Multi-allergen Quantitation and the Impact of ... - ACS Publications

Article pubs.acs.org/JAFC

Multi-allergen Quantitation and the Impact of Thermal Treatment in Industry-Processed Baked Goods by ELISA and Liquid Chromatography-Tandem Mass Spectrometry Christine H. Parker,† Sefat E. Khuda,‡ Marion Pereira,‡ Mark M. Ross,†,∥ Tong-Jen Fu,# Xuebin Fan,⊥ Yan Wu,⊥ Kristina M. Williams,‡ Jonathan DeVries,§ Brian Pulvermacher,§ Binaifer Bedford,# Xi Zhang,⊥ and Lauren S. Jackson*,# †

Center for Food Safety and Applied Nutrition, U.S. Food and Drug Administration, 5100 Paint Branch Parkway, College Park, Maryland 20740, United States ‡ Center for Food Safety and Applied Nutrition, U.S. Food and Drug Administration, 8301 Muirkirk Road, Laurel, Maryland 20708, United States # Center for Food Safety and Applied Nutrition, U.S. Food and Drug Administration, 6502 South Archer Road, Bedford Park, Illinois 60501, United States § James Ford Bell Technical Center, General Mills, 9000 Plymouth Avenue North, Golden Valley, Minnesota 55427, United States ⊥ Institute for Food Safety and Health, Illinois Institute of Technology, Bedford Park, Illinois 60501, United States S Supporting Information *

ABSTRACT: Undeclared food allergens account for 30−40% of food recalls in the United States. Compliance with ingredient labeling regulations and the implementation of effective manufacturing allergen control plans require the use of reliable methods for allergen detection and quantitation in complex food products. The objectives of this work were to (1) produce industryprocessed model foods incurred with egg, milk, and peanut allergens, (2) compare analytical method performance for allergen quantitation in thermally processed bakery products, and (3) determine the effects of thermal treatment on allergen detection. Control and allergen-incurred cereal bars and muffins were formulated in a pilot-scale industry processing facility. Quantitation of egg, milk, and peanut in incurred baked goods was compared at various processing stages using commercial enzyme-linked immunosorbent assay (ELISA) kits and a novel multi-allergen liquid chromatography (LC)-tandem mass spectrometry (MS/ MS) multiple-reaction monitoring (MRM) method. Thermal processing was determined to negatively affect the recovery and quantitation of egg, milk, and peanut to different extents depending on the allergen, matrix, and analytical test method. The Morinaga ELISA and LC-MS/MS quantitative methods reported the highest recovery across all monitored allergens, whereas the ELISA Systems, Neogen BioKits, Neogen Veratox, and R-Biopharm ELISA Kits underperformed in the determination of allergen content of industry-processed bakery products. KEYWORDS: food allergens, baked goods, thermal processing,, ELISA, mass spectrometry

■

INTRODUCTION

composition and the manner in which the food has been processed can mask or alter allergen markers, thereby impairing or negating detection and quantitation of food allergens. Thermal processing, in particular, has been shown to alter the physiochemical properties of allergen proteins depending on the mode of thermal processing (dry versus wet) and environmental factors including temperature, heat duration, pH, and accessory matrix components.2,3 Uncertainties in the effects of processing on the physical and chemical properties of a protein, therefore, pose a significant challenge in the development of reliable methods for allergen detection and quantitation. Protein-based analytical methods traditionally involve immunochemical detection protocols. Commercial immuno-

Food allergy is a growing public health concern affecting up to 5% of the adult population and 8% of young children.1 The Food Allergen Labeling and Consumer Protection Act (FALCPA) of 2004 mandated the declaration of major food allergens (milk, egg, fish, crustacean shellfish, tree nuts, peanut, wheat, and soy) on labels for packaged foods sold in the United States. Undeclared allergens, however, can inadvertently appear in a product from ineffective equipment sanitation procedures, cross-contact during manufacturing, and incorrect labeling. To effectively safeguard the food-allergic population, the food industry and regulatory bodies require reliable analytical methods for allergen detection in complex food products. The methods currently used for the detection of allergens in foods target allergen markers (i.e., proteins, peptides, DNA) to indicate the presence of allergenic ingredients. Despite the abundance of available analytical tools, the selection of appropriate detection methods for allergens can be challenging, in part due to the inherent complexity of food. Food © XXXX American Chemical Society

Received: September 3, 2015 Revised: November 11, 2015 Accepted: November 23, 2015

A

DOI: 10.1021/acs.jafc.5b04287 J. Agric. Food Chem. XXXX, XXX, XXX−XXX

Article

Journal of Agricultural and Food Chemistry

Figure 1. Industrial process flow diagram for production of (A) cereal bar and (B) muffin incurred bakery products.

chemical-based technologies (e.g., ELISA test kits, lateral flow devices, and dipstick tests) are the most commonly used platform for routine food analysis due to their relative ease of use, high precision, sensitivity, and potential for standardization. Quantitative ELISA kits are designed to extract target allergen protein(s) and selectively detect target antigens through immunochemical recognition. Concentration, typically expressed as parts per million (ppm; μg/g) of allergen or protein, is calculated from calibration curves generated with standard reference materials. Despite the abundance of commercially available immunochemical food allergen detection methods, quantitation results from the same sample may vary between kits from different manufacturers due to differences in protein extraction, reference materials, antibody selectivity, and the format of analytical results. For example, antibodies in immunochemical-based methods can be raised to individual allergens such as the allergenic peanut protein Ara h 1 in the Neogen BioKits or total peanut allergen protein as found in the R-Biopharm, Neogen Veratox, and Morinaga peanut ELISA kits.4,5 The evaluation of ELISA kits to detect allergenic foods furthermore requires an assessment of the assay sensitivity and specificity in the food matrix of interest. Numerous published studies have examined the performance of commercial ELISA methods for analyzing thermally processed foods, reporting underestimations of allergen concentrations for egg, milk, and peanut due in part to reduced solubility and impaired immunochemical recognition of chemically or physically altered proteins.6−12 Of the food-allergic pediatric population, 30.4% of children report allergies to multiple foods.13 Due to the prevalence of food-allergic individuals and frequent use of multiple allergen ingredients to achieve the desired flavor and texture in foods, there is an increasing need to develop methods for the simultaneous measurement of multiple allergens. The majority of commercially available allergen detection methods are singleallergen assays, resulting in increased labor costs and reduced economic efficiency when repetitive evaluations of multipleanalyte mixtures are needed. Multi-allergen detection methods

