Nine commercially available gluten-digesting enzyme supplements were obtained and stored, dry at 4°C. The contents of each supplement are described in Table 1, the protein concentration of enzyme extracts is described in Supplementary Table 1.

Proteins were analyzed by Urea-SDS-PAGE and either stained with 0.06% (w/v) colloidal Coomassie Blue or electro-blotted to nitrocellulose membranes (iBlot2, Novex) developed with anti-gliadin-HRP 1/1,000× (Sigma), detected by chemiluminescence (Amersham), and calibrated against pre-stained proteins (Invitrogen) as described (46).

Gliadin was purified from a single wheat cultivar, cv Baxter, from 10 g white flour, as described (47). Freeze-dried cv Baxter gliadin was dissolved at 50 mg/ml in 8 mol/L urea, 1% (w/v) dithiothreitol (DTT), and 20 mmol/L triethylamine-HCl pH 6 (Urea-DTT-TEA). High molecular weight and low molecular weight glutenin subunits (HMWGS and LMWGS, respectively), were isolated from the pellet remaining after gliadin extraction, as described (47). These subunits were prepared to assist in the evaluation of the purity of the isolated gliadin (Figure 1).

Purified gliadin (cv Baxter) 370 mg, was dissolved in 7.4 ml of Urea-DTT-TEA, diluted ~25 × to 200 ml with MilliQ water (to reduce the urea concentration below 320 mmol/L, where trypsin is active) (46) and 25 mg of trypsin (Sigma TPCK treated bovine pancreas in 1 mmol/L HCl) added to give trypsin: gliadin ratio of 1: 20. The pH was adjusted to 7, and the gliadin was digested with trypsin for 2 h at RT during which time the initially cloudy suspension cleared. The pH was adjusted to 3.0 with 1 mol/L HCl, 50 mg of pepsin (Sigma, porcine gastric mucosa in 1 mmol/L HCl, 50% v/v glycerol) added, pH readjusted to 3.0, and the trypsin-treated gliadin digested over-night at RT. The clear digest was freeze-dried and re-dissolved in the original volume of MilliQ water to re-establish a protein concentration equivalent to 50 mg/ml (TP-gliadin). The TP-gliadin was examined by SDS-PAGE and high-performance liquid chromatography (HPLC) (Figure 1 and Supplementary Figure 3). This pre-digestion process was the reverse of that encountered physiologically where pepsin digestion at pH 3 would occur first. This was done deliberately so that residual peptides would be more soluble at the more acid pH due to residual buffer when reconstituted in MilliQ water as above. The reverse treatment, i.e., pre-digestion with pepsin and then trypsin (PT-gliadin), produced a less stable peptide solution which tended to precipitate after storage at 4°C or thawing after freezing, requiring a fresh PT-gliadin preparation each day which was impractical. The changed presentation of peptides each day would make a comparison of kinetic rates on different days impossible. TP-gliadin solution was stable, and the solution could be aliquoted, frozen, and thawed when required. This presented exactly the same concentration and composition of peptides each day, a basic requirement for relative kinetic analysis of digestions in different experiments. Although the peptide patterns of the two alternate gliadin preparations differ on HPLC-MS/MS (Colgrave pers. commun.), there was no significant difference in the rate of digestion by supplement 1, when digestion of the two alternate gliadin hydrolysates was compared with Ridascreen ELISA. This is interpreted as follows: HPLC-MS identifies the composition of individual peptides, so peptides that differ by one or two residues between the two treatments can be resolved and identified. On the other hand, ELISA is an averaging technique—so not one epitope is measured but the response is averaged over many epitopes on many peptides, minimizing the differences in rates of digestion of the two gliadin preparations measured by ELISA.

Relative kinetic constants of epitope digestion were calculated for the nine supplements. This enabled the comparison of the relative rates of gliadin digestion by enzyme supplements with the recommended enzyme dose of one supplement in a fasting stomach volume of 100 ml (48). Therefore, a working (1×) concentration of enzyme consisted of one pill in 100 ml. The final gliadin concentration in digests was the equivalent of 0.25 g/100 ml of stomach volume.

The final enzyme concentration was reduced in reactions, so that the kinetic parameters of gliadin epitope hydrolysis (initial velocity and half-life) could be accurately calculated. These kinetic parameters were then extrapolated to a final enzyme concentration of 1× as below so that the relative initial velocities could be compared. The final pH was either pH 3.5 or 7.0 to allow examination of the effect of pH on the rate of gliadin consumption. These pH values are similar to those used to simulate stomach or intestinal gluten digestion (49) however, the experiments reported here are not intended to mimic in vivo digestion conditions, rather allow calculation of relative rates of physiologically relevant epitope consumption.

