1. Introduction
Consumption of gluten-containing products is involved in several immune disorders, including celiac disease, wheat allergy and wheat sensitivity. Celiac disease is an autoimmune disorder triggered by ingestion of gluten proteins from Triticeae, and causes small intestine mucosa damage [
1]. Gluten proteins are referred to as seed storage prolamin proteins; in wheat they are called gliadins and glutenins, in barley they are called hordeins, and in rye they are called secalins [
2]. To maintain a healthy condition, celiac patients must practice strict avoidance and keep a gluten-free diet. Therefore, residual gluten or cross contamination in products needs to be quantified correctly. The Codex Alimentarius standard 118-1979 (2008) [
3] defined the current threshold for gluten-free products as 20 mg/kg, and the European Union [
4], the United States of America [
5] and Canada [
6] also adopted the same threshold. The most commonly used quantification method is the enzyme-linked immunosorbent assay (ELISA) method and several commercial kits based on different gluten antibodies are available on the market, including R5, G12, A1, and Skeritt antibodies. The R5 Mendez based ELISA method is the type I method suggested by the Codex Alimentarius standard. The R5 antibody is a monoclonal antibody raised against rye secalins and mainly recognises the epitope QQPFP [
7]. Ideally, in a sandwich type R5 ELISA designed for intact protein detection, the R5 antibody detects gliadin fractions from a wheat sample extracted from a food matrix. The measured gliadin content is calibrated against a gliadin standard, and finally total gluten content of the sample is calculated from gliadin concentration with a conversion factor 2 based on the theoretic ratio (1:1) of gliadins:glutenins. Because of the cross reaction of the R5 antibody with other non-wheat Triticeae prolamins, the assay is also used for detection of barley and rye prolamins, because usually the contamination source is unknown from a sample. However, the quantification of barley prolamin using a wheat gliadin standard caused about five times overestimation in R5 sandwich ELISA [
8,
9,
10]. A separate barley standard was needed for calibration, thus in our previous study [
10], we isolated barley C-hordein and the calibration with C-hordein achieved more accurate quantitation than the gliadin standard when determining the barley contamination in gluten-free oats. The primary structure of C-hordein, much like the homologues in wheat and rye (ω-gliadins and ω-secalins, respectively) consists almost entirely of repeats of the QQPFP motif, which is the main recognition sequence for the R5 antibody [
2]. C-hordein from three selected barley cultivars with different RP-HPLC patterns had equal and good R5 recognition. To calibrate the total hordein amount, isolated C-hordein was mixed with an inert protein in proportion (40%) to represent the reactivity of total hordein against the R5 antibody [
10]. Thus, the approach of a reference material using one group of gluten protein was first proposed.
There are several challenges in gluten detection by ELISA methods that have been critically discussed, including the extraction methods, the antibody specificity and detection, and the calibration step with the reference material [
11,
12]. Using a generic extraction method and a common calibrator, a study revealed that the comparability of all gluten detection commercial kits was very limited [
13]. The lack of a certified reference material has been one major challenge. Several gluten detection kits use gluten or gliadin isolate as the reference material for their calibrations. Of these, the Prolamin working group (PWG) gliadin standard, made with a mixture of 28 cultivars from the UK, Germany and France from 1999, was the best characterised. The PWG gliadin standard consisted of a solution with a total protein content of 96.7%, of which the gliadin content was 86.4%. Of this, the α/β-gliadins comprised 41.7%, γ-gliadins 47.0%, ω1,2-gliadin 6.3% and ω5-gliadin 5.0% [
14]. However, this standard was not accepted as certified reference material as it did not have sufficient purity and it was not reproducible [
11]. Additionally, none of the wheat cultivars used are currently important on the market and the stock of this batch of standard material will run out soon. This brings up the issue of the development of new reference materials for wheat gluten quantification and for other cereals. The strategies proposed are the use of whole wheat flour, gliadin or gluten isolate from a mixture of cultivars or from one single cultivar, incurred matrix, or a single protein [
11,
15,
16]. Based on these strategies and led by the PWG, five cultivars were selected after characterisation as the basis for the development of a new reference material [
15,
16]. The gluten composition of these five cultivar flours varied between harvest years, but a blend of the five cultivars overcame this variability and showed advantages over use of a single cultivar in ELISA responses [
17,
18]. A gluten isolate or a gliadin isolate from the blend of the five cultivars showed the same protein composition as the native flour [
18]. The varying gluten composition by genotype and environmental factors certainly increased the difficulty of satisfying the criteria of reference material. A series of reference materials that are suitable for different cereals and food matrices is required for reliable gluten quantification.
