Advertisement

Analytical and Bioanalytical Chemistry

, Volume 410, Issue 22, pp 5653–5662 | Cite as

In house validation of a high resolution mass spectrometry Orbitrap-based method for multiple allergen detection in a processed model food

  • Rosa Pilolli
  • Elisabetta De Angelis
  • Linda Monaci
Research Paper
Part of the following topical collections:
  1. Food Safety Analysis

Abstract

In recent years, mass spectrometry (MS) has been establishing its role in the development of analytical methods for multiple allergen detection, but most analyses are being carried out on low-resolution mass spectrometers such as triple quadrupole or ion traps. In this investigation, performance provided by a high resolution (HR) hybrid quadrupole-Orbitrap™ MS platform for the multiple allergens detection in processed food matrix is presented. In particular, three different acquisition modes were compared: full-MS, targeted-selected ion monitoring with data-dependent fragmentation (t-SIM/dd2), and parallel reaction monitoring. In order to challenge the HR-MS platform, the sample preparation was kept as simple as possible, limited to a 30-min ultrasound-aided protein extraction followed by clean-up with disposable size exclusion cartridges. Selected peptide markers tracing for five allergenic ingredients namely skim milk, whole egg, soy flour, ground hazelnut, and ground peanut were monitored in home-made cookies chosen as model processed matrix. Timed t-SIM/dd2 was found the best choice as a good compromise between sensitivity and accuracy, accomplishing the detection of 17 peptides originating from the five allergens in the same run. The optimized method was validated in-house through the evaluation of matrix and processing effects, recoveries, and precision. The selected quantitative markers for each allergenic ingredient provided quantification of 60–100 μgingred/g allergenic ingredient/matrix in incurred cookies.

Keywords

High resolution mass spectrometry Multi-allergen detection Processed matrix Incurred samples Peptide marker In house validation 

Introduction

The development of sensitive and reliable reference methods for allergen detection represents a priority of food industries and regulatory bodies in order to enforce the current allergen labeling regulation and promptly detect any inadvertent contamination occurring along the food chain [1]. Different approaches have been proposed on this regard over the years, mainly directed towards the allergenic protein itself or a marker indicative of its presence in a food, with relative advantages and/or limitations, even in the absence of internationally approved reference methods. In this context, the role of mass spectrometry (MS) has considerably changed over the last two decades moving from protein characterization and elucidation to qualitative and quantitative tool for the unambiguous detection of allergenic ingredients in foods [2]. MS methods allow multiplex allergen detection with high molecular-level specificity in a single run, enabling the unequivocal identification of allergens in a variety of food commodities. The first MS screening method for food allergens dates back 2011, when Heick et al. developed a multi-allergen MS methods for the simultaneous detection of seven allergens in one run [3]. Since then, several other methods were proposed mainly employing low-resolution mass analyzers such as triple quadrupole or ion traps [4, 5, 6, 7, 8, 9, 10, 11, 12]. However, the lack of consensus in analytical performance definitions (e.g., limit of detection/quantification and recovery), in spiking procedures and in reporting units prevents full comparability of the results obtained among different laboratories.

In general, quantitative MS analysis requires preparation of calibration standards in real matrix due to matrix effects that can significantly alter the MS signal. As a consequence, method performance assessment during method development is typically achieved by preparing spiked food materials where the manufactured food is spiked with multiple allergenic ingredients or representative extract [13, 14, 15]. However, this procedure might lead to an over-estimation of the results since both recovery and sensitivity can be impaired by the type of food processing affecting protein stability and extractability. Therefore, incurred food materials should be prepared when undertaking a method validation study.

Beyond using correct protocols for the inclusion of allergenic ingredients within a food product, the reporting unit must also be taken into account. Regulatory bodies define a food allergen as a whole food commodity [16], excluding the direct association to a specific protein or DNA sequence, therefore referring to the total allergenic commodity in the total food matrix. In order to translate these quantities into allergen protein thresholds capable of eliciting an allergic reaction, the trend of the scientific community has been to report the contamination levels as total protein content of the allergenic commodity in the complex food matrix. This would facilitate to perform a direct correlation of method sensitivity with the minimum doses (if known) eliciting immunoreaction.

So far only a few studies have been accomplished to develop multi-target approaches for the detection of different classes of allergens into incurred food matrices [3, 7, 17, 18], and to the best of our knowledge none of them was based on HR-MS platforms. In this investigation, we reported for the first time the application of high resolution Orbitrap™-based mass spectrometry to the multiple detection of different allergenic ingredients (milk, egg, soy, hazelnut, and peanut) in both spiked and incurred cookies, selected as processed and complex model food. In particular, different acquisition schemes were compared, and aiming at exploiting the potential of HR-MS platform, sample preparation was kept to a minimum. Three acquisition modes, full-MS (FS), targeted-selected ion monitoring with data-dependent fragmentation (t-SIM/dd2), and parallel reaction monitoring (PRM), were independently optimized and compared in terms of sensitivity. In addition, performances provided by such hybrid HR-MS platform were compared with an optimized HPLC-SRM method recently developed in our lab employing a dual pressure linear ion traps MS platform for the analysis of the same processed food [18].