have recently been developed using a multiplexed enzyme immunoassay,14−17 mass spectrometry,18−22 and optical thinfilm biochips with PCR detection.23 Mass spectrometry (MS) methods have the advantage of providing highly multiplexed allergen detection with molecular level specificity in a single analysis. Criteria pertinent for the selection of signature peptide markers in the implementation of multiple reaction monitoring (MRM) assays include analyte selectivity to minimize interference from matrix components, peptide/protein specificity and uniqueness, optimization of chromatographic and mass spectrometric performance, and characterization of potential protein post-translational modifications.24,25 Following these criteria, the selection of representative peptide allergen markers enabled the simultaneous detection and identification of multiple allergen proteins in a broad variety of foods.18−22 In the development of robust multi-allergen LC-MS/MS methods, however, the implementation of enhanced protein solubilization strategies, the characterization of processing effects (thermal and nonthermal) in allergen-incurred commercially processed food samples, and the incorporation of reference standards (i.e., isotopically labeled proteins or peptides) as chemical surrogates to the native allergen must additionally be considered to optimize the selection of allergen peptide markers and enable accurate quantitation in targeted mass spectrometry methods.26,27 This study aims to compare the performance of commercial immunochemical assays with that of a novel multi-allergen MS method for the detection and quantitation of egg, milk, and peanut in thermally processed bakery products (cereal bars and muffins) formulated in a commercial pilot-scale food production facility. Control and allergen-incurred cereal bars (1000 and 5000 ppm) and muffins (100 and 5000 ppm) were prepared in collaboration with General Mills. A comparison of method performance for each food matrix and incurred concentration provided important insights regarding effects of thermal processing, incurred allergen concentration, food formulation, and preferred allergen detection targets for quantitation. Understanding the advantages and limitations of B


Article


the instructed results format for each immunochemical method. When results were expressed as ppm allergen reagent, protein content values [whole egg powder (47.35%), nonfat dry milk (35.10%), or peanut (25.80%)] reported in the Food Allergen Handbook, 9th ed.29 and reproduced from the U.S. Department of Agriculture (USDA) National Nutrient Database (release 24) were used to convert to ppm of allergen protein. Recovery for each quantitative method was defined as the experimentally measured allergen protein concentration (Cmeasured), corrected for moisture content, divided by the incurred allergen protein concentration by dry weight (Cincurred), and expressed as a percentage: (Cmeasured/Cincurred) × 100. For all allergen-incurred samples, the protein content as measured by a micro-Kjeldahl method for each allergen ingredient (Supporting Information Table 1) was used to calculate the incurred allergen protein concentration (Cincurred). Preparation of Allergen Extracts for LC-MS/MS. All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless otherwise noted. Sample preparations were performed using Eppendorf Protein LoBind microcentrifuge tubes (Hauppauge, NY, USA). Optima grade solvents for LC-MS sample preparation and analysis were purchased from Fisher Scientific (Pittsburgh, PA, USA). Samples were weighed (0.100 ± 0.002 g) in 2.0 mL microcentrifuge tubes and defatted using a 1:13 (w/v) sample/hexane mixture by endover-end rotation (0.23g) at ambient temperature for 15 min. The hexane layer was decanted, and the defatting process was repeated two additional times. The defatted sample was vacuum-dried and resuspended in 1300 μL of extraction buffer [2 M urea, 50 mM Tris-buffered saline (TBS), pH 8.0, 25 mM dithiothreitol] supplemented with 10 pmol of yeast alcohol dehydrogenase (sample processing internal standard) and a plant protease inhibitor cocktail [4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride (AEBSF), bestatin, pepstatin A, E-64, leupeptin, and 1,10-phenanthroline in dimethyl sulfoxide (DMSO)]. The solubilized sample mixture was vortexed for 5 min at 1400 rpm (Eppendorf ThermoMixer), sonicated in a water bath for 10 min at 4 °C, and allowed to rotate end-over-end (0.23g) for 15 min, at room temperature. Centrifugation at 14000g for 10 min (20 °C) pelleted the undissolved sample debris. The supernatant was collected, and the remaining sample pellet was reextracted (1300 μL) using a barocycler pressure-assisted extraction chamber (NEP3229, Pressure Biosciences, South Easton, MA, USA). Samples were extracted in pulse tubes for 25 cycles at 0−35 kpsi in 15 s intervals. The extraction supernatant was clarified by centrifugation at 14000g for 10 min at 20 °C and combined with the primary sample extract. Estimated protein concentrations for all samples were determined using a Qubit Fluorometer (Invitrogen Life Technologies, Grand Island, NY, USA). A modified filter-aided sample preparation (FASP) sample concentration and digestion protocol was implemented for all sample extracts.30 Samples corresponding to 2 mg of total extracted protein were concentrated on a pre-equilibrated Amicon Ultra-0.5 mL centrifugal filter with Ultracel-10 membrane (Millipore, Billerica, MA, USA). The membrane was washed with 1 M urea and 50 mM ammonium bicarbonate (AmBC) (2 × 100 μL) to dilute the total urea concentration below a tolerated threshold for trypsin digestion. The concentrated sample (20 μL) was reconstituted in a Waters RapiGest SF acid labile surfactant (ALS) (Manchester, UK) at a final concentration of 0.1% in 50 mM AmBC (100 μL total volume). A digestion internal standard, yeast enolase (20 pmol), was added to each extract, and the samples were reduced and alkylated with dithiothreitol and iodoacetamide at 10 and 25 mM final concentrations, respectively. Sample membranes were washed with 1 M urea and 50 mM AmBC (100 μL) after each successive step to remove excess reagent. A sequencing grade modified trypsin (Promega, Madison, WI, USA) digestion was performed on membrane at a 1:100 enzyme to total protein concentration ratio in 50 mM AmBC overnight (16 h) at 37 °C (100 μL total volume). Peptide digestions were collected by centrifugation at 14000g for 20 min. The membrane was washed with an additional aliquot of 50 μL of 50 mM AmBC to ensure sufficient peptide collection. The digestions were quenched with the addition of 0.1% trifluoroacetic acid (TFA) and 2% acetonitrile and incubated for 30 min at 37 °C to degrade the ALS

these methods is crucial to enforce regulatory compliance, support allergen management in the food industry, and, most importantly, mitigate the risk to consumers with food allergies.