Duplicate tablets of digestive supplement 1 were ground to a powder in a mortar and pestle and prepared as 11.1× concentrated solutions by pouring the powders into 9 ml of 50 mmol/L Na acetate, 50 mmol/L NaCl, adjusted to pH 2.5 (MM3.5 buffer), designed to yield a final pH 3.5 in the enzyme reaction. Duplicate extracts of supplements 2–9 (powdered in capsule form) were made by pouring the powders into 9 ml of MM3.5. Contents were mixed by vortexing 1 min, regularly inverting the tube every 10 s for 30 min at RT, and clarified by centrifugation at 2,000 g/ 5 min. It was essential to use MM3.5 buffer to extract active enzymes from AnPEP-containing pills. No activity was observed if AnPEP supplements were extracted at pH 7 (Supplementary Figure 1), at either RT or 4°C. The addition of 500 mmol/L NaCl did not improve the extraction of AnPEP activity. The reason for this is unknown but may be a result of an interaction between the AnPEP enzyme and the inert binder causing inadvertent pelleting of the enzyme.

Enzyme solutions were prepared as above and diluted as required to 0.11 ×, 0.011 ×, or 0.0011 × enzyme concentrations as indicated, with an appropriate volume of phosphate-buffered saline (PBS) buffer containing 0.1% (w/v) Tween 20 (PBST). Enzyme reactions were prepared in 200 μl polypropylene ELISA wells by adding 20 μl of enzyme supplement appropriately diluted with PBST as indicated, to a final volume of 200 μl, further diluting the supplement concentration by 0.1 ×. The reaction contained 168 μl of either MM3.5 or MM7 buffer (50 mmol/L Na-phosphate, 50 mmol/L NaCl adjusted to pH 7.0 and designed to yield a final pH 7.0 in the enzyme reaction) and 12 μl of TP-gliadin (41.8 mg/mL in Urea-DTT-TEA). Enzyme alone (EA) control solutions contained 20 μl of appropriately diluted supplement as above, diluted to a final volume of 200 μl by addition of 168 μl of either MM3.5 or MM7, and 12 μl of Urea-DTT-TEA in place of cv Baxter gliadin. A zero-time control consisted of 168 μl of either MM3.5 or MM7, 12 μl TP-gliadin (41.8 mg/ml in Urea-DTT-TEA), and 20 μl of PBST in place of an enzyme. Reactions were equilibrated at 30°C with gentle agitation. The proteolysis was started by the addition of enzyme unless noted, and 20 μl aliquots were taken at the indicated time from 10 to 60 min, diluted by either 1/1,000× (Gluten-Tec Elisa) or 1/4,000 (Ridascreen Elisa) with PBST. PBST was used in place of the Gluten-Tec sample diluent, which produced a gel upon heating which prevented accurate pipetting. Diluted reaction aliquots (100 μl) were transferred to PCR tubes, heated at 95°C for 20 min to eliminate any residual proteolytic activity. The solutions were cooled, centrifuged briefly to collect droplets from the lid, and aliquots (40 μl) were taken for ELISA as below.

Digestions for western blot analysis were as above but contained either 0.5 mg of Baxter gliadin or TP-gliadin but at 1.1× enzyme concentration (i.e., 10× higher than for kinetic experiments) at both pH 3.5 and pH 7.0.

For the Gluten-Tec competitive ELISA (EuroProxima B.V., Arnhem, The Netherlands), dilute, heat-treated reaction aliquots (40 μl) were further diluted with 10 μl PBST and added to ELISA wells with 50 μl 1/100× dilute antibody-horseradish peroxidase conjugate and left at 4°C for 3 h to equilibrate with gentle agitation. ELISA wells were rinsed 5× with manufacturer wash buffer and tapped dry. Manufacturer substrate (100 μl) was added and left to react at 30°C for 15 min and stopped with 100 μl stop reagent. The A450 of wells was read on a Perkin Elmer plate reader and calibrated against 0–210 ng dilutions of cv Baxter gliadin (Supplementary Figure 2C).

