- Email: [email protected]
Analysis of acrylamide in different foodstuffs using liquid chromatography–tandem mass spectrometry and gas chromatography–tandem mass spectrometry
Analytica Chimica Acta 520 (2004) 207–215
Analysis of acrylamide in different foodstuffs using liquid chromatography–tandem mass spectrometry and gas chromatography–tandem mass spectrometry K. Hoenicke∗ , R. Gatermann, W. Harder, L. Hartig Eurofins/Wiertz-Eggert-Jörissen GmbH, Stenzelring 14 b, D-2107 Hamburg, Germany Received 16 December 2003; received in revised form 12 March 2004; accepted 12 March 2004
Abstract Acrylamide levels over a wide range of different food products were analysed using both liquid chromatography–tandem mass spectrometry (HPLC–MS–MS) and gas chromatography–tandem mass spectrometry (GC–MS–MS). Two different sample preparation methods for HPLC–MS–MS analysis were developed and optimised with respect to a high sample throughput on the one hand, and a robust and reliable analysis of difficult matrices on the other hand. The first method is applicable to various foods like potato chips, French fries, cereals, bread, and roasted coffee, allowing the analysis of up to 60 samples per technician and day. The second preparation method is not as simple and fast but enables analysis of difficult matrices like cacao, soluble coffee, molasses, or malt. In addition, this method produces extracts which are also well suited for GC–MS–MS analysis. GC–MS–MS has proven to be a sensitive and selective method offering two transitions for acrylamide even at low levels up to 1 g kg−1 . For the respective methods the repeatability (n = 10), given as coefficient of variation, ranged from 3% (acrylamide content of 550 g kg−1 ) to 12% (acrylamide content of 8 g kg−1 ) depending on the food matrix. The repeatability (n = 3) for different food samples spiked with acrylamide (5–1500 g kg−1 ) ranged from 1 to 20% depending on the spiking level and the food matrix. The limit of quantification (referred to a signal-to-noise ratio of 9:1) was 30 g kg−1 for HPLC–MS–MS and 5 g kg−1 for GC–MS–MS. It could be demonstrated that measurement uncertainties were not only a result of analytical variability but also of inhomogeneity and stability of the acrylamide in food. © 2004 Elsevier B.V. All rights reserved. Keywords: Acrylamide; HPLC–MS–MS; GC–MS–MS; Sample preparation; Complex matrices; Inhomogeneity; Stability
1. Introduction In April 2002 the Swedish National Food Administration reported the finding of alarmingly high levels of acrylamide in heat-treated potato products and other baked goods [1]. Many researchers have confirmed the presence of acrylamide in different processed foods and it was shown that its concentration might reach levels as high as several mg kg−1 depending on the composition and the way of processing [2–6]. Numerous paths of formation have been discussed, predominantly the thermal degradation of the free amino ∗ Corresponding author. Tel.: +49-40-752709-479; fax: +49-40-752709-35. E-mail address: [email protected] (K. Hoenicke).
0003-2670/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2004.03.086
acid asparagine via an Amadori product presupposing the interaction of reducing sugars like fructose and glucose [7–10]. Consequently, the Maillard reaction, responsible for characteristic browning and flavour is considered to be responsible for the elevated levels of acrylamide in baked, fried, or roasted food. Since acrylamide is considered a potential carcinogen, worldwide monitoring of this substance in various food products has started. Many data have been published in several journals and magazines of diverse scientific quality, but only a small number of articles were concerned with the technical aspects of the actual measurements. There are primarily two approaches for the analysis of acrylamide in food, based either on gas chromatography with mass spectrometric detectionc (GC–MS) or liquid chromatography with tandem mass spectrometric detection (HPLC–MS–MS) [11].
208
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
The analyses of acrylamide was first performed by GC–MS following bromination to 2,3-dibromopropionamide [12]. Although this method is very sensitive, the derivatisation is laborious and time consuming. Recently, some methods were developed which omit the time consuming derivatisation step and measure acrylamide directly after extraction and clean-up using either GC–MS [13] or HPLC–MS–MS [2,3]. The HPLC method has the advantage that the transfer of acrylamide into an organic solvent or the removal of water after aqueous extraction is not necessary. An interlaboratory comparison study organised in September 2002 by the Federal Institute for Risk Assessment (BfR, previously BgVV) in Germany revealed that the quality of acrylamide analysis depends on the respective food matrix. Consistent results were found for crisp bread or butter cookies, while the analysis of cacao powder caused problems [14]. In a recent proficiency study, only approximately 50% of the participating laboratories reported satisfactory analysis values. Many did not report data or reported data below the limit of quantification (LOQ), others were outside of the accepted range. Comparable results were obtained in an interlaboratory comparison study, organised by the Institute for Reference Materials and Measurement (IRMM) of the Directorate General Joint Research Centre (DG JRC), for crisp bread with acrylamide levels close to the LOQ [15]. The outcome of both proficiency tests suggest that there is not only a lack of analytical methods for the analysis in complex matrices and at low levels but also a lack of communication between the laboratories identifying the pros and cons of used methods. This paper presents a simple and fast method for the detection of acrylamide applicable to a wide range of food category groups like potato chips, French fries, breakfast cereals, bakery products, confectioneries, nuts, and roasted coffee. The method is suitable for routine analysis and allows for a high sample throughput. Additionally, an alternative method for the reliable quantification of acrylamide in complex matrices like molasses, malt, partially cacao or soluble coffee or in lower concentration (up to 5 g kg−1 ), e.g. baby food, is described.