The aim of this study was to investigate the feasibility of a single protein group, barley C-hordein, for calibration of wheat gluten in R5 ELISA assay. We collected 27 common wheat cultivars that are important in the recent market and investigated their gluten compositions and their total gluten reactivity against the R5 antibody. Based on their R5 reactivities, we proposed the use of barley 10% C-hordein for the calibration of wheat gluten. To evaluate the calibration in raw and heat-treated foods, three wheat cultivars, with varying protein composition and R5 reactivities, were selected and spiked in gluten-free oat flour and oat biscuits made from the spiked flour, respectively.
2. Materials and Methods
2.1. Materials
Based on production in 2016–2017, samples of 27 high-yielding common wheat cultivars (Triticum aestivum L.) were collected from seven countries, including Anniina, Quarna and Amaretto from Finland; Julius, Brons, and Hereford from Sweden; Julius, Kerubino, and Patras from Germany; Cellule and Apache from France; Siskin, Lili, Crusoe, Zulu, Claire, Revelation, and Britannia from the UK; Brandon, Steller, Foremost, and Penhold from Canada; Gregory, Lancer, Spitfire, Suntop, and Mace from Australia. Hull-less barley (Hordeum vulgare) cultivar Jorma was obtained from Villala, Finland. All chemicals were analytical grade.
2.2. Gluten Composition Analysis by Reverse-Phase-High Performance Liquid Chromatography (RP-HPLC)
The grain seeds were milled with a sample mill (Koneteollisuus Oy, KT-30, Klaukkala, Finland). The protein extraction method was slightly modified following modified Osborne sequential extraction [
19], 100 mg flour was extracted by 1 mL 0.4 M NaCl + 0.067 M HKNaPO
4 (pH 7.6) at room temperature (20–23 °C) for 10 min for two times. These two extractions were combined to form the Albumin + Globulin (Alb + Glo) fraction. The gliadin fraction was then extracted with 0.5 mL 50% (
v/v) propan-1-ol at 60 °C for 10 min three times and combined [
20]. The glutenin fraction was extracted in 1 mL of 50% (
v/v) propan-1-ol with 5% (
v/v) β-mercaptoethanol at 60 °C twice and the extracts were combined. Duplicate samples of these three protein fractions were obtained from each cultivar. After filtration through a 0.45 µm GHP membrane (Pall Corporation, Ann Arbor, MI, USA), these fractions were analysed by RP-HPLC using the Agilent Technologies 1200 series system with a diode array detector (Agilent, Santa Clara, CA, USA). Protein solutions were separated at 50 °C on a SUPELCO Discovery Bio Wide Pore C8, 5 µm, 25 cm × 4.6 mm (Sigma-Aldrich, St.Louis, MO, USA) with matching guard column 2 cm × 4 mm on a gradient of 2 min, 0% B; 4 min, 24% B; 52 min, 56%; 58 min, 90% B; 65 min, 0% B, where buffer B consisted of acetonitrile with 0.1% (
v/v) trifluoroacetic acid and buffer A consisted of 0.1% (
v/v) trifluoroacetic acid in mQ water. The injection volume for Alb + Glo fraction was 50 µL, for gliadin fraction it was 25 µL and for glutenin fraction it was 50 µL. UV detection was set to 210 nm. Protein content was calculated based on the peak area using bovine serum albumin (BSA) as the standard in the linear range (0–80 µg). Cultivar-specific conversion factors were calculated as its total gluten proportion divided by its total gliadin proportion.
