ-gliadin genes, certain mutations in the ω-gliadin genes of the Cas9-positive pSSLOmega9 T₀ line were inherited by its progeny (Fig. 4B).

Fig. 4.

CRISPR/Cas9-mediated InDels in ω-gliadin genes of wheat lines. (A) Illustrative examples of deletions in sgOmega4 and sgOmega2. (B) The heritability of mutations in ω-gliadins across generations of Cas9-negative lines. (C) The frequency of InDels in Sanger clones is expressed as a percentage, calculated by dividing the total events within each mutation length interval by the overall number of mutations identified in ω-gliadins of pSSLOmega8 and pSSLOmega9 lines. (D) InDels identified in γ-gliadin genes from Omegamma lines.

Lines generated by combining pSSLGamma17 and pSSLOmega9 constructs were denoted as Omegamma lines. These lines were screened by NGS amplicon sequencing for the γ-gliadin genes, indicating the presence of significant deletions in these genes. In these lines, we have found independent deletions with the two sgRNAs of plasmid pSSLGamma17, and in combination with the sgRNAs from plasmid pSSLOmega9 (Fig. 4D). Specifically, in the γ3 contig, there was a deletion between sgOmega2 and sgGamma9, underscoring the potential use of paired sgRNAs to delete specific undesired fragments from target genes.

As previously noted, it is important to highlight that not all sgRNAs exhibit the same editing efficiency. For the sgOmega1 and sgOmega5 sgRNAs it was not possible to score the efficiency as their target sites overlapped with sequencing primers (Supplementary Table S2). However, protein profiles obtained with A-PAGE gels clearly show the missing bands in the ω5 region in several wheat lines transformed with plasmids containing these sgRNAs (Fig. 2). Conversely, sgOmega4 and sgOmega2, present in constructs pSSLOmega8 and pSSLOmega9, respectively, demonstrated notably high editing efficiencies. In particular, sgOmega2 achieved values as high as 50%.

Mutations in γ- and ω-gliadin genes lead to a drastic decrease of the gluten content in edited plants

Homozygous lines edited with the different constructs were selected and analyzed by RP-HPLC to quantify the gluten protein fractions (Fig. 5A, B). The RP-HPLC results revealed a substantial reduction in the CRISPR/Cas target fractions, albeit with considerable variability depending on the line and construct used. Lines generated using plasmids pSSLGamma16 and pSSLOmega8 exhibited the most noteworthy reductions in the γ-gliadin fraction. Conversely, lines derived from plasmid pSSLOmega9 showed the most significant decreases in ω-gliadins, together with some lines from the pSSLOmega8 plasmid. Notably, these lines also exhibited a significant reduction in the α-gliadin fraction, an outcome not observed for all the lines (Fig. 5A, B). In addition to the decrease in gliadins, several lines showed a significant increase in glutenins (Fig. 5B). This phenomenon was observed for the pSSLGamma16 and pSSLOmega8 constructs, in which the strong reduction of ω- and γ-gliadins led to the increase of HMW-GSs and LMW-GSs. Interestingly, some lines exhibiting a significant reduction of the α-gliadins showed a decrease in the content HMW and/or LMW glutenin subunits, such as for lines AE608 (pSSLOmega9) and AC389 (pSSLGamma16) (Fig. 5B).

Fig. 5.

Prolamins and gluten content in edited wheat lines. (A) RP-HPLC gliadin profile in edited lines and the WT. (B) Quantification of gliadin fractions in edited lines compared with the WT line. (C) The percentage of reduction of gluten content (%) measured by R5 and G12 MoAbs of T₂ lines edited with constructs pSSLGamma16, pSSLOmega8, and pSSLOmega9. *P<0.05; **P<0.01; ***P<0.001.