■

MATERIALS AND METHODS

Formulation of Allergen-Incurred Bakery Foods: Cereal Bars and Muffins. Control (0 ppm of allergen ingredient) and allergenincurred cereal bars (1000 and 5000 ppm) and muffins (100 and 5000 ppm) were prepared in collaboration with General Mills (Minneapolis, MN, USA) and designed to simulate conventional ingredient formulation and processing used in food manufacturing systems. Bakery products were incurred prior to heat treatment with dry allergen ingredients: whole egg powder (Rembrandt Foods, Rembrandt, IN, USA), nonfat dry milk (Dairy America, Fresno, CA, USA), and partially defatted (12% fat) dark roast peanut flour (Golden Peanut Specialty Products, Blakely, GA, USA). Each allergen-incurred concentration (100, 1000, and 5000 ppm) and control (0 ppm) sample matrix was prepared and processed as a single batch replicate. Percent nitrogen content (Micro-Analysis, Inc., Wilmington, DE, USA) was measured using a micro-Kjeldahl method for each allergen source material (Supporting Information Table 1). Product formulation is reported in Supporting Information Tables 2 and 3 for the control and allergen-incurred (low and high concentration) cereal bar and muffin foods, respectively. In brief, cereal bar dough formulated with a mixture of rolled oats, rice crisp, and corn flour was rolled into sheets of 15 mm thickness and 50 cm width. Cereal bars were baked in a rotating oven at 177 °C for 30 min. In muffin formulation, source ingredients were homogenized and the batter was deposited into lined trays with approximately 125 g of batter per muffin. Muffins were baked in a four-shelf rotating oven at 177 °C for 48 min. Moisture content analyses were completed for samples obtained at all processing stages (Supporting Information Table 4). Samples were stored at −20 °C. A processing flow diagram is shown in Figure 1 to illustrate the production of cereal bar (A) and muffin (B) samples, respectively. Three representative subsamples (500 g) were collected at various stages of processing (mixer, after roller/depositor, and final product), and each was homogenized with a food processor (Cuisinart DLC-X; Stamford, CT, USA) to ensure uniform distribution of raw ingredients, product assembly, and heat-treated final products.28 In all situations, good manufacturing practices were utilized to limit allergen crosscontact during production of bakery products manufactured in this investigation. Detailed records were collected throughout the entire manufacturing process including environmental conditions, temperature, time, and type of processing equipment (Supporting Information). Protein Quantitation by ELISA. The commercial ELISA kits used for allergen quantitation were: ELISA Systems Peanut (Queensland, Australia); Morinaga Egg, Milk, and Peanut (Yokohama, Japan); Neogen BioKits Peanut (Lansing, MI, USA); Neogen Veratox Egg, Milk, and Peanut (Lansing, MI, USA); and R-Biopharm RIDASCREEN FAST Ei/Egg Protein, Milk, Peanut (Darmstadt, Germany). A summary of antibody specificities, format of analytical results, and ranges of quantitation for each commercial assay are reported in Supporting Information Table 5. ELISA kits with corresponding analytical standards were used according to manufacturer protocols. ELISA results were determined spectrophotometrically using a SpectraMax M5 (Molecular Devices Corp., Sunnyvale, CA, USA) or BioTek μQuant (Winooski, VT, USA) plate reader and corresponding data analysis software (Molecular Devices Softmax Pro 5.3 or BioTek Gen5). For quantitation, standard curves were generated using a recommended linear or nonlinear regression model provided by the manufacturer or a four-parameter logistic calibration model, if none was provided. Two analytical aliquots of each subsample (A, B, C) obtained at each processing step (mixer, after roller/depositor, and final product) were extracted, diluted within an acceptable range of calibration, and assayed with at minimum duplicate wells for each allergen immunoassay. Measured allergen concentrations were quantitated following C


Article


precut uncoated silica tip emitter (360 μm o.d. × 20 μm i.d.) capillary with a 10 μm i.d. tip. Source parameters were as follows: curtain gas, 20 arbitrary units (au); CAD gas, high (12 au); ion spray voltage, 2400 V; source temperature, 150 °C; gas 1 pressure, 20 au; gas 2 pressure, 0 au. Using a syringe pump for infusion, peptide-dependent parameters (declustering potential, collision energy, and collision cell exit potential) were optimized for standards of each isotopically labeled heavy and native peptide target. Consensus values for declustering potential (80 V) and entrance potential (10 V) were determined for each analyte. Individual peptide transitions are reported in Supporting Information Table 6 with corresponding retention times, optimized collision energy voltages, and optimized collision cell exit potentials. The MS/MS data were collected in a scheduled MRM mode with low and unit resolution for quadrupole 1 (Q1) and quadrupole 3 (Q3), respectively, a 5 ms pause between mass ranges, a MRM detection window of 300 s, and a targeted scan time of 1.65 s. The optimized multi-allergen quantitative LC-MS/MS method monitored a total of 126 transitions (3 transitions/peptide) from 38 allergen peptide targets (13C15N-labeled synthetic and native peptides) in egg, milk, and peanut. Two internal standard peptides were monitored from each rabbit phosphorylase B and yeast enolase for quality control. Allergen quantitation was calculated from matrixmatched sample calibration curves generated for the most abundant MRM transition of each isotopically labeled synthetic peptide. All quantitation ions are reported with signal-to-noise ratios greater than 10:1 and confirmatory ions with ratios greater than 3:1. Retention time alignment and confirmed ion ratios between quantitation and confirmatory transitions were used to authenticate the detection of individual peptide targets. MS/MS product ion ratios for peptides with three monitored transitions were reported within a ±20% tolerance limit of relative ion abundances for synthetic peptide standards. Using a scheduled MRM method, analyte quantitation required a minimum of 10 points across the peak. Quantitative data analysis was performed using a modified targeted proteomics platform in Skyline v 2.6.31