2.3. Isolation of Total Gluten and Their R5 Reactivity
To evaluate the total gluten R5 reactivity, gluten isolates of the 27 wheat cultivars were prepared by modified Osborne fractionation. Albumins and globulins were removed by extraction three times with the buffer described in
Section 2.2, total gluten of each cultivar was extracted from 5 g flour by 30 mL 50% (
v/v) propan-1-ol with 5% (
v/v) β-mercaptoethanol at 60 °C for 30 min twice. After centrifugation (18,000×
g) for 10 min, the supernatant was collected and dialysed (SnakeSkin Dialysis Tubing, 3.5K MWCO, ThermoFisher, Rockford, IL, USA) against mQ-H
2O with at least three changes. The supernatant was then lyophilized and the nitrogen content was determined by the Dumas combustion method (VarioMax CN, Elementar Analysensysteme GmbH, Langenselbold, Germany) and multiplied by 5.7 to give the protein content [
21]. The 0.25 g gluten isolate was dissolved in 2.5 mL patented cocktail solution (R7006) and gluten content was quantified using the Ridascreen Gliadin R7001 following kit instructions (R-Biopharm, Darmstadt, Germany). In order to calculate the EC50 value of the total gluten R5 reactivity, a series of dilutions of each gluten solution was prepared using at least six measuring points on the curve. The EC50 value indicates the half concentration of the maximal antibody binding and was calculated using a non-linear four parameter curve fit by Graphpad Prism 8 (San Diego, CA). A C-hordein isolate preparation has been described [
10], briefly hordeins were extracted by aqueous alcohol solution and separated with an ion exchange chromatographic method, the C-hordein fraction was collected, dialyzed and lyophilized. A stock solution of C-hordein and BSA in 60% (
v/v) ethanol were made at the same concentration, and then mixed at 10% (1 C-hordein: 9 BSA), 20% (2 C-hordein: 8 BSA) and 30% (3 C-hordein: 7 BSA) (
v/v). The 10%, 20% and 30% C-hordein standard solutions were further diluted to fit into the ELISA reaction curve.
2.4. Isolation of Gluten Subunits and R5 Sandwich Responses
Based on retention time, fractions comprising ω5-gliadin, ω1,2-gliadin, α-gliadin, and γ-gliadin in the gliadin fraction, HMW-glutenin and low molecular weight (LMW)-glutenin in the glutenin fraction of cultivar Crusoe were collected from RP-HPLC separation [
19]. The protein content was determined as before. The fractions were dried under nitrogen flow first, and then with vacuum by SpeedVac (Savant SC110A Concentrator, San Diego, CA, USA). The fractions were dissolved in 0.25 mL cocktail solution and then in 80% (
v/v) ethanol, the gluten subunits were diluted into a suitable range for analysis in R5 sandwich ELISA (Ridascreen Gliadin R7001, R-Biopharm, Darmstadt, Germany). The fractions of ω1,2-gliadins of 5 cultivars (Amaretto, Anniina, Brandon, Claire and Lili), and the fractions of α-gliadins and γ-gliadins of 4 cultivars (Amaretto, Apache, Brandon and Foremost) with distinct HPLC patterns were also collected and analysed with sandwich R5 ELISA as before. A series of dilutions was prepared and the EC50 value was calculated as in 2.3.