The gluten content in parts per million (ppm) of homozygous edited lines was analyzed with the R5 MoAb developed by Valdés et al. (2003). According to the Codex Alimentarius, alcohol-soluble gluten proteins (gliadins) should be extracted with ~40–70% ethanol and quantified using an immunochemical method, for which the R5 ELISA is recommended (FAO/WHO Codex Alimentarius Commission, 2008). The gliadin content of gluten is generally taken as 50%, so the total gluten content (ppm) has to be calculated by multiplying the gliadin content by a factor of 2. The gluten content of the WT line is ~190 540 ppm (190 540 mg kg^–1) measured with R5 MoAb (Fig. 5C). Notably, all edited lines exhibited significantly lower gluten content values compared with the WT, with variations depending on the specific CRISPR construct and line. Those lines derived from the pSSLOmega8 plasmid consistently demonstrated lower gluten contents than those obtained with the pSSLGamma16 and pSSLOmega9 plasmids. Among them, lines AD673 and AD700 displayed gluten contents of 6383 ppm and 4447 ppm, respectively, indicating a reduction of 96.7% and 97.7%, respectively, in comparison with that of the WT. Another noteworthy line is AD800, revealing a 90.5% reduction in gluten content. Interestingly, this group of lines originated from the pSSLOmega8 CRISPR construct and achieved substantial reductions, in particular showing a large reduction of ω1,2-gliadins (Fig. 2), which contain the CD-related epitopes. In contrast, lines AD407, AD408, AC389, and AC391, derived from the pSSLGamma16 construct, exhibited gluten content reductions from 66% to 79%. A remarkable correlation was found between gluten content (ppm) measured by R5 MoAb and the content of each gliadin fraction determined by RP-HPLC: the γ-gliadin presented the highest R² value (0.93), followed by ω-gliadins with 0.81. The gluten content was also measured with the G12 MoAb, which is an alternative for determining gluten content in foods (Morón et al., 2008). As for R5, the pSSLOmega8 lines exhibited lower gluten content than the pSSLGamma16 or pSSLOmega9 lines (Fig. 5C). AD673 and AD700 were still the lines with the highest gluten reduction, reaching values of 17 692 ppm and 22 603 ppm, respectively. Notably, all the lines assayed presented reductions of >50% in their gluten content measured by this MoAb, highlighting the effectiveness of these CRISPR/Cas constructs (Fig. 5C). Despite the significant reduction of gliadin/gluten content, the shape of the grains of the transformed lines was like that of the WT, including the lines coming from Omega-43 T0, which had a high reduction of ω- and γ-gliadins (Supplementary Fig. S8).

Combining CRISPR/Cas targeted mutations by cross-breeding

Some lines from the progeny of the Omega-43 line were chosen and used as parentals in crosses with previously established α-gliadin-deficient CRISPR lines (Sánchez-León et al., 2018). Specifically, Cas9-positive (V730) and Cas9-negative (V653) α-gliadin-deficient CRISPR lines were selected for this purpose. The resulting offspring from these crosses were analyzed by A-PAGE (Fig. 6A), cultivated in a greenhouse, self-fertilized, and the descendants of two distinct crossings, V653 (Cas9–)×AE412 (Cas9+) and V730 (Cas9+)×AE412 (Cas9+), respectively, exhibited a non-segregating A-PAGE profile in their F₃ progeny. Subsequently, these lines were subjected to RP-HPLC and R5 and G12 analyses.

Fig. 6.

Crossing wheat edited lines. (A) The A-PAGE gliadin profile of the crosses V653×AE412 and V730×AE412. Gliadin quantification (B) and its RP-HPLC profiles (C) of the offspring of AM500 and AM501. (D) Gluten content (ppm) measured by R5 and G12 monoclonal antibodies of AM500 and AM501 offspring.

As depicted in Fig. 6A, crosses involving the V730 (Cas9+) parent yielded more distinct gliadin profiles, with fewer bands and weaker intensity, compared with those with the V653 (Cas9–) parent, particularly in the ω- and γ-gliadin regions. Nevertheless, bands persist in the ω5-gliadin region for both crosses. In line with the differences observed in the gliadin profiles (A-PAGE), RP-HPLC quantification of gliadin fractions showed significant reductions in both crosses mainly for the α- and γ-gliadins, but also for the ω-gliadins compared with the WT (Fig. 6B, C). Both crosses showed a reduction in the LMW-GSs, and the V730×AE412 line also presented a significant increase in HMW-GSs (Fig. 6B). Finally, R5 results confirmed a significant reduction of gluten content in the offspring of both AM500 and AM501 crosses (Fig. 6D), with values of 36 804 ppm and 26 345 ppm, respectively. Both lines also presented a significant reduction of the gluten content measured by G12. By this MoAb, the AM501 line presented less content than its parent V730, reaching a mean value of 26 455 ppm (Fig. 6D).