surfactant. Prior to LC-MS/MS analysis, a 12.5 fmol/μL (25 fmol on column) solution of a rabbit phosphorylase B digest standard (Waters Corp.) was spiked into each sample as an injection quality control. LCMS certified clear glass total recovery vials with presplit PTFE/Silicone septa (Waters Corp.) were utilized as sample injection vials for all samples. Preparation of Isotopically Labeled Synthetic Peptides for Quantitation. Lyophilized isotopically labeled (13C15N) synthetic peptide standards (21st Century Biochemicals, Marborough, MA, USA) were reconstituted in 94% water, 5% acetonitrile, 1% formic acid, 1 pmol/μL angiotensin I, and 10 fmol/μL rabbit phosphorylase B digest. Peptide concentrations were calculated on the basis of monoisotopic molecular weight, peptide purity determined by HPLC with UV detection at 215 nm (≥97%), and peptide content measured by amino acid analysis. Standard stock solutions were stored at −20 °C or colder for 1 month. Matrix-matched calibration curves were prepared over the concentration range of 0.05−200 fmol/μL using neat stock solution of synthetic peptide standards diluted in control (0 ppm) cereal bar and muffin lysates. Matrix-matched calibration solutions were analyzed in technical injection replicates and evaluated as sample preparation duplicates. Interleaved blanks and quality control reference standards were incorporated throughout sample analysis to minimize the extent of peptide carry-over. Sample digests were enriched with 13C15N-labeled synthetic peptide standards prior to injection at 12.5 and 2.5 fmol/μL concentrations for corresponding high (5000 ppm) and low (1000, 100, and 0 ppm) allergen incursion concentrations, respectively. Matrix samples were diluted appropriately such that incurred peptide peak areas were measured within an acceptable range of calibration. Samples were analyzed in technical replicates and evaluated as sample preparation duplicates with interleaved blanks. Ion signal ratios for 13C15N-labeled and native peptide transitions were measured to calculate peptide concentrations. Normalized native peak areas were translated into parts per million of allergen protein by incorporating appropriate conversion factors including nominal protein molecular weight, relative allergen protein content in total protein composition, total extractable protein, moisture content, and sample preparation recovery (Supporting Information Figure 1). Relative allergen protein composition and total extractable protein concentrations were empirically calculated from individual allergen reference materials used in product formulation (whole egg powder, nonfat dry milk, and dark roast peanut flour). Sample preparation recovery was characterized by comparing the peak area response for individual isotopically labeled synthetic standards added at the time of digestion or prior to injection in a 0 ppm sample matrix control. The average analyte recovery during sample preparation was calculated to be 47.8 ± 6.0% for the cereal bar matrix and 40.4 ± 7.5% for the muffin matrix at multiple test concentrations. Liquid Chromatography and Tandem Mass Spectrometry for Quantitation. A Waters nanoACQUITY ultraperformance liquid chromatography (UPLC) system was operated under single-pump trapping mode, where samples were loaded onto a Waters nanoACQUITY UPLC Symmetry trapping column (20 mm × 180 μm i.d.) with 5 μm C18 particles (100 Å) at 5 μL/min for 3 min with 99.4% water, 0.5% acetonitrile, and 0.1% formic acid. Reverse phase separation was completed on a Waters ACQUITY UPLC analytical capillary column (100 mm × 100 μm i.d.) packed with 1.7 μm C18 BEH particles (130 Å). The column flow rate was maintained at 300 nL/min with a column temperature of 40 °C. Peptides were eluted with a linear gradient of 3−40% B in 30 min prior to re-equilibration (55 min total run time), where mobile phase A consisted of 0.1% (v/v) formic acid in water and mobile phase B of 0.1% (v/v) formic acid in acetonitrile. The weak needle wash solvent consisted of 96.9% water, 3% acetonitrile, and 0.1% TFA, and the strong needle wash solvent composition matched that of mobile phase B. The injection volume was 2 μL (5 μL sample loop). The autosampler temperature was thermostated to 4 °C. Mass spectrometry analysis was accomplished using an AB Sciex 6500 QTRAP (Foster City, CA, USA) operated in positive ionization mode. The AB Sciex NanoSpray III ion source was configured with a

■

RESULTS AND DISCUSSION ELISA. Bakery products are the most frequently recalled food type for undeclared allergens as listed in the FDA Reportable Food Registry.32 In this work, two model bakery food samples were selected representing a low-moisture (cereal bar) and a high-moisture (muffin) product. Food products were formulated in an industrial processing facility, and allergenic ingredients were incurred prior to heat treatment to ensure that the samples being studied were similar in nature to commercially available products. The use of such incurred food materials provided the capability to meaningfully evaluate processing effects on food allergen proteins in a complex food matrix and enabled a critical assessment of analytical method performance between commercial immunochemical kits. Quantitative measurements of egg, milk, and peanut allergen concentrations were compared for commercial immunochemical assays, where data are reported for samples collected after raw ingredient homogenization (mixer) and post-baking (final product). In both the cereal bar and muffin bakery products, parts per million allergen concentrations for the cereal bar samples obtained after the roller or for muffin samples taken at the depositor were determined to be statistically indistinguishable (two-tailed pairwise t test; p < 0.05) from the samples obtained at the mixer and therefore are excluded from data presentation. In all situations, no false positives were observed as all test kits yielded concentrations below the limits of detection for the allergen-free control samples. Calculated percent recoveries for each allergen and corresponding ELISA test kit in the 5000 ppm incurred cereal bar and 5000 ppm incurred muffin samples are reported in Figures 2−4, where optimal test kit performance (100% D


Article


Figure 2. Comparison of commercial ELISA kit performance for whole egg powder in 5000 ppm incurred (A) cereal bar and (B) muffin samples collected at the mixer (blue) and final product (red). Measured concentrations are normalized to ppm (μg/g) allergen protein and corrected for moisture content. The dotted line in each graph represents allergen recovery at 100%. Method performance as influenced by heat processing is reported above each ELISA kit as a ratio of measured allergen protein concentration in the final product and mixer samples.