Materials and methods

Reagents

Trypsin Gold Mass Spectrometry Grade was purchased from Promega (Milan, Italy). Solvent and other reagents, together with the allergenic ingredients egg powder (EP) and skimmed milk powder (MP) were purchased from Sigma–Aldrich (Milan, Italy). Pre-cooked soy flour (SF) was purchased from a local retailer. Roasted peanuts and hazelnuts were provided by Besana s.p.a. (San Gennaro Vesuviano, NA, Italy), minced and used without defatting. Cellulose acetate syringe filters, 1.2 μm (size 25 mm), were purchased from Labochem Science S.r.l. (Sant’Agata di Battiati, CT, Italy), and polytetrafluoroethylene syringe filters, 0.2 μm (size 4 mm), were purchased from Sartorius Italy S.r.l. (Muggiò, MB, Italy). Disposable desalting cartridges PD-10 were purchased from GE Healthcare Life Sciences (Milan, Italy).

Production of allergens mixture stock solution

A raw mixture of egg powder, skimmed milk powder, pre-cooked soy flour, ground roasted peanuts, and ground roasted hazelnuts was prepared by weighting 5 mg of each ingredient on a plastic tube. The mixture was extracted with 24 mL of Tris–HCl buffer 20 mM at pH = 8.2 with ultrasound assistance (probe sonicator Vibracell™, Sonics & Materials Inc., Newton, CT, USA) in the following conditions: time 30 min, amplitude 70, pulse 10sON/2sOFF, and output ca 14 W (overheating was controlled with an ice bath). The extract was centrifuged (15 min at 1800 rcf) and the supernatant was filtered through 1.2 μm acetate cellulose membranes. The resulting mix allergens stock solution (concentration equivalent to 208 μgingr/mL) was used both for building up standard calibration curves and for recovery calculation. The stock solution was cleaned-up with disposable size exclusion cartridge (SEC) following instruction (loading: 2.5 mL of sample, elution 3.5 mL of 50 mM ammonium bicarbonate). The eluate was diluted with 50 mM ammonium bicarbonate to cover the concentration range 5–100 μgingr/mL, providing full calibration curves referred to as “standard solution curves”.

Production of spiked and incurred cookie samples

Blank, spiked, and incurred cookie samples were prepared according to the recipe already described in a previous paper [18]. All the prepared cookies were ground mechanically and sifted with a 1-mm sieve. Allergen concentrations reported hereafter were defined as microgram per gram allergenic ingredient/matrix (μgingr/g), unless otherwise specified.

Of ground cookie, 1.2 g was extracted with 24 mL of Tris–HCl buffer 20 mM at pH = 8.2 with ultrasound assistance in the same conditions reported before (time 30 min, amplitude 70, pulse 10sON/2sOFF, and output ca 14 W). The extract was centrifuged for 15 min at 1800 rcf and the supernatant filtered through 1.2 μm acetate cellulose membranes. Of the filtered sample, 2.5 mL was cleaned-up by disposable SEC and eluted with 3.5 mL of 50 mM ammonium bicarbonate.

Calibration curves of both incurred and spiked samples were prepared by serial dilutions of spiked/incurred cookie extracts with appropriate volumes of blank cookie extract covering the concentration range of 20–600 μg/g (spiked samples—six concentration levels), and 10–600 μg/g (incurred samples—six concentration levels).

Enzymatic digestion

Trypsin digestion was carried out in 300 μL of the SEC eluted fraction (either standard solutions or cookie samples). Before digestion, protein denaturation (15 min at 95 °C), reduction (addition of 15 μL of 50 mM dithiothreitol solution and incubation for 30 min at 60 °C), and alkylation were performed (addition of 30 μL of 100 mM iodoacetamide solution and incubation for 30 min at room temperature). Finally, 4 μL of trypsin solution (1 μg/μL in acetic acid 50 mM) was added to the sample so that a ratio always higher than 1/50 (enzyme/protein) was attained to facilitate a complete enzymatic digestion. Reaction was stopped after about 14 h by acidification (HCl 1 M) and the final digests were 0.2 μm-filtered before analysis.

LC-MS analysis

For LC-MS analyses, a system consisting of an UHPLC pump provided with an autosampler, an electrospray ionization interface connected to a hybrid quadrupole-Orbitrap™ mass spectrometer Q-Exactive Plus (Thermo Fisher Scientific, San Josè, USA) was used. Peptide separation was accomplished on an Acclaim PepMap100, C18 column, (3 μm, 100 Å, 1 × 150 mm) at a flow rate of 60 μL/min, using a binary gradient composed H2O + 0.1% formic acid (solvent A) and CH3CN/H2O 80:20 + 0.1% formic acid (solvent B). The gradient elution program was as follows: 0–25 min linear from 15 to 40% B; step change to 50% B then isocratic for 5 min; step change to 90% B then isocratic for 10 min; step change to 15% B then isocratic for 15 min for column conditioning. The MS system runs under three acquisition modes with the following instrumental parameters:
  • FS: resolution 70 k, AGC target 3e6, maximum injection time 100 ms, scan range 400–1500 m/z.

  • Timed t-SIM/dd2: time-scheduled inclusion list with 2-min wide window centered on the average retention time (tR), SIM microscan 1, resolution 35 k, AGC target 5e5, maximum injection time 100 ms, loop count 8, isolation window 2.0 m/z and offset 0.4 m/z, MS/MS resolution 17.5 k, AGC target 2e5, maximum injection time 100 ms, loop count 10, isolation window 2.0 m/z and offset 0.4 m/z, normalized collision energy 30, underfill ratio 0.3%, charge exclusion 1,4-8,>8, peptide match preferred, exclude isotopes on.

  • PRM: time-scheduled inclusion list with 2-min wide window centered on the average tR, MS/MS resolution 35 k, AGC target 5e5, maximum injection time 130 ms, isolation window 2.0 m/z, isolation offset 0.4 m/z, normalized collision energy 30.