Discussion

Wheat is a common food allergen and, when consumed, it can trigger adverse reactions in a significant and increasing percentage of the population (Wieser et al., 2020). While a GFD is essential for managing gluten-related disorders, there are some disadvantages associated with this diet (Vici et al., 2016; Marciniak et al., 2021). Breeding wheat varieties to lack immunogenic gluten peptides can lead to several benefits to alleviate these pathologies, including access to a broader range of food choices, improved quality of life, increased consumer acceptance, reduced costs, increased sustainability, and improved nutritional value. However, obtaining such immunogenic-free wheat is highly challenging due to the inherent characteristics of wheat and the complex nature of gluten.

Gliadins are the major proteins responsible for triggering adverse reactions to wheat in CD (Arentz-Hansen et al., 2002). These proteins are highly resistant to digestion in the gastrointestinal tract, leading to the formation of small peptides that can trigger the immune response (Shan et al., 2002). However, not all gliadins exhibit the same stimulatory capacity. Specifically, the α-gliadins harbor the 33-mer peptide, consisting of six overlapping copies of three adaptive immune response-related CD epitopes, which are recognized as the most stimulatory in the immunogenic response (Tye-Din et al., 2010). Additionally, the α-gliadins also contain the p31-43 peptide, associated with innate immunity (Maiuri et al., 2003). Therefore, this group of gluten proteins has been the first target of CRISPR/Cas to develop wheat lines that lack immunogenicity (Sánchez-León et al., 2018). Although γ-gliadins have less stimulatory capacity compared with α-gliadins (Tye-Din et al., 2010), they harbor a substantial number of epitopes (Marín-Sanz et al., 2023). Unlike α-gliadins, where ~50% of the genes lack CD epitopes, ~85% of γ-gliadin genes contain them (Marín-Sanz et al., 2023).

On the other hand, wheat ω-gliadins include a highly DQ2.5 restricted stimulatory epitope, with two versions known as DQ2.5_glia_ω1 and DQ2.5_glia_ω2 (Tye-Din et al., 2010; Sollid et al., 2020). In CS and BW208, these CD epitopes are present in the ω1,2-gliadins but absent in ω5-gliadins (Fig. 1B). It is interesting to highlight that the gene sequences carrying these epitopes are found in both the D and A subgenomes. However, in the D subgenome, these sequences represent functional genes, while in the A subgenome they are pseudogenes, suggesting that, analogous to α- and γ-gliadins (Molberg et al., 2005; Ozuna et al., 2015; Marín-Sanz et al., 2023), the D subgenome is likely to harbor the most immunogenic CD epitopes from ω-gliadins. In contrast, ω5-gliadins exhibit a higher abundance of IgE-binding epitopes than ω1,2-gliadins (Matsuo et al., 2015). Therefore, the ω1,2-gliadins could be more relevant to CD, while ω5-gliadins are associated with WDEIA. This has important implications because CRISPR/Cas targeted mutations in specific ω-gliadin fractions will facilitate the development of ‘à la carte’ products for those health conditions.

CRISPR/Cas9-mediated editing of γ- and ω-gliadins

In a previous study, we used CRISPR/Cas to introduce targeted mutations in the α-gliadin genes of durum and bread wheat (Sánchez-León et al., 2018). This study extends our approach by employing multiplexed CRISPR/Cas to introduce mutations in the γ- and ω-gliadin genes. To design the sgRNAs, the sequences of the γ- and ω-gliadin genes from CS and the BW208 cultivar were used. Both groups of genes are structurally different and, within each family, there is a large variability between them, making it impossible to find highly conserved regions for the design of individual sgRNAs for targeting all genes. Consequently, four constructs harboring eight RNA guides were designed to cover the highest number of genes possible, many of them targeting both gene families. This genetic variability was also evident when the same protospacer sequence was flanked at the 3' end by different PAM sequences. Due to the repetitive nature of these genes, some sgRNAs are repeated in multiple hits within certain genes, which could provide enhanced coverage for targeted mutations.

