Introduction

Food allergy is a growing topic in the life sciences, affecting about 2 % of the global population (Ortolani et al. 2001). The presence of allergens could be intentional or could occur accidentally via cross-contamination at any stage of food production. One hundred nineteen food allergic reactions (22.4 %) are considered to be due to hidden allergens (Añíbarro et al. 2007). Since it is known that food allergens could be life-threatening to sensitive allergic individuals, legislation in several countries, especially in the USA and in the European Union, were put in place to increase food safety (European Commission 2000, 2003, 2006).

Taking into account that the dose eliciting allergic reactions has been estimated in the order of subunits or units of micrograms depending on the allergenic food (Bindslev-Jensen et al. 2002), a considerable reduction of the detection and quantification limits as well as an improvement in the reliability of the analytical methods available for allergen determination are of paramount importance to safeguard consumers’ health.

Analytical methods for qualitative and quantitative determinations of food allergens including immunochemical, nucleic acid-based and mass spectrometric (MS)-based methods have been recently reviewed (Kirsch et al. 2009; Monaci and Visconti 2009; Sancho et al. 2010; Sancho and Mills 2010; Schubert-Ullrich et al. 2009). In the last years, the number of applications of MS techniques in this field has considerably increased for unambiguous identification and accurate quantification of proteins and peptides (Monaci and Visconti 2009; Kerkaert et al. 2010; Kuppannan et al. 2011; Monaci and van Hengel 2008; Monaci et al. 2010; Shefcheck et al. 2006). In this context, our research group successfully proposed analytical methods for the selective determination of hidden allergens in foods by selecting univocal biomarker peptides (Careri et al. 2008; Bignardi et al. 2010; Mattarozzi et al. 2012). Beyond the accuracy, one of the main advantages of MS-based methods is the possibility of multi-allergen determination by simultaneously monitoring of several food allergens in a single analysis (Bignardi et al. 2010; Mattarozzi et al. 2012; Heick et al. 2011; Ansari et al. 2011). In particular, in our most recent work, we devised a MS-based method capable of investigating simultaneously four lupin allergens in pasta and biscuits in one single liquid chromatography–electrospray–tandem mass spectrometry (LC-ESI-MS2) run (Mattarozzi et al. 2012). The food matrix can have a considerable effect on the extraction and determination of allergenic food residues (Taylor et al. 2009). Chocolate is one of the most challenging food matrices. Indeed, dark chocolate contains a large amount of polyphenolic compounds which can react with food proteins, thus masking the target proteins under investigation. Some commercially available enzyme-linked immunosorbent assay (ELISA) test kits have been used for the analysis of hazelnut and peanut content in dark chocolate samples (Hurst et al. 2002; Pele et al. 2007; Poms et al. 2005). However, it has to be stated that the most current available commercial ELISA test kits are suitable for qualitative screening but fail for accurate quantitative assays (Hurst et al. 2002; Poms et al. 2005).

In this context, taking into account the complexity of dark chocolate matrix, and that peanut and tree nuts are among the major food allergens (Luccioli 2012) eliciting the most severe reactions (Eller et al. 2012), the scope of the present work was the improvement of the sensitivity of the previously devised shotgun LC-ESI-MS2 method (Bignardi et al. 2010) through the addition of a rapid size-exclusion solid-phase extraction step for sample clean-up, and its application on dark chocolate and biscuit samples for five nut multi-allergen determination. In fact, the reliability of the LC-MS analysis strongly depends on the overall quality of the sample extract. The sample treatment protocol was devoted to the analysis of five proteins corresponding to the five most important nut allergens (Ana o 2, cashewnut; Cor a 9, hazelnut; Pru 1, almond; Jug r 4, walnut; Ara h3/4, peanut). To assess analytical method performance, validation of the method was performed in terms of sensitivity, selectivity, accuracy, recovery, limits of detection and quantification in the food matrices under investigation. The method was then applied for the analysis of nut traces in commercially available commodities.