Results and discussion

Home-made cookies, selected as processed food matrix, were fortified with five different allergenic ingredients, including skim milk powder, whole egg powder, soy flour, ground hazelnut, and ground peanut (not defatted), representing allergens included in the EU regulation for mandatory food labeling. We prepared both spiked and incurred cookie samples, where allergens were added after or before the thermal processing, respectively. A sharp and easy sample preparation was carried out, limited to a 30-min ultrasound-aided protein extraction with buffer solution followed by clean-up on size exclusion disposable cartridges, as already detailed in a previous investigation [18].

Comparison of three different acquisition modes for multiple allergen detection in spiked cookies

Several acquisition modes can be set on the Q-Exactive™ platform to perform a multi-target analysis. In this paper, we investigated three different acquisition schemes including Full scan acquisition (FS), timed-selection ion monitoring and data-dependent acquisition (t-SIM/dd2), and parallel reaction monitoring acquisition (PRM) for multi-allergen analysis. Each approach enables to perform quantitative determination of peptide markers in processed foods, with different features. The FS mode allows acquiring a broader information on a wide m/z range of precursor ions with high resolution/accuracy without any preliminary peptide markers selection. Such untargeted data acquisition scheme typically represents a source of information also for further retrospective analysis, but as main drawback the FS mode does not allow the reliable identification of amino acids sequence based on MS/MS spectra matching. On the contrary, the t-SIM/dd2 mode allows a highly reliable peptide identification on account of the accurate detection of both the peptide precursor and the relevant MS/MS fragmentation pattern; however, as targeted analysis, it requires preliminary selection of markers and any additional information about other species is lost. Finally, PRM approach represents an alternative targeted choice, which would help in analyzing complex samples since only MS/MS pattern of selected ions are acquired, reducing contribution from the matrix, but no information about the precursor ions are available. As general statement, the choice between the two last options should be guided basing on the signal background which is directly related to the complexity of the system under study. PRM is MS/MS based so has inherently higher specificity, however the overall “base” sensitivity is lower as the intensity of the available precursor ions is distributed across multiple fragments; however with highly complex matrix, PRM works properly in improving signal-to-noise ratio. As targeted analysis modes, both t-SIM/dd2 and PRM required information about the peptide markers identifying the presence of the specific allergenic ingredients, therefore an untargeted Full-MS/dd2 acquisition was first accomplished on spiked cookie extracts. Data were processed via a commercial software for protein identification Proteome Discoverer v.1.4 (Thermo Fisher Scientific), setting some filtering constrains in order to highlight the most reliable peptide markers. In particular, only matches with high peptide confidence (false discovery rate ≤ 0.01), high accuracy on precursor ion identification ≤ 5 ppm, and peptides retrieved as unique for the allergenic protein were taken into consideration. All these features, confirmed by the visual inspection of the MS/MS assigned spectra, allowed compiling the list of 17 peptide markers reported in Table 1, which were used to set up both the t-SIM/dd2 and PRM acquisition modes. Given the high number of targets to be monitored simultaneously, the cycle time for each acquisition mode was taken into consideration to guarantee a proper sampling speed across the chromatographic peak and allow sensitive and accurate quantitative information to be retrieved.
Table 1

List of the unique peptide markers selected for each allergenic ingredient, both the theoretical m/z ratios and the experimental tR were used for building up timed t-SIM/dd2 and PRM instrumental methods

Allergenic ingredient

Target protein

Peptide marker sequence

Theoretical m/z (charge)

tR (min)

EGG

Gal d2

GGLEPINFQTAADQAR (GGL)

844.4236 (+ 2)

12.6 ± 0.1

EGG

Gal d2

LTEWTSSNVMEER (LTE)

799.3618 (+ 2)

5.0 ± 0.2

EGG

Gal d2

YPILPEYLQCVK (YPI)

761.9023(+ 2)

16.6 ± 0.1

MILK

Bos d 5

TPEVDDEALEK (TPE)

623.2959 (+ 2)

3.1 ± 0.1

MILK

Bos d 9

YLGYLEQLLR (YLG)

634.3559 (+ 2)

20.5 ± 0.1

MILK

Bos d 9

FFVAPFPEVFGK (FFV)

692.8686 (+ 2)

21.7 ± 0.1

SOY

Gly m 5

ESYFVDAQPK (ESY)

592.2851 (+ 2)

5.6 ± 0.1

SOY

Gly m 6

SQSDNFEYVSFK (SQS)

725.8279 (+ 2)

11.9 ± 0.1

SOY

Gly m 6

FYLAGNQEQEFLK (FYL)

793.8961 (+ 2)

14.1 ± 0.1

HAZ

Cor a9

ADIYTEQVGR (ADI)

576.2882 (+ 2)

3.2 ± 0.1

HAZ

Cor a9

TNDNAQISPLAGR (TND)

678.8469 (+ 2)

3.5 ± 0.1

HAZ

Cor a9

QGQVLTIPQNFAVAK (QGQ)

807.4541 (+ 2)

13.9 ± 0.1

HAZ

Cor a9

ALPDDVLANAFQISR (ALP)

815.4334 (+ 2)

19.9 ± 0.1

PEA

Ara h1

AMVIVVVNK (AMV)

486.7992 (+ 2)

7.6 ± 0.1

PEA

Ara h1

GTGNLELVAVR (GTG)

564.8222 (+ 2)

9.1 ± 0.1

PEA

Ara h1

EGEQEWGTPGSEVR (EGE)

780.8499 (+ 2)

5.0 ± 0.2

PEA

Ara h1

VLLEENAGGEQEER (VLL)

786.8786 (+ 2)