Figure 3. Comparison of commercial ELISA kit performance for nonfat dry milk (NFDM) in 5000 ppm incurred (A) cereal bar and (B) muffin samples collected at the mixer (blue) and final product (red). Measured ppm concentrations are normalized to ppm (μg/g) allergen and corrected for moisture content. The dotted line in each graph represents allergen recovery at 100%. Method performance as influenced by heat processing is reported above each ELISA kit as a ratio of measured allergen protein concentration in the final product and mixer samples.

recovery) is illustrated as a dotted line in each graph. In this work, recovery is defined as the experimentally measured allergen protein concentration (Cmeasured) divided by the incurred allergen protein concentration by dry weight (Cincurred), and expressed as a percentage. Method performance as a function of heat processing (baking) is reported for each allergen and corresponding ELISA as a ratio of measured allergen protein concentration for samples collected at the final product and the mixer (FP:M). ELISA platforms exhibiting equivalent (or near-equivalent) measured concentrations in the pre- and post-processed matrix (FP:M = 1.0) were selected as reliable immunochemical methods for allergen quantitation in bakery products. ELISA results for low-level allergen-incurred samples are displayed in Supporting Information Figures 2 and 3 for the 1000 ppm cereal bar and 100 ppm muffin samples, respectively, and used to evaluate the influence of incurred allergen concentration on ELISA performance. Egg. In the 5000 ppm incurred cereal bar and muffin, the Morinaga (MO), Neogen Veratox (NE-V), and R-Biopharm RIDASCREEN (RB) ELISA kits estimated the incurred level of egg for samples obtained at the mixer with recoveries of 104.9 ± 6.7% (MO), 119.4 ± 11.0% (NE-V), and 166.7 ± 39.9% (RB) for the cereal bars and 125.2 ± 14.2% (MO), 197.4 ± 66.2% (NE-V), and 203.6 ± 21.37% (RB) for the muffins when averaged across all subsamples (Figure 2). Overestimations in egg concentration for samples collected at the mixer were more prevalent in the muffin as compared to the cereal bar. In the final product, substantial reductions in measured egg concentrations were noted for the NE-V and RB ELISA kits with reported recoveries of 2.1 ± 0.6% (NE-V) and 3.2 ± 0.5%

(RB) for the cereal bar and 8.6 ± 3.8% (NE-V) and 6.0 ± 1.2% (RB) for the muffin. In contrast, the MO kit yielded the highest percent recovery in the muffin samples at 99.6 ± 8.2% and the cereal bar samples at 76.7 ± 8.5%. Higher egg ELISA recovery values were displayed for muffin samples collected at the final product compared to baked cereal bar samples (Figure 2 and Supporting Information Figures 2 and 3). During baking, the maximum internal temperature of the high-moisture muffin matrix is predicted to be lower than that encountered in the low-moisture cereal bar. As a result, protein solubility and epitope recognition of selective targets in the muffins were likely not affected by the baking process to the same extent as they were in the cereal bar matrix. Differences in moisture content and matrix composition were therefore identified as plausible factors that can influence the performance of commercial ELISA kits and result in inconsistencies between the quantitation of egg proteins for immunochemical methods. The most prominent factor influencing the detection and quantitation of egg in examined food samples was heat processing (baking) (FP:M < 1). Within a food matrix, allergenic egg proteins can be denatured or chemically altered at elevated temperatures, which can hamper protein solubilization and antibody recognition. Although the protein allergen ovomucoid has a higher resistance to thermal denaturation compared to egg allergen proteins ovalbumin, ovotransferrin, and lysozyme,6,33−37 at elevated heat treatments (≥177 °C), changes in immunoreactivity of egg white proteins result in substantial ELISA underestimations of egg concentration.6 A similar finding is demonstrated in this work for the final product cereal bar and muffin samples when examined by the E


Article


119.8 ± 18.8% (MO), 130.1 ± 20.5% (NE-V), and 212.1 ± 62.2% (RB). For all milk ELISA kits evaluated, the muffin sample matrix yielded higher protein recoveries for NFDM in samples collected at the mixer as compared to the cereal bar. In contrast to egg, commercial milk allergen immunochemical methods yielded higher recoveries from baked cereal bar and muffin samples. The percent recoveries of NFDM from the 5000 ppm cereal bar final product samples were 80.6 ± 8.2% (MO), 49.4 ± 15.9% (NE-V), and 67.7 ± 4.3% (RB). In the case of the muffin final product samples, percent recovery values were considerably reduced as measured with the NE-V ELISA (17.3 ± 3.3%), whereas the MO (93.8 ± 5.3%) and RB (114.6 ± 17.5%) yielded calculated milk concentrations similar to the incurred concentration. The impact of heat processing on immunochemical detection methods can be evaluated by calculating the ratio of reported allergen concentration in the final product to samples collected at the mixer. Differences in antibody specificity, matrix composition, and protein extraction efficiency were identified as the primary factors influencing variations in the results obtained with different immunochemical methods for the detection of NFDM. Whereas the MO (casein) and RB (casein and β-lactoglobulin) immunochemical assays are characterized by antibody specificity to select milk allergen proteins, the NE‑V ELISA uses antibodies that recognize total milk protein. In cow’s milk, caseins and serum albumin have higher heat stability and, therefore, favored quantitative performance, as compared to the whey proteins, α-lactalbumin, β-lactoglobulin, and lactoferrin.39 The higher heat resistance of selective casein targets in the MO ELISA contribute to the higher reported recovery (FP:M) in samples collected post-baking, whereas the inclusion of whey protein targets for antibody specificity in the NE-V and RB assays limits detectability in the cereal bar and muffin final product samples. When heated, β-lactoglobulin interacts with food components by forming intermolecular disulfide bonds.40 The resulting alterations in the conformation of β-lactoglobulin and the formation of higher order molecular structures inhibit immunochemical recognition in the absence of appropriate additives (i.e., reducing agent) in the extraction buffer formulation. For the ELISA kits evaluated, NFDM recovery was lower for the cereal bar samples as compared to the muffin for samples collected at both the mixer and final product. Interestingly, however, whereas the trends between recovery values for the mixer and final product samples were consistent for both food matrices in the MO and RB ELISA kits, the NE-V ELISA displayed a more substantial change in recovery for the muffin (FP:M = 0.13) compared to the cereal bar (FP:M = 0.63). Subtle changes in protein chemistry resulting from processing conditions and interactions with secondary food ingredients are potential factors affecting the versatility of the NE-V assay for NFDM detection in high-moisture muffin-baked commodities. Consistent data trends are reported for the 1000 ppm cereal bar and 100 ppm muffin matrix samples (Supporting Information Figures 2 and 3), whereby the NE-V assay displayed lower allergen recovery (FP:M) in the muffin sample matrix as compared to the cereal bar. Peanut. The dark roast peanut flour used to incur cereal bar and muffin samples was analyzed with five quantitative ELISA kits, ELISA Systems (ES), Morinaga (MO), Neogen Veratox (NE-V), Neogen BioKits (NE-B), and R-Biopharm (RB), to evaluate the ability of the different kits to detect peanut in this highly processed peanut ingredient. In all examined immu-

Figure 4. Comparison of commercial ELISA kit performance for dark roast peanut flour in 5000 ppm incurred (A) cereal bar and (B) muffin samples collected at the mixer (blue) and final product (red). Measured ppm concentrations are normalized to ppm (μg/g) allergen and corrected for moisture content. The dotted line in each graph represents allergen recovery at 100%. Method performance as influenced by heat processing is reported above each ELISA kit as a ratio of measured allergen protein concentration in the final product and mixer samples.