Using the designed constructs, 59 independent lines containing the Cas9 gene (Cas9+) were identified, out of a total of 152 that survived in vitro selection. The T₁ progeny of 20 lines showed alterations in the γ- and/or ω-gliadin regions, as determined by A-PAGE gels. Subsequent Sanger sequencing or NGS of these 20 lines confirmed mutations in the γ- and/or ω-gliadin genes. Overall, the editing efficiency was 34% (20/59) for both gene families. In contrast, 10 randomly selected lines containing the Cas9 gene but lacking alterations in the γ- and ω-gliadin regions were sequenced, and no mutations were found in these regions. This shows the utility of A-PAGE screening in rapidly identifying mutants in these target genes. In addition, A-PAGE screening facilitated the detection of segregant lines for the target fractions, highlighting that probably not all alleles carry the mutations in T₀ plants. Nevertheless, the selection of T₁ profiles enabled the generation of homozygous lines exhibiting diverse silencing profiles (Fig. 2; Supplementary Figs S3–S5).

The spectrum of mutations obtained with the different sgRNAs and plasmids was complex, but in most cases, large deletions predominate over small deletions and/or insertions. This could be largely explained by the fact that sgRNAs have more than one hit in the target genes, and cleaving multiple sites could result in large deletions. This is perfectly illustrated by the mutations found for ω1,2-gliadins in the Omega-25 and Omega-35 lines (Supplementary Fig. S7). While contig ω13 has only one hit of the sgRNA and provides a deletion of 24 bp, contig ω8 has four hits of the sgRNAs, leading to larger deletions involving from one to four sgRNA sites. In general, the more sgRNA sites, the larger the deletions. However, the combined mapping of several sgRNAs in the same γ- and/or ω-gliadin gene also causes small mutations (1–4 nt) and not always the deletion of the fragment in between. The frequency and length of the inserted fragments in both γ- and ω-gliadins were low, in contrast to those previously described for the α-gliadins, where the frequency of insertions was higher, despite using only one sgRNA (Sánchez-León et al., 2018). Moreover, small or large insertions have not been found in single cases but in combination with deletions in different sgRNAs. Yu et al. (2024) also used CRISPR/Cas to knock out wheat γ- and ω-gliadin genes, detecting deletions in the range of 18 bp and 111 bp, and large deletions covering even several functional genes. In previous CRISPR work on α-gliadins, gene losses were also reported in durum and bread wheat (Guzmán-López et al., 2021a). This could indicate that the loss of long chromosome fragments, covering several genes, is a common phenomenon when several sgRNAs are used against gene families, with high homology and structurally close in the genome.

When multiple sgRNAs target the same gene, perfect cleavages can result in the complete excision of the fragment between them. This feature can be widely exploited to remove highly immunogenic fragments and replace them with peptide sequences that provide functional proteins of the same characteristics but lacking CD-related epitopes. In this work, we report examples of fragment excision affecting several γ- and ω-gliadin genes. In many cases, this excision provided novel protein bands as they are evidenced in the protein profiles of edited lines. Although targeted replacement is still very challenging in plant genome editing, several methods have been developed to exploit non-homologous end-joining for high-efficiency targeted insertion in rice and Setaria viridis at high efficiency (Lu et al., 2020; Kumar et al., 2023). In the case of wheat gliadins, this challenge is even greater as they are multigene families with various immunogenic complex variants (Tye-Din et al., 2010; Ozuna et al., 2015; Huo et al., 2018b; Marín-Sanz et al., 2023). In the gliadin genes, the regions with the most immunogenic epitopes are found in the proline- and glutamine-rich repeat regions, primarily clustered around the 33-mer and 26-mer immunogenic peptides present in the α- and γ-gliadins, respectively (Shan et al., 2002, 2005). For example, there are α-gliadin variants containing 1–5 CD epitopes, variants that are even more abundant than the 33-mer (Sánchez-León et al., 2021), but also contain other variants without CD epitopes whose sequence can serve as a template for this replacement. However, the number of epitopes present in γ-gliadins is much greater than in α-gliadins and extends beyond the 26-mer (Marín-Sanz et al., 2023). In these cases, synthetic biology, by completely redesigning the CD immunogenic region, will play a key role. Several works have already shown that the substitution of certain amino acids present in the nine amino acid core sequence of the DQ2.5 restricted epitopes, recognized by antigen-presenting cells, abolishes immunogenicity (Anderson et al., 2006; Ruiz-Carnicer et al., 2019).