3.1 ± 0.1

Several instrumental parameters were independently optimized including resolution, AGC target and maximum injection time in order to achieve a good compromise among detection sensitivity, accuracy, and cycle time. In FS mode, the reliability of peptide identification and quantification was totally dependent on the accuracy in the precursor ion detection, therefore the top three resolution levels available on the Q-Exactive™ mass spectrometer were tested (namely 280, 140, and 70 k at 200 m/z), corresponding to different transient lengths. Under optimized chromatographic conditions, an average peak width at half a maximum of 15 s was obtained, therefore even at the highest resolution level (transient length of 1024 ms) we achieved the minimum of 10 acquisition points across the chromatographic peak, meeting the general recommendations for accurate quantitative analysis. According to our results, the resolution set at 70 k was deemed satisfactory enabling an accuracy in the detection of all marker peptides better than 1 ppm, with good sensitivity. As for t-SIM/dd2 acquisition, given the intrinsically longer cycle time (independent acquisitions of the precursor ions each combined with the relevant MS/MS spectrum), we set a lower resolution of 35 k (at 200 m/z) for the t-SIM acquisition and of 17.5 k for MS/MS acquisition. In addition, we also included a time scheduling of the markers acquisition windows based on their retention time; given the good chromatographic separation achieved, a maximum of four peptides were monitored simultaneously in the same time range with no need for multiplex C-trap filling. An example of the sampling speed across the peak was shown in the Electronic Supplementary Material (ESM) in Fig. S1 for the first 4 min of the chromatographic run where the maximum of four partially overlapping markers was present. Given the targeted nature of the analysis mode, the proposed instrumental set-up represented a good compromise between accuracy, sensitivity, and cycle time for the selected peptides. Analogously, similar set-up was kept also for PRM acquisitions as time scheduling of the acquisitions (2-min-long windows centered on the average retention time), resolution (35 k at 200 m/z), AGC target, but with a slightly higher maximum injection time (see “Materials and methods” section for details).

In order to compare the performance of the three acquisition modes, the blank and spiked cookies samples (600 μg/g) were progressively diluted to simulate intermediate spiking levels, covering the concentration range from 20 to 600 μg/g for full calibration curves. Such set of samples was subjected to clean-up and enzymatic digestion like previously described and the resulting peptide pools separated by HPLC and analyzed under optimized conditions. Extracted ion chromatograms (XIC) with an accuracy of 5 ppm were generated for each peptide marker considering the ion current of the precursor ion or of its fragments, when MS/MS spectra were acquired (sum of the four most intense fragments). Concerning the t-SIM/dd2 acquisitions, which provided both information on precursor and fragments ions, it was observed that a higher sensitivity can be achieved monitoring the intact peptide ions, whereas MS/MS spectra were considered only for qualitative identification purposes. XIC peak areas were integrated for all the dilution levels in the three acquisition modes and the experimental data were interpolated by linear regression (see ESM Fig. S2). For each acquisition mode, an evaluation of the sensitivity provided by each peptide recorded was performed based on the direct comparison of the curve slopes, the final goal being to identify the best quantitative markers. In Table 2, we summarized the fitting parameters for the best quantitative markers selected. For these peptides, response linearity was confirmed across the whole concentration range under investigation with linear correlation coefficients, R2, ranging from 0.977 to 0.998 among the different peptide markers (Table 2). The three modes provided a slight difference in the sensitivities achieved and in all cases, the t-SIM/dd2 turn out the best compromise among sensitivity, specificity, and reliability, given the dual information available about both the precursor ion and the MS/MS spectrum. An overlay of typical XICs referred to the peptide markers acquired in t-SIM/dd2 mode at 100 μg/g concentration level is shown in Fig. 1 (spiked cookie).
Table 2

Summary of the fitting parameters (slope and linear correlation coefficients, R2) calculated by linear regression of experimental XIC peak areas (5 ppm accurate extraction of theoretical precursor ions) recorded for spiked cookies samples in three different acquisition modes

Analysis mode

Parameter

LTE, 799.362 (EGG)

YLG, 634.356 (MILK)

FYL, 793.896 (SOY)

ALP, 815.433 (HAZ)

GTG, 564.822 (PEA)

Full-MS

R 2

0.993

0.991

0.997

0.996

0.981

Slope

3370 ± 90

10,500 ± 300

1860 ± 30

6040 ± 120

1170 ± 50

t-SIM/dd2

R 2

0.997

0.995

0.998

0.998

0.992

Slope

10,900 ± 200

22,300 ± 600

6300 ± 900

20,400 ± 300

3570 ± 100

PRM

R 2

0.990

0.992

0.977

0.989

0.989

Slope

1280 ± 40

10,300 ± 300

420 ± 30

3780 ± 130

710 ± 30

Fig. 1

Typical extracted ion chromatograms (5 ppm accuracy on the theoretical m/z) acquired in t-SIM/dd2 acquisition mode for spiked cookies (100 μg/g) under optimized conditions

Evaluation of the t-SIM/dd2 method for multi-allergens detection in incurred cookies

The optimized t-SIM/dd2 instrumental settings were finally applied to the analysis of thermally processed samples, produced at laboratory scale by direct inclusion of the allergenic ingredients in the recipe at known amount before dough making and baking (incurred cookies). We expected that both the matrix composition and the food thermal processing would affect peptide detection at different extents on the base of the specific allergenic commodity/marker protein.