NE-V (egg white protein) and RB (ovomucoid and ovalbumin) immunochemical assays. The MO (ovalbumin) ELISA, however, provides improved detection of egg proteins in thermally processed foods by including a surfactant, 1% sodium dodecyl sulfate (SDS), and a reducing agent, 7% 2mercaptoethanol, in the extraction buffer, which enables enhanced protein solubilization from processed foods.38 To enable protein recognition, the MO method uses antibodies that were developed to recognize the denatured form of allergen proteins and that are compatible with stringent extraction conditions.38 In this work, differences in protein extraction efficiency for heat-processed baked foods were determined to have a greater influence on immunochemical detection as compared to antibody specificity for the detection of egg. Similar trends are reported for the 1000 ppm cereal bar and 100 ppm muffin matrix samples (Supporting Information Figures 2 and 3), whereby the MO kit displays higher allergen recovery (FP:M) in the final product samples for all evaluated ELISA methods. Milk. The MO casein (94.1 ± 8.6% recovery) ELISA and NE-V total milk (78.4 ± 22.5% recovery) ELISA methods yielded amounts of nonfat dry milk (NFDM) within the expected values for 5000 ppm allergen-incurred cereal bar samples obtained at the mixer (Figure 3). The RB kit, however, overestimated the quantity of milk with a mean recovery of 146.8 ± 18.3% for all cereal bar subsamples. Similar results were reported for the muffin samples obtained at the mixer: F


Article


ES, NE-B, NE-V, and RB kits were concluded to be most sensitive in the detection of peanut allergens Ara h 3 followed by Ara h 1.10 The exception is the MO kit, which displays the highest sensitivity in the recognition of the 2S albumin protein allergens Ara h 2 and Ara h 6.10 In a raw peanut extract, Ara h 3 is identified to be the most abundant protein allergen (50−60%) followed by Ara h 1 at 12−16% total content.42−44 Several studies have characterized the propensity for heat-induced advanced glycation endproduct (AGE) modifications on peanut allergens Ara h 1 and Ara h 3.26,45 Changes in protein solubility and immunoreactivity for select allergens, as a result of thermal processing, limit the ability of ELISA kits to accurately quantify the amount of peanut protein in the dark roast peanut flour. 7,46−48 Consequently, immunochemical recognition of the most abundant and, in this instance, most chemically reactive proteins in peanut (Ara h 1 and Ara h 3) explains the reduced protein recoveries for the dark roast peanut flour in the majority of ELISA kits examined. The MO kit has the highest sensitivity in the detection of Ara h 2 and Ara h 6. These 2S albumin proteins are recognized to be the most relevant allergens in terms of allergenicity,49−51 and each is present in peanut kernels at a concentration of 6− 9%.43,52 Characterized by a protein structure heavily dominated by the formation of multiple disulfide bonds, the 2S albumin class of protein allergens has been identified to be less susceptible to protein digestion, aggregation, and thermal modification and, therefore, may be a better target for quantitation of peanut in processed foods.26,45,53 Furthermore, the MO kit differs significantly from the other examined kits with respect to sample extraction buffer containing the reducing agent 2-mercaptoethanol. The reduction of disulfide bonds serves to denature proteins in heat-processed samples, resulting in improved solubilization of aggregated and cross-linked proteins. Reliable quantitation not only is dependent on antibody specificity for immunochemical targets but also may be influenced by other food matrix constituents (e.g., fat, carbohydrates, emulsifiers), which chemically or physically modify allergen proteins and interfere with extraction and detection protocols. Across all examined ELISA methods, percent recovery of dark roast peanut was consistently higher for the muffin matrix. Interestingly, however, the Ara h 1specific NE-B assay exhibited a lower recovery in the highmoisture processed muffin (0.2 ± 0.3% recovery) as compared to the low-moisture cereal bar (3.0 ± 0.7% recovery). Reduced Ara h 1 solubility after moist heat treatment has been reported.7 Koppelman et al.54 demonstrated that heating Ara h 1 in solution resulted in an irreversible structural transition between 80 and 90 °C, leading to an increase in the β-structure and concomitant aggregation of the protein. In the high-moisture muffin matrix, the internal sample temperature was predicted to remain ≤100 °C during baking, potentially resulting in the aggregation of Ara h 1 and, consequently, the reduced recovery in the NE-B assay. Multi-allergen LC-MS/MS Quantitation. Method Characterization. The development of a multi-allergen mass spectrometry method required the critical evaluation of sample extraction parameters for enhanced protein recovery of egg, milk, and peanut allergens in the context of thermally processed, complex, food matrices. Working outside the confines of antibody recognition, mass spectrometry allows the use of more stringent extraction conditions for protein