Materials and Methods

Reagents and Chemicals

Water was obtained with a MilliQ element A10 System (S. Francisco, CA, USA). Bicarbonate buffer (pH 8.0), acetonitrile (HPLC purity), trifluoroacetic acid (TFA), formic acid (analytical reagent grade), trypsin from bovine pancreas, dithiothreitol and iodoacetamide were purchased from Sigma Aldrich (Milan, Italy). Roasted peanuts (Arachis hypogea), walnuts (Juglans regia), hazelnuts (Corylus avellana), almonds (Prunus dulcis), cashewnuts (Anacardium occidentale), commercial biscuits and dark chocolate were obtained at a local food store. Biscuits and dark chocolates used as blank samples did not report the precautionary label “may contain trace of nuts, milk and soy”.

Sample Treatment

The protein extracts were prepared by adding 10 mL of 50 mM bicarbonate buffer (pH 8.0) to 1 g of ground sample, and then performing a 6 h extraction at 60 °C. Food extract protein was then reduced (10 mM dithiothreitol, 45 min, 37 °C) and alkylated (40 mM iodoacetamide, 45 min, room temperature, dark room). Protein purification by size-exclusion step was performed according to our previous work (Mattarozzi et al. 2012). More precisely, to purify proteins from the sample extracts the size-exclusion columns Bio-Spin 6 (Bio-Rad Laboratories, Milan, Italy) with a 6 kDa exclusion limit were used. Firstly, the column was centrifuged for 2 min in a swinging bucket centrifuge at 1,000×g in order to remove the packing buffer, then it was conditioned three times with 500 μL of 50 mM bicarbonate buffer (pH 8.0) and centrifuged for 1 min. One hundred microliters of the extract were loaded directly on the column and centrifuged for 4 min. The purified sample (100 μL) was thus collected in a 2.0 mL microcentrifuge tube. Enzymatic digestion was carried out by adding 8 μL of 4 mg/mL trypsin solution to 100 μL of purified protein extract (protein/enzyme ratio 50:1) and carried out at 50 °C overnight. The digestion reaction was quenched with 2 μL TFA.

LC-MS Analytical Method

LC separation was carried out on a HPLC system (Thermo Electron Corporation, San José, CA, USA) with a flow-rate of 200 μL/min. A C18 Kinetex (100 mm × 2.1 mm, 2.7 μm particles) (Phenomenex, CA, USA) column thermostated at 25 °C was used. Gradient elution separation was carried out under the following conditions [(A) aqueous formic acid 0.1 % solution (v/v) and (B) 0.08 % (v/v) formic acid in acetonitrile]: solvent B was set at 2 % for 2 min, then delivered by a linear gradient from 2 to 25 % in 9 min and to 60 % in 1 min. Solvent B was maintained at 60 % for 1 min before column re-equilibration (7 min).

Mass spectrometer device was an LTQ XL linear ion trap (Thermo Electron Corporation) equipped with a pneumatically assisted ESI interface. The sheath gas (nitrogen, 99.9 % purity) and the auxiliary gas (nitrogen, 99.9 % purity) were delivered at flow rates of 45 and 5 arbitrary units, respectively. Source conditions were as follows: ESI voltage 3.5 kV, capillary voltage 20 V, tube lens 20 V, capillary temperature 270 °C. In a preliminary step, raw tryptic digest extracts from a mixture of the five nuts under investigation were analysed in data-dependent acquisition mode under full scan conditions in the 300–1300 m/z mass range and excluding any singly charged species. Collision gas (He) pressure was set at 2.1 × 10−3 mbar in the collision cell, with a collision energy equal to 30 arbitrary units. The product ion spectra obtained from data-dependent acquisition were processed using BioWorks 3.3 software (Thermo Electron Corporation) with stringent criteria, against a FASTA database containing only the five proteins selected and specifying iodoacetamine derivative of cysteine (+57) as a fixed modification. From the peptide list generated, two biomarker peptides for each protein were selected with the same criteria used in the previous work (Bignardi et al. 2010), i.e. signal intensity of the most abundant fragment of MS/MS spectrum, no post translational modification sites, absence of missed cleavages and sequence specificity (BLAST search, www.blast.ncbi.nlm.nih.gov/Blast.cgi). Transitions for selected reaction monitoring (SRM) analysis were carefully selected on the basis of sensitivity and selectivity criteria (Table 1).