In order to characterize such effects, we devised an ad hoc experiment in which we generated and compared three sets of calibration curves: (i) allergens standard curves, namely protein extracts from dry mixture of allergens powders/minces in buffer solution (no interfering compound from the matrix, no thermal processing); (ii) multi-allergens spiked cookies curves (allergens not baked into the food matrix); (iii) multi-allergens incurred cookies curves (allergens baked into the food matrix). The same sample preparation was applied to all three sets of samples (see “Materials and methods” section for details) and the integrated XIC peak areas plotted as a function of the allergenic ingredient concentration. The slopes calculated by the linear regression fitting of the calibration curves based on the experimental data obtained were featured as indicators of the detection sensitivity (see Fig. 2). The three sets provided huge differences in sensitivity as displayed in the graph by the semi-logarithmic scale utilized to graphically compare the slopes. According to what shown, both the matrix composition and the thermal processing applied to the food appeared to negatively affect the detection sensitivity. In order to assess the entity of these effects, we calculated the percentage ratio between spiked cookies slopes (blue bars in Fig. 2) and standard solutions slopes (dark blue bars in Fig. 2) for each peptide marker as an estimator of the matrix effect and the percentage ratio between incurred cookies slopes (light blue bars in Fig. 2) and spiked cookies slopes (blue bars in Fig. 2) as indicator of the processing effect. The first two columns of Table 3 reported such evaluation for all the monitored peptide markers and, in addition, we also calculated in the third column their combined effect. Noteworthy, matrix and processing effects in combination can lower the sensitivity by 2–3 orders of magnitude in the worst case, with a certain variability among the five allergenic ingredients on the base of the specific peptide marker monitored. Hazelnut and peanut displayed a relatively more resistant behavior to the thermal processing, likely due to the previous roasting process both ingredients had already undergone. This evidence furtherly supported the need for the inclusion of incurred food materials when undertaking a method development and validation study.
Fig. 2

Evaluation of matrix and food processing effects on the detection sensitivity by direct comparison of the calibration curves slopes referred to allergen standard solutions, allergen spiked cookies and allergens incurred cookies

Table 3

Summary of the main features calculated for each peptide marker. Matrix, processing, and their combined effects on detection sensitivity were calculated by dividing calibration curves slopes like reported in the first line. Recoveries were averaged over two different concentration levels. The intra-day and inter-days precision was reported as percent coefficient of variation (CV%) at a fixed concentration (600 μg/g) and was evaluated for five measurements over five working days. The inter-days precision was calculated only when the mean values on different days resulted not to be significantly different at 95% confidence level by one-way ANOVA test

 

Peptide markers

Matrix effect (%)

\( \frac{Slope_{spiked}}{Slope_{standard}}\times 100 \)

Processing effect (%)

\( \frac{Slope_{incurred}}{Slope_{spiked}}\times 100 \)

Combined effect (‰)

\( \frac{Slope_{incurred}}{Slope_{standard}}\times 1000 \)

Averaged recovery (%)

Intra-day CV % (%)

Inter-day CV % (%)

Egg

GGL (844,424)

3.03 ± 0.11

6.3 ± 0.3

1.90 ± 0.08

60.4 ± 1.5

4

4

Egg

LTE (799,362)

13.8 ± 0.6

1.42 ± 0.12

1.96 ± 0.18

95 ± 7

7

N.A.*

Egg

YPI (761,902)

2.51 ± 0.11

4.4 ± 0.3

1.11 ± 0.06

51 ± 4

7

7

Milk

TPE (623,296)

1.22 ± 0.06

4.6 ± 1.3

0.56 ± 0.15

67 ± 3

50

40

Milk

YLG (634,356)

2.63 ± 0.11

26.9 ± 1.0

7.1 ± 0.2

69.8 ± 1.6

3

6

Milk

FFV (692,869)

10.3 ± 1.4

3.1 ± 1.0

3.2 ± 1.1

55 ± 7

40

N.A.*

Soybean

ESY (592,285)

3.8 ± 0.2

16.6 ± 0.8

6.3 ± 0.3

57.7 ± 0.8

6

15

Soybean

SQS (725,828)

1.46 ± 0.15

2.9 ± 0.5

0.42 ± 0.06

63.4 ± 1.4

10

11

Soybean

FYL (793,896)

6.78 ± 0.19

2.83 ± 0.12

1.92 ± 0.09

70.9 ± 1.7

2

4

Hazelnut

ADI (576,288)

16.0 ± 0.9

22.2 ± 1.0

35.5 ± 1.8

52 ± 4

4

N.A.*

Hazelnut

TND (678,847)

21.4 ± 1.7

8.6 ± 0.7

18 ± 2

55 ± 7

9

N.A.*

Hazelnut

QGQ (807,454)

16.4 ± 1.0

11.2 ± 0.7

18.3 ± 0.8

60 ± 7

4

3

Hazelnut

ALP (815,433)

29.7 ± 1.9

9.1 ± 0.4

27.0 ± 1.7

57 ± 14

10

12

Peanut

AMV (486,799)

3.16 ± 0.18

6.2 ± 1.2

1.9 ± 0.4

59 ± 4

3

N.A.*

Peanut

GTG (564,822)

4.0 ± 0.2

16.4 ± 0.9

6.6 ± 0.3

64.5 ± 0.8

3

4

Peanut

EGE (780,850)

7.2 ± 0.3

34.1 ± 1.4

24.6 ± 1.0

68 ± 4

11

N.A.*

Peanut

VLL (786,879)

10.0 ± 0.9

4.0 ± 0.2

4.0 ± 0.4

60.4 ± 0.5

17

N.A.*

*not applicable

In house validation study of the developed LC-HR MS method

A further step is method validation, to assure that the method is suitable for its intended purpose and that the same method will perform equally in all laboratories.