2.5. Spiking Oat Flour and Oat Biscuits and Calibration with C-Hordein Standard
Wheat cultivars of high, medium and low total gluten R5 reactivity, Steller, Zulu and Apache, respectively, were selected for spiking tests to demonstrate the performance of the calibration when spiked with different cultivars. The three wheat flours were spiked into gluten-free oat flour (Provena, Raisio, Finland) at 1 g/kg (1000 mg/kg) concentration in three consecutive steps (10 × 10 × 10) for better homogeneity. Depending on the flour total protein content and gluten composition, the spiked gluten concentration was around 100 ppm. In order to investigate the effect of food processing/heat treatment on the ELISA analysis, oat biscuits were prepared by mixing eggs 20 g, sugar 50 g, and butter 50 g with 100 g of the spiked oat flour. The biscuits were baked in a conventional oven at 180 °C for 16 min. After cooling to room temperature, the loss of moisture was measured. The gluten content of the spiked oat flour and biscuits was measured following Ridascreen Gliadin R7001 ELISA. Negative control biscuits were made from the gluten-free oat flour to ensure no contamination occurred during the biscuit-making process. The gluten content was calculated by (1) calibration with the gliadin standard, then multiplied with the conversion factor 2; (2) calibration with gliadin standard and then multiplied with the cultivar-specific conversion factor obtained; and (3) calibration with 10% C-hordein standard and no conversion. The gluten protein recovery was determined by , where theoretical gluten content = nitrogen content × 5.7 × gluten proportion from HPLC. The spiked flour and biscuit samples were prepared in two biological replicates; for measurement of each biological replicate, two extraction replicates were made and measured in four technical replicates in ELISA and calculated with two dilution factors.
2.6. Statistical Analysis
The significance test of difference of gluten protein compositions results from RP-HPLC was conducted by SPSS 10.0, using one-way ANOVA analysis with Tukey’s HSD test; within each protein group the significance (p < 0.05) was indicated by different letters. The significance test of difference of protein recovery from the spiking test was determined by one-way ANOVA analysis and Tukey’s test by Graphpad Prism 8; the level of significance was indicated by asterisk (ns, not significant; * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001)
4. Discussion
This study investigated the prolamin compositions of the wheat cultivars on the current market and their total gluten reactivities towards the R5 antibody. The results showed that the gluten composition varied greatly among popular common wheat cultivars; gliadins were the main component and the ratio of gliadins to glutenins ranged from 2.07 to 5.34. Gliadins were also the main recognition for the R5 antibody, of which ω1.2-gliadins had the strongest reactivity. In addition, the same type of gliadins from different cultivars also varied in R5 reactivity. The complexity of gluten composition and the varying R5 reactivity of each gluten type explained the large range of total gluten R5 reactivity of these 27 cultivars. Thus, it would be challenging to establish and maintain a reference material based on a single cultivar or a mixture of wheat cultivars. The calibration with 10% C-hordein standard achieved the same protein recovery as the gliadin standard with cultivar-specific conversion factor in all three wheat cultivar spiked oat flour or biscuits.
The main recognition of R5 antibody is epitope QQPFP, the amount of which in prolamin sequences determines the varying reactivity. For example, eighteen repeats of QQPFP were found in ω1.2-gliadins (Uniprot entry D2KKB1), 3 repeats from γ-gliadins (P06659), 5 repeats from γ-gliadins (P21292), and 1 repeat from α-gliadins (P04721, P04723, P04725, P18573), explaining the main R5 reactivity of gluten from these prolamins [
22]. Because of the complexity of wheat prolamins, this study showed the same prolamin group from different cultivars varied to some extent. No R5 epitope was found in ω5-gliadins (Q40215) and only one was found in LMW-glutenins (P13615); this accounted for their minimum reactivities. Although no R5 epitope was found in HMW-glutenin x-type (P10388), and y-type (P10387), they had some limited reactivity against the antibody. A similar phenomenon was observed in Western blot with the R5 antibody [
20,
23]. In addition, the R5 antibody also recognises homologous epitopes such as LQPFP, QLPYP, QQSFP, QQPYP and PQPFP [
24]. The mild recognition of HMW-glutenins complicated total gluten quantitation when, theoretically, the R5 antibody detects only gliadins and therefore renders the conversion factor incorrect. The varying total gluten R5 reactivity can be partially explained by the proportion of ω1.2-gliadins. For example, cv. Steller and cv. Brandon contained 7.5% and 9.0% of ω1.2-gliadins in wheat proteins, respectively, and also had strong total gluten R5 reactivity with EC50 values of 30.7 ng/mL and 35 ng/mL, respectively. On the other hand, cv. Brons and cv. Apache had only 1.9% and 2.0% ω1.2-gliadins, respectively, and consequently had weak total gluten R5 reactivity with EC50 values of 325.6 ng/mL and 376.6 ng/mL, respectively. However, a good correlation of the EC50 and ω1.2-gliadins cannot be established (
Figure S4). Wheat cultivar Claire had a low ω1.2-gliadins proportion of 2.8% but a strong total gluten R5 reactivity with an EC50 value of 38.6 ng/mL. This may be explained by a high ratio of gliadins to glutenin (4.24), such that α- and γ-gliadins contributed mostly to the total gluten R5 reactivity.