Table 1 SRM transitions monitored for the five allergen proteins (MS2 mode, collision energy 30 eV)
Full size table

Method Validation

Validation of the whole analytical method was performed on fortified samples of biscuits and dark chocolate according to Eurachem guidelines (Eurachem 1998). To this aim, food samples were fortified by adding a homogenized mixture of the five ground nuts at different amounts before extraction. To obtain a homogeneous nut mixture, same amounts of the five ground nuts were mixed and spun on a mixer for 30 min, by refrigerating with liquid nitrogen. Measurements were carried out by monitoring the most abundant MS2 transition for each peptide (Table 1). The detection limits (LOD) and the quantification limits (LOQ) were calculated from the calibration curve as 3 s/slope and 10 s/slope, respectively, where s is the standard deviation of the blank signal obtained from ten independent blank measurements. Linear dynamic range was investigated starting from LOQ values of each peptide to 10,000 mg/kg (ten concentration levels, two replicates for each level). Precision was assessed as RSD% for each compound in terms of intra-day repeatability and inter-day repeatability. For this purpose, the within-day repeatability was evaluated by performing three independent extractions of each matrix fortified with 50 and 100 mg nuts/kg matrix and three LC-MS2 injections for each extract in the same day. The inter-day repeatability was calculated on 5 days by performing five independent extractions of the matrix fortified with 50 and 100 mg nuts/kg matrix and three LC-MS2 injections for each extract. Matrix effect was tested by performing a t test between curves calculated on matrix samples and on standard solutions. Recovery of the whole method was evaluated by analysing in triplicate a matrix fortified sample at two concentration levels, i.e. 5 and 100 mg nuts/kg matrix in the case of biscuits, 50 and 100 mg nuts/kg matrix in the case of dark chocolate, using different batches of nuts and matrices. The recovery was calculated as a ratio in percent (%) of determined and added nut content (\( =(\text{determined}\,\text{milligrams}\,\text{per}\,\text{nut}/\text{per}\,\text{kilogram}\,\text{matrix}/\text{added}\,\text{milligrams}\,\text{nut}/\text{kilogram}\,\text{matrix})\times 100 \)).

Results and Discussion

Sample Treatment and Peptide Selection

For the analysis of complex matrices as dark chocolate, an efficient sample clean-up step is fundamental for an accurate detection and quantification of trace allergenic proteins. Since chemical and physical matrix properties have significant effect on the extraction efficiency, it is important the use of an extraction and purification procedure applicable for the determination of hidden allergens in most food matrices. For these reasons, with the aim of improving method accuracy and sensitivity in terms of limit of detection and quantification we evaluated the use of a clean-up step by size-exclusion chromatography before enzymatic digestion. For this purpose, we used columns having an exclusion limit of 6 kDa specifically designed to clean protein samples quickly (the process takes about 10 min) in such a way to retain low molecular weight components as carbohydrates and lipids. Furthermore, with the aim of improving the digestion protocol in terms of repeatability and efficiency, protein structures were unfolded by reduction/alkylation steps performed just before enzymatic digestion. The implementation of this sample preparation procedure resulted in a sequence coverage average of about 50 %, a greater value with respect to that obtained in the previous work (∼35 %) (Bignardi et al. 2010). On the basis of the new sequence coverage and increased signal intensity, for all allergens except for Cor a 9, it was possible to select some peptides exhibiting higher sensitivity with respect to those chosen in the previous study (Bignardi et al. 2010). It has to be stated that the peptide LDALEPTNR, related to walnut, occurs also in Car i 4, a major allergen in pecan nut.