The validation process includes a number of steps to demonstrate that the developed method complies with the established performance criteria set in the guidelines issued by different international standardization bodies. As far as validation of MS methods for food allergens detection is concerning, there has been a need over the years to harmonize analytical methods for food allergen analyses. The full validation of a multi-allergen method is currently hard to achieve and a limitation is represented by the few reference materials currently available on the market. Both reference allergenic ingredients and reference-incurred matrices require a systematical characterization of total protein and allergen content, a proper calculation of the total protein content, and a guaranteed homogeneity and stability of the allergenic commodities used for the validation study. In absence of these, a preliminary characterization of the raw materials at least in terms of total protein content and homogeneity should be carried out, unless such information were made already available from the provider, for confident inclusion in validation studies. Taking into considerations all these limitations, we featured in the following section a first in-house validation of the LC-HR MS method under development.

Trueness

Since no certified reference materials for baked food matrices incurred with the five selected allergens are available, the most feasible option to calculate the trueness in a single-laboratory method validation is by recovery experiments. We compared spiked cookies fortified with the five allergenic ingredients at two concentration levels 600–300 μg/g, and submitted to the whole sample preparation with blank cookies fortified only after the SEC-based clean-up with a proper amount of mix allergen stock solution corresponding to 100% recovery. The percentage ratio of XIC peak areas for each marker recorded from these paired samples represented the actual recovery of the analytical method, which however does not account for the contribution of the thermal processing. In Table 3 the averaged recoveries at both concentration levels tested were reported, and for all the peptide markers the values ranged from 51 to 95%.

Precision

Method repeatability within the same laboratory was calculated as intra-day and inter-day precision of the analytical method (percent coefficient of variation in peak areas at a fixed concentration, CV %). In particular, we tested first the intra-day repeatability by analyzing three independent extracts prepared and each injected five times. The experimental data were compared by one-way ANOVA at 95% confidence level and the three samples resulted not to be significantly different for all peptide markers, thus allowing pooling all data from the first day. The calculated intra-day CV% was lower than 17% for most peptides, which is deemed to be an acceptable variability in absence of any internal standard employed in the study taking into account that the sample prep also included an enzymatic digestion step. Only 2 peptides out of 17 both belonging to milk proteins (TPE, FFV) showed a higher variability comprised between 50 and 40%, probably ascribed to the uncomplete release by proteolysis and/or poor stability in processed matrix (see Table 3). The latter were then excluded from further discussion. In addition, the analyses of the same samples were iterated over five consecutive working days to evaluate the inter-day precision. The mean values obtained on different days were compared by a one-way ANOVA test at 95% confidence level, turning out not to be significantly different only for nine peptide markers resulted to be very stable peptides. In these cases, the data acquired over different days were averaged providing an inter-day CV% ranging from 3 to 15% depending on the marker (see Table 3).

Sensitivity

Method sensitivity was assessed in incurred cookie samples by calculation of both limit of detection (LOD) and quantification (LOQ) from the linear regression fitting parameters of the calibration curves. LOD and LOQ were defined as 3 and 10 times, respectively, the standard deviation on the calibration line intercept divided by the slope of the calibration curve. Notwithstanding several options for LOD/LOQ calculation have been reported in literature, we deemed this approach to be more confident since it relies on the trueness and precision of the whole data set acquired on a defined concentration range. In Table 4, we reported a summary of the fitting parameters and the LOD/LOQ values calculated for five selected peptides that resulted to be the best quantitative markers for each allergenic ingredient. Noteworthy this selection was coherent with the features presented in Table 3, because the most sensitive markers represented also the best compromise among sensitivity to processing effect, recovery, intra-day and inter-day precision. The analytical method provided very interesting sensitivity with LOD values ranging from 17 to 30 μg/g and LOQ values ranging from 60 to 100 μg/g, the best performance achieved for milk and hazelnut markers. As an example, Fig. 3 showed an overlay of typical XIC peaks and relevant isotopic pattern as a function of the allergen concentration for the hazelnut quantitative marker. Even at the lowest concentration investigated, 10 μg/g, a mass accuracy better than 3 ppm was achieved for peptide identification with a still clear and meaningful isotopic pattern. However, it is worthy to be noticed that the lowest the concentration the worst the quality of MS/MS fragmentation spectrum obtained (e.g., for ALP peptide the dd2 fragmentation was not activated for concentration levels ≤ 20 μg/g).
Table 4

Summary of the fitting parameters calculated by linear regression of experimental XIC peak areas (5 ppm accuracy in the extraction of theoretical precursor ions) recorded for incurred samples in t-SIM/dd2 mode. Limit of detection (LOD) and quantification (LOQ) were calculated as follow: LOD = 3 × SDintercept/slope; LOQ = 10 × SDintercept/slope. (*values scaled as μg of total protein per g of matrix by means of correction factors)

Quantitative peptide marker

R 2

Slope

LOD-LOQ (μgingred/gmatrix)

LOD-LOQ (μgprot/gmatrix)*

GGL, 844.424 (EGG)

0.993

605 ± 14

30–100

14–50*

YLG, 634.356 (MILK)

0.997

6000 ± 80

17–60

6–20*

FYL, 793.896 (SOY)

0.995

445 ± 8

30–90

10–30*

ALP, 815.433 (HAZ)

0.996

1850 ± 30

20–80

4–12*

GTG, 564.822 (PEA)

0.995

375 ± 7

30–90

7–24*

Fig. 3

Overlay of typical extracted ion chromatogram peaks and isotopic patterns for the peptide ALP (m/z 815.433) selected as hazelnut quantitative marker into incurred cookie samples as a function of the incurring concentration (defined as whole allergen commodity). Mass accuracy in precursor ion identification was always better than 3 ppm