In this study, the ratio of gliadins to glutenins of 27 wheat cultivars ranged from 2.07 to 5.34 based on Osborne sequential extraction. In previous studies, with similar RP-HPLC methods, the ratio ranged from 1.51 to 3.14 (54 cultivars) [
25], 1.7 to 4.2 (13 cultivars) [
26], 2.0 to 4.2 (23 cultivars) [
16], and 1.93 to 3.10 (5 cultivars) [
15]. These data suggest that the conversion factor 2, recommended by the Codex standard 118-1979, is higher than the actual conversion factor for common wheat (1.19–1.48 this study, 1.24 to 1.50 [
16], 1.32–1.66 [
25]). One must be aware that the Osborne sequential extraction based on solubility cannot provide clear-cut classification of wheat proteins, because of the fact that some of the glutenins or albumin/globulins co-extracted into gliadin fraction or even co-eluted in the RP-HPLC. For example, amylase/trypsin inhibitors were found to be co-extracted in the gliadin fraction and co-eluted in the ω-gliadins fraction in RP-HPLC [
27,
28,
29].
Due to varying compositions of wheat cultivars and varying R5 reactivities of gluten types, the total gluten R5 reactivities of these cultivars resulted in a large range. C-hordeins, similar to ω1.2-gliadins, have good binding to the R5 antibody and were selected as the base of the reference material, for their consistently higher proportion of barley and their relative ease of preparative isolation. Ideally, a reference material would comprise a total gluten isolate, but its solubility and stability is poor in a solution due to the aggregative nature of gluten proteins. Thus, a gliadin standard was developed because gliadins are monomeric and soluble in aqueous alcohol solution. The R5 antibody detected epitopes related to QQPFP and C-hordein is a good source of QQPFP epitopes. Interestingly, although three spiked wheat flours had varying gluten composition and R5 reactivity, in both raw and cooked foods, calibration with 10% C-hordein achieved comparable results as the gliadin standard with the cultivar-specific conversion factors. The reason might be that calibration with 10% C-hordein represents the average of total gluten. The calibration with the gliadin standard represents only one part of gluten and therefore to achieve correct quantitation of total gluten, a cultivar-specific conversion factor is needed. However, this is not normally possible as the contaminant source/cultivar is unknown.
Using a conversion factor of 2 achieved higher protein recovery compared to the theoretical gluten content. This raises a question concerning the efficiency of the two-step extraction procedure, firstly with the “cocktail solution” containing guanidine hydrochloride and 2-mercaptoethanol, and secondly with an aqueous alcohol solution. This procedure improved the gliadin extractability for unheated and heated food compared to the conventional extraction method with 60% aqueous alcohol [
30]. Another extraction buffer with reducing agent tris(2-carboxyethyl)phosphine (TCEP) achieved similar recovery [
31,
32]. In those tests the spike was isolated gliadin, of which the extractability is higher than that of whole flour [
33]. In a spiked flour test, the gliadin recovery determined by the R5 and G12 ELISA assay against RP-HPLC results of gluten-free flour ranged from 72.8% to 162.5% and its cookies ranged from 49.1% to 192.0% [
15]. In a comparison study of ELISA kits, the 2-step extraction procedure of vital gluten as in R-Biopharm, Agraquant or Transia showed, on average, a lower extraction efficiency than a Japanese official allergen detection extraction method [
13]. A higher conversion factor of 2, in a way, compensated for the incomprehensive nature of the two-step extraction procedure. A better extraction procedure is needed to improve the extractability of gluten proteins, such as the introduction of a prolonged extraction time or a multi-step extraction.