The Ridascreen competitive ELISA was performed as per the manufacturer's instructions (R-Biopharm AG, Darmstadt, Germany). Dilute, heat-treated reaction aliquots (40 μl) were diluted with 10 μl PBST and added to Ridascreen Competitive ELISA wells with 50 μl of 1/10× dilute antibody-horseradish peroxidase conjugate and left at 30°C for 30 min to equilibrate with gentle agitation. ELISA wells were rinsed 5× with manufacturer wash buffer and tapped dry. Manufacturers substrate (100 μl) was added and left to react at 30°C for 10 min and stopped with 100 μl of stop reagent. The A450 was read with a Perkin Elmer plate-reader and calibrated against 0–30 ng dilutions of cv Baxter gliadin (Supplementary Figure 2A).

Dilute and heat-treated solutions (10 min at 95°C) still damaged the ELISA antibodies, due to residual proteolytic activity from the supplements. This was demonstrated by decreased signal levels (Supplementary Figures 2B,D). No damage was evident when diluted reactions were heated for 20 min at 95°C.

Duplicate absorbances from the plate reader were transformed into amounts of Baxter gliadin using the appropriate standard curve and biological duplicates plotted vs. time using GraphPad Prism (version 8.4.2 for Windows, GraphPad Software, San Diego, CA, USA, www.graphpad.com). The decreasing gliadin amounts remaining over time were fitted to first-order non-linear decay models and initial rate (±SE) and half-life (±SE) calculated. In general, the data well-fitted the curves with R² values >0.9. However, where rates of digestion were low and the data were noisy, fitting data to non-linear regressions did not return reasonable estimates of initial rate or half-life, and the data were therefore fitted to linear regressions which returned more reasonable estimates of initial rate and half-life. With linear regressions, only the SE of the initial rate could be calculated by the software.

More rapid gliadin hydrolysis was indicated by a higher initial rate and shorter half-life. The initial rate and half-life of gliadin digestion were normalized to enzyme concentrations of 1× by dividing the observed initial rate and SE by the dilution and multiplying the observed half-life and SE by the dilution. The significance of differences in normalized mean initial rate and half-life were examined by the one-way ANOVA with Tukey's multiple comparison test at the indicated significance level.

Protein concentrations of the nine enzyme supplements were determined by the method of Bradford, calibrated against gamma-globulin (50). The protein concentration varied from 0.23 mg/ml (Supplement 4) to about 1 mg/ml (Supplements 1, 6, 7, 8, and 9, Supplementary Table 1); however, the dose of digestive supplements was related to a dose of one pill dissolved in 100 ml.

Gliadin was isolated from a single wheat cultivar cv Baxter. Analysis by Urea-SDS-PAGE showed that the gliadin was uncontaminated with HMWTGS and LMWTGS (Figure 1, BG). Urea-SDS-PAGE of trypsin-treated gliadin showed in addition to a trypsin band at 25 kDa (Figure 1, lane BT, *trypsin) a number of small peptides that ran at <17 kDa (Figure 1, lane BT). Urea-SDS-PAGE of TP-gliadin produced in addition to trypsin and pepsin bands (Figure 1, lane BTP, *trypsin and *pepsin, at 25, and 39 kDa, respectively), smaller peptides that ran ahead of the electrophoretic front <10 kDa. HPLC showed that the TP-gliadin consisted of over 40 small peptides (Supplementary Figure 3) and confirmed digestion.

Digestion of both native gliadin and digested TP-gliadin was firstly examined by western blotting of Urea-SDS-PAGE gels of all extracts, but at 1.1× enzyme concentration (i.e., 10× higher than for kinetic experiments) at both pH 3.5 and pH 7 (Figure 2). Lanes were loaded with equal protein loads equivalent to 2 μg of protein. Enzyme activity was demonstrated by the absence of western bands for native or TP-gliadin. Native gliadin runs as a series of dominant bands at 30–40 kDa. The TP-gliadin runs largely at the electrophoretic front with a small amount of partially degraded peptides in the range 20–30 kDa (e.g., Figure 2D, ellipse). Supplement 1 was active under all conditions, removing both native and TP-gliadin fragments. At this high enzyme concentration, the DPP-IV supplements 5, 6, 7, and 8 were also active with native gliadin at both pH (Figures 2A,B). Supplements 2, 3, and 4 had no activity with native gliadin (Figures 2A,B) as expected for AnPEP enzymes. Supplement 9 was active with native gliadin but only at pH 3.5 (Figure 2A) and appeared active with TP-gliadin but only at pH 7. The AnPEP supplements 2, 3, and 4 were active with TP-gliadin at pH 3.5 (Figure 2C) but had no activity at pH 7 (Figure 2D) as expected. The DPP-IV supplements 5, 6, 7, and 8 were only active with TP-gliadin at pH 7 (Figure 2D) and had little activity at pH 3.5 (Figure 2C).