Representative LC-IT-MS2 chromatograms of the ten targeted peptides from the analysis of a tryptic digest of purified biscuit and chocolate blank matrices fortified with five nut mix are illustrated in Figs. 1 and 2.

Fig. 1
figure 1

Selected reaction monitoring LC-ESI-MS2 chromatograms of the ten targeted peptides from a tryptic digest of a purified biscuit sample. Blank matrix was fortified with a mixture of the five nuts 0.01 % (w/w)

Full size image
Fig. 2
figure 2

Selected reaction monitoring LC-ESI-MS2 chromatograms of the ten targeted peptides from a tryptic digest of a purified dark chocolate sample. Blank matrix was fortified with the five nuts 0.1 % (w/w)

Full size image

Method Validation

Method validation was carried out in terms of limits of detection and quantification, linear dynamic ranges, accuracy, recovery, and selectivity.

As illustrated in Tables 2 and 3, excellent LOD and LOQ values were obtained for each nut. In particular, LODs ranging from 0.1 to 1.3 mg nut/kg for biscuits and from 5 to 15 mg nut/kg for chocolate were calculated. LOQ values were in the 0.3–4.5 mg nut/kg biscuit and in the 18–50 mg nut/kg chocolate ranges. As expected, for all peptides, LOD and LOQ values in chocolate resulted higher, of about one order of magnitude, than those obtained in biscuits owing to increased complexity of the matrix.

Table 2 Validation data for the quantitative analysis of nut allergens in biscuits under MS2 acquisition mode
Full size table
Table 3 Validation data for the quantitative analysis of nut allergens in dark chocolate under MS2 acquisition mode
Full size table

For biscuit samples, linearity was excellent up to four orders of magnitude for all peptides with determination coefficients (r 2 ≥ 0.99). For chocolate samples, a good linear dynamic range (r 2 ≥ 0.97) up to three orders of magnitude was achieved. Linearity was demonstrated by variance homogeneity assessment (p > 0.05) and by applying the Mandel fitting test after calculation of linear curve equations for each targeted peptide; p values greater than 0.01 allowed to confirm that the quadratic model was not significantly better than the linear one. In addition, no significant intercepts confirmed the absence of constant systematic errors at a confidence level of 95 %.

Good results were obtained also for method precision, intra-day precision ranging between 3 and 11 % (n = 9). For inter-day precision RSD% values lower than 19 % (n = 75) were always obtained.

To verify the matrix effect, systematic proportional errors were calculated with a t test by comparing calibration curve slopes obtained for the aqueous tryptic digests and the matrix tryptic digests (p < 0.05). For both types of matrix, significant ion suppression was observed from matrices to standard solutions in terms of a decrease of curve slopes. In the case of biscuits, a reduction in sensitivity from 6 to 33 % with respect to the aqueous tryptic digest was evidenced. On the other hand, in the case of dark chocolate, a stronger suppression effect was observed, with a lack of sensitivity in the range of 16–58 % owing to the increased complexity of the matrix compared to biscuits. Hence, to overcome matrix effect, matrix-matched calibration curves were built up to perform a label-free quantification method. Method recoveries obtained in the case of biscuits were in the 84(±6)–106(±4)% range at the low and high fortification level. Similarly, recoveries from chocolate were found to vary between 98(±5) and 108(±6)% at the 50 and 100 mg nuts/kg matrix spiking level, respectively.

Enhanced performances of the new approach were evaluated by comparing it to the previous one (Bignardi et al. 2010) in terms of quality parameters. For all nuts, values of both LODs and LOQs, lower of at least one order of magnitude, were obtained in biscuits, with the advantage to detect and quantify five allergenic proteins at very low concentration levels in the same chromatographic run. On the basis of these findings, it can be inferred that the introduction of the size-exclusion purification step resulted to be helpful both for allergen detectability and for linearity, since a wider linear range of response was observed.