In this paper, the concentration levels tested and the LOD/LOQ values obtained were defined as microgram of whole allergenic commodity per gram of matrix, following the rationale behind the spiking procedure and the EU regulation. However, since it is advised to report the contamination levels as total protein content of the allergenic ingredient, in agreement with the definition of minimum threshold doses eliciting an immunoreaction, a tentative conversion of the experimental LOD/LOQ values was provided in Table 4. As for the conversion factors, we referred to the total protein content of allergenic ingredients reported by the USDA Food Composition Database [19] for the following ingredients: dried whole egg (code 01133, protein content 48.05%), dried nonfat milk (code 01091, protein content 36.16%), dry-roasted peanuts without salt (code 16390, protein content 24.35%), dry-roasted hazelnut without salt (code 12122, protein content 15.03%). In addition, since for soybean artificial contamination we used a commercial pre-cooked flour purchased from a local retailer, we used the information on food composition reported on the label: protein content 35.4%.

The overall performance provided by the hybrid quadrupole/Orbitrap™ mass spectrometer operating in t-SIM/dd2 mode was quite challenging, considering that a sharp and fast sample preparation protocol was used without any further customization. In order to highlight the relevance of our achievements, we presented in Table 5 a comparison of the main features of the presented micro-HPLC-HR-MS approach with our previous work based on a low-resolution dual pressure linear ion trap. In that work, a 50-fold higher matrix equivalent was injected on the column, thanks to the use of an automated on-line pre-enrichment step on a C18 cartridge. As a matter of the fact, by converting the LOQ values (μg/g) into absolute quantifiable amount of allergenic ingredient (LOQ values multiplied by injected matrix equivalent), results obtained proved to be quite challenging. In prospective, a further optimization of the sample preparation for example by addition of some denaturing agents in the extraction buffer (to improve the extractability from processed foods) or by implementation of a peptide enrichment step could further improve the detection sensitivity.
Table 5

Comparison of the main features of the presented micro-HPLC-HR-MS approach with our previous work based on a low-resolution dual pressure linear ion trap [18], both applied to the multiple detection of different allergenic ingredients in home-made incurred cookies. In order to highlight the difference in sensitivity, a conversion of LOQ values in absolute amount of quantifiable allergenic ingredient was performed based on the injected matrix equivalent

 

This work

Reference [18]

Sample preparation

- Ultrasound-aided protein extraction

- Clean-up by size exclusion disposable cartridge

- Protein thermal denaturation, reduction and alkylation

- Tryptic digestion

Chromatographic set-up

Micro-HPLC

Automated on-line SPE enrichment/clean-up coupled to HPLC

Injected matrix equivalent

0.0714 mgmatrix

3.57 mgmatrix

Mass spectrometer and analysis mode

Hybrid quadrupole/Orbitrap mass spectrometer, t-SIM/dd2 mode

Dual pressure linear ion trap, SRM mode

Sensitivity: LOQ values and minimum quantifiable allergen

EGG

100 μgingred/gmatrix

7 ngingred

30 μg ingred/gmatrix

110 ngingred

MILK

60 μgingred/gmatrix

4 ngingred

20 μg ingred/gmatrix

70 ngingred

SOY

90 μgingred/gmatrix

6 ngingred

19 μg ingred/gmatrix

70 ngingred

HAZ

80 μgingred/gmatrix

6 ngingred

20 μg ingred/gmatrix

70 ngingred

PEA

90 μgingred/gmatrix

6 ngingred

40 μg ingred/gmatrix

140 ngingred

Conclusions

In the present investigation, we presented the first example of a micro-HPLC-HR-MS method based on a simple analytical workflow for multiple detection of different allergenic ingredients in processed food matrix. A sharp protocol for sample preparation was tested in order to challenge the HR-MS performances provided by a hybrid quadrupole-Orbitrap™-based mass spectrometer. Several acquisition modes were investigated scouting for the most suitable option to be applied for the multiple detection of allergens in complex food matrices. Timed t-SIM/dd2 turned out the best choice as good compromise between sensitivity and accuracy, accomplishing the detection of 17 peptides, belonging to five allergens in the same run and providing also MS/MS spectra. The selected quantitative markers for each allergenic ingredient provided limits of detection in incurred cookie samples ranging between 17 and 30 μg/g (defined as whole commodity) depending on the specific allergen. The sensitivity achieved could be further improved by an ad hoc optimization of the sample preparation with selective enrichment of marker peptides. In addition, future development of the work could include the use of stable isotope-labeled forms of the peptides as external calibrants for proper recovery evaluation and absolute quantitative estimates.

Notes

Acknowledgments

Roberto Schena is kindly acknowledged for his technical aid in performing MS measurements. Besana group S.p.A. is also acknowledged for kindly providing hazelnuts and peanuts.

Funding

The work was funded by the project Safe & Smart—Nuove tecnologie abilitanti per la food safety e l’integrità delle filiere agro-alimentari in uno scenario globale—National CL.AN-Cluster agroalimentare nazionale programma area 2. The equipment used in this work was supported by the “Biodiversità per la valorizzazione e sicurezza delle produzioni alimentari tipiche pugliesi, BioNet-PTP” project (Cod. 73) funded by Programma Operativo Regionale Puglia FESR 2000-2006 - Risorse liberate - Obiettivo Convergenza.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.

Informed consent

Not applicable.