Residual proteolytic activity in diluted samples had to be inactivated before ELISA assay to avoid damage to the ELISA antibodies. Proteolytic activity was remarkably stable and resisted most methods commonly used to destroy proteolytic activity. Adjustment of supplements 1, 2, 6, 7, and 8 to pH 12.3 (final 1 mol/L NaOH) for 20 min did not inhibit subsequent gliadin proteolysis. Dilution to final concentrations of 2% (w/v) SDS, 1% (w/v) DTT, and heating at pH 7 at 95°C/5 min did not inhibit the subsequent hydrolysis of gliadin by supplement 1 but did inhibit the activity of supplement 2. Incubation of supplement 1 in final concentrations of 4 mol/L urea, 1% (w/v) SDS for 30 min at RT did not inhibit hydrolysis of gliadin. Incubation in 80% (v/v) acetone at 70°C for 30 min did not inhibit gliadin proteolysis in supplements 1, 5, 6, 7, or 8. Supplements 1, 2, 5, and 6 heated at 95°C/10 min still had slight activity, whereas all other supplements were inhibited by this treatment. It was necessary to heat the final dilutions at 95°C/20 min to remove all remaining proteolytic activity from all supplements (Supplementary Figure 2). Supplement 1 was stable to the action of both trypsin and pepsin (Supplementary Figure 4).

Residual epitopes were measured in duplicate by Gluten-Tec competitive ELISA and from the lines of best fit (Figure 3), the initial rate and half-life of gliadin digestion were determined and corrected to 1× enzyme dilution at pH 3.5 (Supplementary Table 1) and pH 7.0 (Supplementary Table 2). At pH 3.5, supplements 1, 2, 3, and 4 showed the highest rates of gliadin digestion (Figure 3A) while at pH 7.0 supplements 1, 2, and 4 showed the highest rates of gliadin digestion (Figure 3C). Preparation 5 (Figure 3A) showed a modest rate of digestion and supplements 6, 7, 8, and 9 showed little activity at pH 3.5 (Figure 3B). At pH 7, supplements 3 and 5 showed modest rates of digestion (Figure 3C) but supplements 6, 7, 8, and 9 had low activity (Figure 3D).

Residual epitopes were measured in duplicate by Ridascreen competitive ELISA and from the lines of best fit (Figure 4), the initial rate and half-life of gliadin digestion were determined and corrected to 1× enzyme dilution at pH 3.5 (Supplementary Table 3) and pH 7.0 (Supplementary Table 4).

At pH 3.5, supplements 1, 2, and 3 showed the highest rates of gliadin digestion (Figure 4A). Supplements 4, 5, 6, 7, 8, and 9 showed little activity at pH 3.5 (Figures 4A,B). At pH 7.0, supplements 1, 5, 6, 7, 8, and 9 showed the highest rates of gliadin digestion (Figures 4C,D). At pH 7.0, supplements 2, 3, and 4 showed low rates of digestion (Figure 4C) with Ridascreen ELISA.

The significance of the difference in the initial rates and half-lives of gliadin digestion were determined by ANOVA at pH 3.5 and pH 7.0 for Gluten-Tec ELISA (Figure 5) and Ridascreen ELISA (Figure 6).

Aspergillus niger prolyl endopeptidase containing enzymes 2 and 3 had the highest initial rate and shortest half-lives, followed by the caricain supplement 1, at about half the rate of supplement 2 and 3 but only when measured by Gluten-Tec ELISA at pH 3.5 (Figures 5A,B). The scatter of the initial rates of supplements 2 and 3 reduced the significance of the differences in mean values. The caricain containing supplement 1 had a significantly higher initial rate and shorter half-life (Figures 5C, 6C) than all other supplements when measured with Gluten-Tec at pH 7.0 (Figures 5C,D). The DPP-IV supplements 5, 6, 7, and 8 and the fungal/bacterial containing supplement 9 had little activity at either pH with Gluten-Tec ELISA.

The initial rates of the caricain supplement 1 were statistically larger, and the half-life was smaller than all other supplements when measured by Ridascreen ELISA at both pHs (Figure 6) excepting those of the DPP-IV supplements 5, 6, 7, and 8 which were not significantly different to caricain at pH 7.0 (Figures 6C,D). The fungal/ bacterial containing supplement 9 had little activity at either pH with Ridascreen ELISA.