nochemical assays, the percent allergen composition was underestimated: 3.17 ± 0.02% (ES), 2.2 ± 0.3% (NE-B), 26.8 ± 1.7% (MO), 1.6 ± 0.2% (NE-V), and 1.7 ± 0.1% (RB). Although the partially defatted dark roast peanut flour was measured to have a protein content of 54.68% (Supporting Information Table 1) by micro-Kjeldahl determination, extended roasting conditions likely resulted in thermally induced changes in protein chemistry and reduced concentrations of soluble protein.7,26,41 Analyzed with the same commercial ELISA methods, the 5000 ppm dark roast peanut incurred cereal bar and muffin products (Figure 4) similarly yielded reduced allergen recovery values. Averaged across all subsamples, allergen recoveries were 11.6 ± 0.2% (ES), 43.1 ± 4.1% (MO), 4.9 ± 1.5% (NE-B), 4.9 ± 0.3% (NE-V), and 7.6 ± 0.5% (RB) for cereal bar samples obtained at the mixer and 15.8 ± 1.3% (ES), 41.0 ± 5.1% (MO), 1.9 ± 0.6% (NE-B), 5.6 ± 0.4% (NE-V), and 12.4 ± 0.8% (RB) for muffin mixer samples. The current study demonstrates that protein quantitation by immunochemical methods is limited when food samples formulated with dark roast peanut flour are examined. The performance of commercial ELISA methods to accurately quantify the concentration of peanut was further impaired when the dark roast peanut ingredient was baked into a complex food matrix. The percent recoveries of dark roast peanut from the 5000 ppm cereal bar final product samples were 8.0 ± 1.5% (ES), 37.4 ± 5.8% (MO), 3.0 ± 0.7% (NE-B), 3.2 ± 0.2% (NE-V), and 3.9 ± 0.2% (RB). In the case of the muffin final product samples, percent recovery values were 9.4 ± 1.0% (ES), 27.1 ± 1.6% (MO), 0.2 ± 0.3% (NE-B), 4.6 ± 0.2% (NE-V), and 11.7 ± 0.9% (RB). In both sample types, processing stages (mixer and final product), and allergenincurred concentrations (Figure 4 and Supporting Information Figures 2 and 3), the MO ELISA yielded the highest protein recovery and the NE-B and NE-V assays the lowest recoveries. Method performance, as influenced by heat-processing (baking) and matrix interferences, was expressed as a ratio of allergen protein concentration measured in samples collected at the final product and the mixer. In the cereal bar samples, the average fold change (FP:M) across all immunochemical assays was 0.67 ± 0.13, with the MO kit demonstrating a minimal change in reported concentration at the final product (0.87). The muffin samples yielded a similar method performance (FP:M) across all examined ELISA kits (0.62 ± 0.32). These results suggest that the quantitation of peanut is less sensitive to the baking process used in the manufacturing of the cereal bar and muffin samples compared to the roasting conditions employed in the commercial production of the peanut flour. Although the detection of peanut proteins in processed food products can be accomplished with any of the ELISA kits studied, comparable results cannot be expected when the same sample is analyzed with different immunochemical assays. Variations in antibody specificities between investigated immunochemical assays explain some of the differences in recovery noted between kit manufacturers. The MO, NE-V, and RB kits specify antibody selection to total peanut protein, whereas the NE-B ELISA states specificity to the peanut allergen Ara h 1 and the ELISA Systems to protein allergens Ara h 1 and Ara h 2. As investigated by Jayasena et al.,10 despite reported antibody specificity to selected peanut protein(s) by ELISA manufacturers, reactivity against purified protein allergens Ara h 1, Ara h 2, Ara h 3, and Ara h 6 was reported to some extent in all examined immunochemical assays. The G


Article

Journal of Agricultural and Food Chemistry solubilization when used in conjunction with adequate sample cleanup procedures. In this study, we report the development of an empirically optimized protein extraction protocol incorporating the addition of a chaotropic denaturant (2 M urea) and a reducing agent (25 mM dithiothreitol) in a TBS extraction solution at pH 8.0. Denaturing extraction solutions augment protein solubility for thermally processed food allergens as compared to more traditional (i.e., phosphatebuffered saline) extraction systems.26 Whereas no single extraction condition may be optimally effective for all food allergens, matrix components, and processing conditions,53,55,56 the sample preparation methodology developed in this work effectively extracts egg, milk, and peanut allergen proteins from processed foods representing diverse composition of protein chemistries. Sample lysates were evaluated using an MS1-based comparative proteomic platform to survey potential allergen protein and peptide markers for analytical quantitation. Selection of candidate peptides was based on traditional criteria such as ionization and fragmentation efficiency, tryptic cleavage reproducibility, unique amino acid sequence composition, and chemical properties of amino acid constituents (e.g., posttranslational modifications and hydrophobicity).57 For complex processed food samples, however, characterization of processing effects (thermal and nonthermal),26,39,40,45,53,58−60 relative allergen protein abundance,42−44,52,61 structural diversity, isoform equivalence, and the utilization as an ingredient in processed foods62,63 were additional factors considered. Using these criteria, the following peptides were selected as optimal peptide markers for cereal bar and muffin samples: lysozyme (FESNFNTQATNR, NTDGSTDYGILQINSR) and ovalbumin (ELINSWVESQTNGIIR, GGLEPINFQTAADQAR, HIATNAVLFFGR) from egg, αS1-casein (FFVAPFPEVFGK, HQGLPQEVLNENLLR, YLGYLEQLLR) and β-lactoglobulin (LSFNPTQLEEQCHI, TPEVDDEALEK, VLVLDTDYK) from milk, and Ara h 1 (GTGNLELVAVR, NNPFYFPSR), Ara h 2 (CCNELNEFENNQR, CMCEALQQIMENQSDR, NLPQQCGLR), and Ara h 3 (FNLAGNHEQEFLR, SPDIYNPQAGSLK, WLGLSAEYGNLYR) from peanut. In this work, heavy (13C15N)-isotope-labeled reference peptides were synthesized as chemically identical surrogates for native marker peptides to confirm identifications and calculate concentrations of target molecules. The purity, peptide content, and stability of individual synthetic peptides were characterized for solution-based and matrix-matched samples. Targeted LC-MS/MS average peak areas for allergen peptide markers selected from egg, milk, and peanut are displayed in Figure 5 for the 5000 ppm incurred cereal bar (A) and muffin (B) samples. Comparing the ratio of peak area responses between the final product and mixer for different food matrix preparations (cereal bar and muffin), variances in candidate peptide behavior as a function of matrix composition and heat-processing platforms were characterized. Peptide targets exhibiting equivalent (or near-equivalent) solubilization in the pre- and post-processed matrix (FP:M = 1.0) were identified as optimal peptide markers. For model bakery products examined in this work, baking effects appeared more pronounced in the cereal bar relative to the muffin with average ion abundance ratios (FP:M) of 0.56 ± 0.27 and 0.68 ± 0.24, respectively, for all monitored allergen peptide marker candidates. Whereas the abundance of egg allergens lysozyme and ovalbumin was most strongly affected by thermal processing in the muffin matrix (0.43 ± 0.07), the milk