Notably, contig γ9 did not show matches for the sgRNAs from the pSSLGamma16 construct, yet several lines exhibited mutations in that contig. Interestingly, the sequence of these sgRNAs was present in the γ9 contig but contains 1–2 mismatches distal to the PAM in regions coincident with the deletions (Fig. 3D). It has been reported that mismatches distal to the PAM are tolerated, although cleavage efficiency steeply decreases as the number of mismatches increases (Modrzejewski et al., 2020). Thus, with one mismatch, the efficiency could decrease by 41%, and with two mismatches by up to 75% (Modrzejewski et al., 2020). In addition, cleavage efficiency also depends on the position of the mismatch, since in mismatches proximal to the PAM the recognition of off-targets is performed by a different conformation of the guide RNA–DNA duplex, which prevents Cas9 activation (Bravo et al., 2022). In contrast, target DNAs with mismatches distal to the PAM are stabilized by the reorganization of a loop in the Ruvc domain, facilitating DNA cleavage, although at a lower efficiency (Bravo et al., 2022).

Surprisingly, none of the eight lines transformed with the pSSLGamma17 plasmid showed mutations in either of the two sgRNAs. However, mutations were detected at these sgRNA cleavage sites in the γ-gliadin genes when the pSSLGamma17 plasmid was used in combination with the pSSLOmega9 plasmid. Moreover, both sgRNAs from the pSSLGamma17 plasmid introduced mutations either individually or in combination with the sgRNAs from the pSSLOmega9 plasmid (Fig. 4D). This highlights the complexity of CRISPR/Cas9 effectiveness, even when the Cas9 enzyme is integrated into the plant genome. A possible explanation could be related to Cas9 expression levels; double transformation with two plasmids containing Cas9 might result in higher endonuclease levels, enhancing editing efficiency with these sgRNAs. Nevertheless, more detailed studies are necessary to better understand the CRISPR/Cas system’s effectiveness when targeting gene families such as gliadins, which is beyond the scope of this work.

Gluten content in edited wheat plants

CRISPR/Cas targeted mutations in the γ- and ω-gliadin genes led to significant changes in the A-PAGE protein profiles of the edited lines. In total, 24 homozygous lines showing changes in gliadin profiles and mutations in these genes were identified, of which 14 were analyzed by RP-HPLC and/or R5 and G12 MoAbs. Protein profiles and quantification by RP-HPLC confirmed that the largest decreases in gliadin fractions occurred in γ-gliadins and ω1,2-gliadins, and finally to a lesser extent in ω5-gliadins, as expected from the design of the sgRNAs to cover the gliadins containing CD epitopes. Some lines analyzed by RP-HPLC showed a significant decrease in the α-gliadin fraction. Some of these lines were edited with sgOmega2 and, although perfect hits were not found in α-gliadin genes of BW208, it was present with 1–3 mismatches outside the seed sequence. As discussed for the γ9 contig, the presence of one mismatch distal to the PAM could be the cause of this decrease in α-gliadins in these specific lines. However, a post-transcriptional regulation or alteration in the processing of the defective proteins cannot be ruled out either. Chen et al. (2022) reported that the lgp1 mutation in a single γ-gliadin gene leads to defective cleavage of the signal peptide, resulting in the accumulation of an excessive amount of unprocessed γ-gliadin and a reduced level of gluten, which disrupts the endoplasmic reticulum.

The decreasing gliadin fractions led to the reduction of gluten measured by R5 and G12 MoAbs. In the food industry, R5 is used to determine the presence of gluten, and it is useful for screening foods to meet gluten-free standards. This technique is not free of controversy, as the gluten content is not determined but rather the gliadin content, and then the total gluten content (ppm) has to be calculated by multiplying the gliadin content by a factor of 2 (Wieser and Koehler, 2009). However, considerable caution must be observed when considering immunogenicity, since the recognition sites of this MoAb are distributed in the gliadin fractions, being only present in 22.5% of DQ CD epitopes. Despite the limitations of R5 providing immunogenicity information, 11 lines transformed with pSSLGamma16, pSSLOmega8, and pSSLOmega9 presented reductions in gluten content as high as 97.7% by this MoAb. These lines also presented strong reductions in the gluten content measured by G12. In general, the results were similar when G12 was used, although, in this case, the gluten content as ppm was higher than that detected by R5. These discrepancies are probably because R5 was raised against rye ω-secalins, homologous to wheat ω-gliadins, while G12 was raised against wheat α-gliadins (Morón et al., 2008), and these lines retain very high values of α-gliadins which are detected by G12.