Supplementary material

216_2018_927_MOESM1_ESM.pdf (220 kb)
ESM 1 (PDF 219 kb)

References

  1. 1.
    Monaci L, Visconti A. Mass spectrometry-based proteomics methods for analysis of food allergens. Trends Anal Chem. 2009;28:581–91.CrossRefGoogle Scholar
  2. 2.
    Monaci L, Pilolli R, De Angelis E, Mamone G. Mass spectrometry in food allergen research. In: Picó Y, editor. Comprehensive analytical chemistry. Elsevier Science and Technology; 2015. pp. 359–393.Google Scholar
  3. 3.
    Heick J, Fischer M, Pöpping B. First screening method for the simultaneous detection of seven allergens by liquid chromatography mass spectrometry. J Chromatogr A. 2011;1218:938–43.CrossRefPubMedGoogle Scholar
  4. 4.
    Pilolli R, De Angelis E, Godula M, Visconti A, Monaci L. Orbitrap™ monostage MS versus hybrid linear ion trap MS: application to multi-allergen screening in wine. J Mass Spectrom 2014;49:1254–1263.Google Scholar
  5. 5.
    Parker CH, Khuda SE, Pereira M, Ross MM, Fu TJ, Fan X, et al. Multi-allergen quantitation and the impact of thermal treatment in industry-processed baked goods by ELISA and liquid chromatography-tandem mass spectrometry. J Agric Food Chem. 2015;63:10669–80.CrossRefPubMedGoogle Scholar
  6. 6.
    Korte R, Monneuse JM, Gemrot E, Metton I, Humpf H-U, Brockmeyer J. New high-performance liquid chromatography coupled mass spectrometry method for the detection of lobster and shrimp allergens in food samples via multiple reaction monitoring and multiple reaction monitoring cubed. J Agric Food Chem. 2016;64:6219–27.CrossRefPubMedGoogle Scholar
  7. 7.
    Planque M, Arnould T, Dieu M, Delahaut P, Renard P, Gillard N. Advances in ultra-high performance liquid chromatography coupled to tandem mass spectrometry for sensitive detection of several food allergens in complex and processed foodstuff. J Chromatogr A. 2016;1464:115–23.CrossRefPubMedGoogle Scholar
  8. 8.
    Korte R, Brockmeyer J. MRM3-based LC-MS multi-method for the detection and quantification of nut allergens. Anal Bioanal Chem. 2016;408:7845–55.CrossRefPubMedGoogle Scholar
  9. 9.
    De Angelis E, Pilolli R, Monaci L. Coupling SPE on-line pre-enrichment with HPLC and MS/MS for the sensitive detection of multiple allergens in wine. Food Control. 2017;73:814–20.CrossRefGoogle Scholar
  10. 10.
    Hoffmann B, Münch S, Schwägele F, Neusüß C, Jira W. A sensitive HPLC-MS/MS screening method for the simultaneous detection of lupine, pea, and soy proteins in meat products. Food Control. 2017;71:200–9.CrossRefGoogle Scholar
  11. 11.
    Huschek G, Bӧnick J, Lӧwenstein Y, Sievers S, Rawel H. Quantification of allergenic plant traces in baked products by targeted proteomics using isotope marked peptides. LWT-Food Sci Technol. 2016;74:286–93.CrossRefGoogle Scholar
  12. 12.
    Chen Q, Zhang J, Ke X, Lai S, Tao B, Yang J, et al. Quantification of bovine β-casein allergen in baked foodstuffs based on ultra-performance liquid chromatography with tandem mass spectrometry. Food Addit Contam Part A. 2015;32:25–34.CrossRefGoogle Scholar
  13. 13.
    Monaci L, Pilolli R, De Angelis E, Godula M, Visconti A. Multi allergen detection in food by micro high performance liquid chromatography coupled to a dual cell linear ion trap mass spectrometry. J Chromatogr A. 2014;1358:136–44.CrossRefPubMedGoogle Scholar
  14. 14.
    Korte R, Lepski S, Brockmeyer J. Comprehensive peptide marker identification for the detection of multiple nut allergens using a non-targeted LC–HRMS multi-method. Anal Bioanal Chem. 2016;408:3059–69.CrossRefPubMedGoogle Scholar
  15. 15.
    Monaci L, De Angelis E, Bavaro SL, Pilolli R. High-resolution Orbitrap™-based mass spectrometry for rapid detection of peanuts in nuts. Food Addit Contam: Part A. 2015;32:1607–16.CrossRefGoogle Scholar
  16. 16.
    European Commission Commission Directive 2007/68/EC. Off J Eur Union. 2007;L310:11–4.Google Scholar
  17. 17.
    Monaci L, Pilolli R, De Angelis E, Carone R, Pascale M. LC-tandem mass spectrometry as a screening tool for multiple detection of allergenic ingredients in complex foods. ACTA IMEKO. 2016;5:5–9.CrossRefGoogle Scholar
  18. 18.
    Pilolli R, De Angelis E, Monaci L. Streamlining the analytical workflow for multiplex MS/MS allergen detection in processed foods. Food Chem. 2017;221:1747–53.CrossRefPubMedGoogle Scholar
  19. 19.
    United States Department of Agriculture Agricultural Research Service. In: USDA Food Composition Databases. USA National Agricultural Library. 2017. https://ndb.nal.usda.gov/ndb/search/list. Accessed 10 July 2017.

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2018

Authors and Affiliations

  • Rosa Pilolli
    • 1
  • Elisabetta De Angelis
    • 1
  • Linda Monaci
    • 1
  1. 1.Institute of Sciences of Food ProductionNational Research Council (ISPA-CNR)BariItaly

Personalised recommendations