Figure 5. LC-MS/MS extracted ion peak areas for allergen peptide markers of egg (lysozyme and ovalbumin), milk (αS1-casein and βlactoglobulin), and peanut (Ara h 1, Ara h 2, and Ara h 3) allergen proteins in 5000 ppm incurred (A) cereal bar and (B) muffin mixer (blue) and final product (red) samples. Peak areas are corrected for moisture content. Method performance as influenced by heat processing is reported above each peptide as a ratio of measured allergen protein concentration in the final product and mixer samples.

allergens αS1-casein and β-lactoglobulin displayed the most prominent deviation (0.37 ± 0.27) in solubilized peptide recovery for the cereal bar samples. In both sample systems, peanut proteins were detected with similar peptide abundances in pre- and post-baked samples. This is largely due to the use of a dark roast (already thermally processed) peanut flour reagent for allergen incursion. Balancing peptide extraction efficiencies with peptide ionization and chromatographic features, a condensed list of allergen peptide markers was selected for quantitation from each allergen protein: NTDGSTDYGILQINSR (lysozyme), GGLEPINFQTAADQAR (ovalbumin), YLGYLEQLLR (αS1-casein), LSFNPTQLEEQCHI (β-lactoglobulin), NNPFYFPSR (Ara h 1), NLPQQCGLR (Ara h 2), and SPDIYNPQAGSLK (Ara h 3). Secondary peptides were monitored for confirmation and validation of protein allergen inferences where peptide identifications were authenticated on the basis of criteria including retention time alignment and fragment ion abundance ratios. Average peak areas for the mixer and final product sample collection in a 1000 ppm allergen-incurred cereal bar and a 100 ppm allergen-incurred muffin are displayed in Supporting Information Figures 4 and 5, respectively, and demonstrate similar trends in peak area responses for selected egg, milk, and peanut peptide targets. Quantitative Performance. Matrix-matched calibration solutions were prepared using control (0 ppm) cereal bar and muffin samples at equally spaced concentration intervals from 0.01 to 200 fmol on-column. Linear calibration regressions had an R2 value of ≥0.99 for all heavy-labeled synthetic peptides. A representative (100 fmol) extracted ion chromatogram is shown in Supporting Information Figure 6 to illustrate the separation of 19 isotopically labeled (13C15N) synthetic H


Article


content were less pronounced for the milk and egg protein allergen proteins. For each allergen, estimated recovery was evaluated as a single peptide marker from the most abundant protein in each allergen source, egg (GGLEPINFQTAADQAR; ovalbumin), milk (YLGYLEQLLR; αS1-casein), and peanut (SPDIYNPQAGSLK; Ara h 3), and compared to the average of all (%ALL) quantitative peptide markers for each allergen source. In cereal bar samples collected after mixing, sample recoveries were calculated for egg (85.0 ± 0.3%GGL; 110.3 ± 8.5%ALL), milk (79.7 ± 12.5%YLG; 63.7 ± 10.6%ALL), and peanut (86.7 ± 2.3%SPD; 64.3 ± 11.0%ALL). After baking, recovery values declined and were consistent between the three incurred allergens: egg (60.8 ± 0.2%GGL; 77.5 ± 0.7%ALL), milk (62.3 ± 10.2%YLG; 46.8 ± 8.0%ALL), and peanut (60.7 ± 1.4%SPD; 48.6 ± 1.2%ALL). The muffin samples collected at the mixer yielded recoveries indicative of efficient sample extractability for egg (98.6 ± 9.9%GGL; 117.8 ± 13.0%ALL), milk (87.7 ± 3.6%YLG; 69.7 ± 4.2%ALL), and peanut (100.2 ± 1.8%SPD; 73.6 ± 5.6%ALL). High protein recoveries were measured also for the baked final products for milk (75.1 ± 4.2%YLG; 56.6 ± 3.2%ALL) and peanut (70.2 ± 1.8%SPD; 55.0 ± 2.0%ALL); however, egg protein recovery displayed reduced values (45.2 ± 1.4%GGL; 53.0 ± 3.9%ALL). Recoveries calculated from the quantitation of a single peptide target were determined to be statistically equivalent to recovery values based on the quantitation of all peptide targets (two-tailed pairwise t test; p < 0.05). Overall, the average allergen-incurred protein recovery in the 5000 ppm samples collected at the mixer were consistent for the muffin (95.5 ± 10.5%) and cereal bar samples (83.8 ± 13.3%), as calculated from egg, milk, and peanut quantitative peptides. For final product samples, the average recovery was determined to be 63.5 ± 4.4% for the muffins and 61.3 ± 10.2% for the cereal bars. From these results it may be concluded that variations in matrix composition (i.e., corn, oat, rice, wheat) or sample moisture content have less of an impact on sample extraction efficiency and thereby analytical method performance than the food-processing conditions (i.e., baking). Similar findings were observed for the 1000 ppm incurred cereal bar and 100 ppm muffin as displayed in Supporting Information Figures 7 and 8, respectively, indicating recovery was not dependent on incurred allergen concentration. In summary, the multi-allergen LC-MS/MS method demonstrates near-equivalent (unbiased) detection of selected egg, milk, and peanut peptides in the final product matrix for both the cereal bar and muffin incurred bakery products. The potential for food processing-induced changes in allergen protein structure and interactions with other food ingredients complicate the detection of allergen proteins and necessitate an allergen analysis strategy that can monitor multiple peptides corresponding to specific allergen proteins for reliable food allergen detection and accurate quantitation. Comparison of ELISA and MS for Allergen Quantitation. Allergen quantitation was compared for cereal bar and muffin samples obtained at several production stages using commercial immunochemical (ELISA) methods and a novel multi-allergen LC-MS/MS assay. With the exception of dark roast peanut flour, ELISA provided reliable quantitative measurements for total allergen content in unbaked foods. Thermal processing, however, negatively affected the ELISA detection of the incurred allergens. A summary of calculated protein recoveries from the 5000 ppm final product cereal bar and muffin is reported in Table 1, where measured protein

peptides over a 30 min optimized LC gradient. Quantitation by isotope dilution mass spectrometry was performed using matrix-matched calibration curves where ion signal ratios for heavy-labeled and native peptide transitions were measured to calculate peptide concentrations. A comparison of calculated percent recovery (

Multi-allergen Quantitation and the Impact of ... - ACS Publications

Recommend Documents