In a recent study, CRISPR/Cas was also used to reduce γ- and ω-gliadins in the cultivar Fielder (Yu et al., 2024). These authors reported that their best line showed a gluten content determined by R5 of 4.58 × 10⁴ ppm (i.e. 45 800 ppm). In our work, seven of the 11 lines analyzed by R5 showed values <45 800 ppm, and two of them showed values of 6383 popm and 4447 ppm, representing seven and 10 times less gluten content, respectively, than their best line. For G12, they report values of 6.29 × 10⁴ ppm (i.e. 62 900 ppm) for their best-edited line. In our work, nine of the 11 lines analyzed by G12 showed values <62 900 ppm, and the two best lines showed values of 22 603 ppm and 17 692 ppm, representing, respectively, 2.8 and 3.6 times less gluten than their best line. Yu et al. (2024) suggested that the Bobwhite genotype used in our studies contains the T1BL.1RS from rye and that is why the R5 values are lower than in Fielder. This suggestion is not correct, as both genotypes presented comparable contents of gluten measured by this MoAb (Supplementary Fig. S1). There is a collection of 129 Bobwhite-derived wheat lines (Pellegrineschi et al., 2002), and ~20 lines do not contain the translocation (Warburton et al., 2002). This is perfectly clear in their protein profiles, where the secalins from the rye genome are visible in the genotypes carrying the translocation. As observed in protein profiles, such a translocation is not present in our BW208 genotype (Supplementary Fig. S1). In addition, R5 was designed against rye secalins (Valdés et al., 2003) and, for this reason, rye and wheat genotypes containing the translocation have higher R5 values than those without the translocation.

The decrease in γ- and ω-gliadins is compensated by the glutenin fraction

In previous work, targeted mutations of α-gliadins by CRISPR/Cas produced compensatory events with the non-target ω-gliadin fraction, and with glutenin fractions, which were increased in some mutant lines (Sánchez-León et al., 2018). In this case, targeted mutagenesis of γ- and ω-gliadins is not compensated with the non-target α-gliadins but with the glutenin fractions, which were highly increased in some mutant lines. Compensatory effects have also been observed in wheat lines with the γ-gliadins or all three gliadin fractions silenced by RNAi (Pistón et al., 2013), as well as in rice (Kawakatsu et al., 2010; Cho et al., 2016), and in maize (Huang et al., 2004). Moreover, mutation of the wheat prolamin-box binding factors (WPBFs) using the targeting-induced local lesions in genomes (tilling) method led to a decrease in gliadins and LMW-GSs and an increase in lysine-rich protein (Moehs et al., 2019). Therefore, this compensatory process seems to be common in most cereals, and it has been proposed that it could be a selective process, where proteins with similar characteristics are compensated first, and then, if necessary, with unrelated proteins (Pistón et al., 2013).

In this study we did not perform quality estimations on the γ- and ω-gliadin mutant lines. However, RNAi lines deficient only in γ-gliadins showed stronger doughs and better tolerance to overmixing (Gil-Humanes et al., 2012a). In contrast, the α-gliadin-deficient CRISPR/Cas lines showed low values in SDSS tests when α-gliadins were not compensated with ω-gliadins. However, SDSS values were greatly increased when α-gliadins were compensated with ω-gliadins and/or HMW-GSs (Sánchez-León et al., 2018). In any case, even in wheat lines with all three gliadin fractions silenced by RNAi, it was possible to produce excellent bread, maintaining not only the main organoleptic and structural characteristics but also improved nutritional properties, since the decrease in gliadins is compensated with proteins rich in lysine (Gil-Humanes et al., 2014a; Garcia-Molina et al., 2017). Doughs prepared from the low-gliadin RNAi lines show a general weakening effect, but dough stability was increased significantly in some of the RNAi lines, indicating better tolerance to overmixing (Gil-Humanes et al., 2014b). Although this functional characterization has yet to be performed on these CRISPR/Cas lines, the compensatory profile observed so far allows us to anticipate a behavior similar to that observed with the RNAi lines.