Sample Analysis

In the final part of the work, an important step was represented by sample analysis. The successfully validated method was applied to evaluate the presence of allergens of the five nuts under investigation in different brands of biscuits and dark chocolate samples. In particular, the applicability of the method was tested following two approaches: the first had the final aim of ascertaining possible discrepancies between allergen content and food allergen labelling in positive samples; the second one had the final goal of investigating presence of hidden nut allergens in samples not including peanuts or tree nuts as ingredients. In all the cases, the identification criteria were retention time and comparison of fragmentation spectra between sample and an aqueous extract. For the first purpose, nut contents were estimated in two biscuit and in two chocolate samples having nuts as declared ingredients. As can be seen from Table 4, the quantitative results are in excellent agreement with the labelled values, proving the reliability and the trueness of the method. Examples of separation of the targeted peptides in a hazelnut-biscuit and a hazelnut-chocolate extract are depicted in Figs. 3 and 4, respectively. For the second purpose, the potentiality of the method for trace multi-allergens detection was proved by analysing five biscuits and five dark chocolates of different brands among commercially available products that do not include peanuts or tree nuts as ingredients. All the biscuit samples resulted to be negative, whereas two of the five dark chocolate samples under investigation resulted positive to hazelnut (Table 5, Fig. 5). Figure 5 reports the chromatographic separation of a chocolate sample showing signals corresponding to both the hazelnut peptides monitored. Figure 5b and c illustrate this result for the product ion mass spectrum of the precursor ion at m/z 577 of the ADIYTEQVGR peptide. A good correspondence of the spectra (14/23 fragment ions) with those recorded by analysing a hazelnut extract allows us to confirm the presence of hazelnut traces in the analysed sample.

Table 4 LC-ESI-MS2 determination of the hazelnut content in the food samples investigated (n = 3)
Full size table
Fig. 3
figure 3

Selected reaction monitoring LC-ESI-MS2 chromatograms of the ten targeted peptides in hazelnut-biscuit (brand 1)

Full size image
Fig. 4
figure 4

Selected reaction monitoring LC-ESI-MS2 chromatograms of the ten targeted peptides in hazelnut-dark chocolate (brand 3)

Full size image
Table 5 LC-ESI-MS2 determination of hidden allergens (%) in the food samples investigated (n = 3)
Full size table
Fig. 5
figure 5

a Selected reaction monitoring LC-ESI-MS2 chromatograms of the ten targeted peptides in a dark chocolate sample not including peanuts or tree nuts in the ingredient label (sample G). b Product ion mass spectrum of the precursor ion at m/z 577 corresponding to the signal at 9.55 min, assigned to the hazelnut ADIYTEQVGR peptide. c BioWorks-matched product ion mass spectrum of the precursor ion at m/z 577, which was attributed by the software to the ADIYTEQVGR peptide (the base peak in the mass spectrum at m/z 567 corresponds to the loss of water from the doubly charged precursor ion)

Full size image

Conclusions

The determination of traces of hidden allergens at trace levels in a complex food samples as dark chocolate requires the application of efficient extraction procedures and accurate analytical methods. This work demonstrated the potentiality and the importance of the size-exclusion-based sample treatment able to efficiently clean-up complex food extracts for the sensitive detection of hidden food allergens. The treatment proved effective in determining a significant improvement both in protein sequence coverage for peptide-based biomarker discovery related to food allergies and in sensitivity of the multi-allergen LC-MS2 method, fully taking advantage of the highly selective SRM analysis. Thanks to the very low limits of detection and quantification reached, the extended linearity range and the rapidity of the analysis, this method could be proposed for strict allergen quality controls of both the final products and the production lines for food manufacturers in order to protect consumer health and safety.