Enhanced gluten reduction in wheat plants through cross-breeding

A step towards obtaining wheat lines with all three gliadin fractions mutated was made by crossing the previously reported CRISPR lines deficient in α-gliadins (Sánchez-León et al., 2018) and lines with the ω- and γ-gliadins mutated in the present work. Several lines and their offspring were generated with the α-, γ-, and ω1,2-gliadins strongly reduced, but with the ω5-gliadins still present in their background. The comparison of the absolute values of R5 reveals that these lines showed values lower than those of their parents. However, some mutated lines in the γ- and ω-gliadins still exhibit R5 values lower than the crosses. Nevertheless, these lines have significantly lower contents of ω-gliadins, which could explain the R5 values, as discussed previously. What is interesting is that the three fractions containing CD epitopes, namely α-, γ-, and ω1,2-gliadins, were strongly eliminated by cross-breeding, paving the way to obtaining wheat lines suitable for CD patients. Additionally, the content of LMW-GSs decreased significantly in these lines. This reduction was previously observed in CRISPR/Cas lines deficient in α-gliadins (Sánchez-León et al., 2018). This may be attributed to the potential co-regulation of genes coding for gliadins and LMW-GSs by a network of transcription factors (Marín-Sanz and Barro, 2022).

Concluding remarks

While non-transgenic low α-gliadin wheat had been previously reported (Sánchez-León et al., 2018), the results presented in this study endorse the feasibility and efficacy of gene editing through CRISPR/Cas9 to simultaneously edit multiple genes in the large and complex ω- and γ-gliadin gene families in polyploid bread wheat, resulting in a collection of wheat lines with a strong reduction in the content of these immunogenic proteins, which is a step toward to the development of non-transgenic wheat lines with immune-safe gluten. Characterization of the lines identified large deletions, which predominate over insertions, and are frequent when several hits of the sgRNAs are present in the same gene. However, to establish the precise nature of the mutations, additional techniques, including the generation of genomic and proteomic data, will be necessary. Whole-genome sequencing can provide important information on the genomic rearrangements that occur in these lines, bypassing the limitations of Sanger sequencing or NGS (Yu et al., 2024). However, the cost, computing resources, and time required are only affordable for a few lines but not when applied to a large number of them. Instead, our next step is to perform stimulation assays with peripheral blood mononuclear cells (PBMCs), using transgene-free single lines deficient in the α-, γ-, and ω-gliadins, or lines from additional crossings, carrying significant alterations in the protein profile and gluten content. These assays will be highly indicative for gaining clinical information on these lines to stimulate the immunogenic response in NCWS and/or CD patients. Selecting the most promising lines, with the least PBMC-stimulatory capacities, and sequencing the complete genome of these very low-stimulatory lines, should allow us to produce fully characterized lines ready for use in the elaboration of food products for people suffering from adverse reactions to wheat.

Supplementary data

The following supplementary data are available at JXB online.

Fig. S1. The T1BL.1RS translocation is present in some wheat varieties.

Fig. S2. Schematic diagram of the expression vector used for plant transformation.

Fig. S3. A-PAGE profile of gliadins of Omega-56 offspring.

Fig. S4. A-PAGE profile of gliadins of Omega-7, Omega-15, and Omega-25 offspring.

Fig. S5. A-PAGE profile of gliadins of Omegamma lines co-transformed with pSSLOmega9 and pSSLGamma17.

Fig. S6. The number of control and mutated γ-gliadin clones.

Fig. S7. InDel characterization of Omega-25 and Omega-35.

Fig. S8. Grains of edited lines.

Table S1. List and sequence of primers for PCR analysis.

Table S2. Editing efficiency of each sgRNA.

Author contributions

FB: conceptualization and funding acquisition; SS-L, MHG-L, MG-C, and ES: methodology; FB and MM-S: formal analysis, data curation, and writing—original draft; FB, MM-S, MHG-L, and SS-L: writing—review & editing.

Conflict of interest

No conflict of interest declared.

Funding

This work was supported by the projects funded by MCIN/AEI/10.13039/501100011033 (grant nos PID2022-142139OB-I00 and TED2021-129733B-I00); European Union (‘NextGenerationEU’/PRTR); Junta de Andalucía (grant no. QUAL21_023 IAS); and ‘Conexión TRIGO’ of the Spanish National Research Council (CSIC).

Data availability

The datasets presented in this study are available in the National Center for Biotechnology Information (NCBI) database under the identifiers: PRJNA1088972 (BioProject) and PP707195-PP707589 (GenBank).