2.2. Sample preparation Depending on the sample matrix, the sample was dried and ground. Dry samples like potato chips, crisp bread, cereals and cookies were ground by an Alexanderwerk mill, mesh size 2 mm (Alexanderwerk, Horsham, PA, USA). Wet samples like bread were dried at 50 ◦ C before grinding. Meat, French fries and other potato products were reduced in size using a mincer. 2.2.1. Routine extraction method The 2 g of the homogenised sample was weighed into a filter, placed on a Witt’scher pot, equipped with a vacuum pump, and defatted by adding 80 ml iso-hexane (40–80 ml min−1 ). The 200 l of internal standard, d3 -acrylamide (10 g ml−1 ), were added to the defatted sample and after an incubation time of 30 min, 20 ml of water was added. In case of high fat matrices like chocolate, marzipan or peanut butter 20 ml of water and 200 l of internal standard, d3 -acrylamide (10 g ml−1 ), were added followed by liquid/liquid extraction with 40 ml iso-hexane. Acrylamide was extracted in an ultrasonic bath (Sonorex RK 510H, Bandelin Elecronic, Berlin, Germany) at 60 ◦ C for 30 min. The sample was purified by adding 20 ml of acetonitrile and 500 l of Carrez I (potassium hexacyanoferrate, c = 150 g l−1 ) and Carrez II (zinc sulphate, c = 300 g l−1 ), respectively. The sample was subsequently centrifuged at 4500 rpm for 10 min (Hettich EBA 8 S, Hettich Zentrifugen, Tuttlingen, Germany) and the supernatant was filtered through a membrane filter (OPTI-Flow, 0.45 m NYL, 25 mm, Wicom, Heppenheim, Germany).
2. Experimental
2.2.2. Alternative extraction method The 50 ml of water and 200 l of internal standard, d3 -acrylamide (10 g ml−1 ), was added to 2 g of the homogenised sample. For baby food 10 g of homogenised sample and 1 ml of internal standard, d3 -acrylamide (1 g ml−1 ), was used. After extraction of acrylamide in a ultrasonic bath at 60 ◦ C for 30 min, 30 ml of iso-hexane and 5 ml of Carrez I (potassium hexacyanoferrate, c = 150 g l−1 ) and Carrez II (zinc sulphate, c = 300 g l−1 ) was added. The sample was centrifuged at 4500 rpm for 10 min. The aqueous phase was saturated with sodium chloride and extracted twice with 50 ml ethyl acetate. The combined organic phases were concentrated to 1 ml using a Turbo Vap 500 (Zyrmark, Idstein, Germany).
2.1. Chemicals
2.3. HPLC–MS–MS analysis
Acrylamide was supplied by Sigma (Deisenhofen, Germany), and deuterium-labelled [2 H3 ]acrylamide (d3 -acrylamide) by CIL (Cambridge Isotope Labs, Andover, MA, USA). All other solvents and chemicals used were of analytical grade. Stock solution of standards (1 mg ml−1 ) and working standard solutions (10 and 1 g ml−1 ) were prepared in methanol/water (20:80 v/v).
The quantification of acrylamide was performed by HPLC–MS–MS with positive electrospray ionisation using an Applied Biosystem API 2000 triple quadrupole mass spectrometer (Darmstadt, Germany) coupled to an Agilent 1100 HPLC system (Waldbronn, Germany) consisting of a 1100 binary pump, a 1100 thermostated autosampler, and a 1100 thermostated column compartment. The
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
column used was a Merck LiChrospher 100 CN column (250 mm × 4 mm I.D., 5 m) (Darmstadt, Germany) together with a Merck LiChrospher 100 RP-18 guard column (4 mm × 4 mm I.D.). The operating conditions were as follows: mobile phase, 50% acetonitrile in 1% acetic acid isocratic for 5 min, following rinsing with 100% acetonitrile for 5 min and reconditioning with 50% acetonitrile in 1% acetic acid for 10 min; flow rate, 0.7 ml min−1 , split 1:5; injection volume, 40 l (aqueous extracts) or 10 l (ethyl acetate extracts); column temperature, 25 ◦ C; autosampler temperature, 20 ◦ C. Acrylamide was identified using multiple reaction monitoring (MRM) set to records m/z 72 > 72, 72 > 55, and 72 > 44, utilising a collision energy of 18 eV. Monitored transitions for the internal standard, d3 -acrylamide, were m/z 75 > 75, 75 > 58, and 75 > 44. The dwell time for each MRM transition was 200 ms. The collisionally activated dissociation (CAD) operated with 4 mTorr pressure of nitrogen as collision gas. The electrospray voltage was set to 5500 V, the source temperature was 350 ◦ C. Quantification was performed by comparison of the peak area ratio of acrylamide with the internal standard, d3 -acrylamide (50 ng ml−1 ), monitored using the MRM transitions at m/z 72 > 55 (acrylamide) and 75 > 58 (d3 -acrylamide). In the case of dried samples (e.g. bread) the loss on drying was included in the acrylamide calculation. 2.4. GC–MS–MS analysis Alternative quantification was performed on an Agilent 6890N gas chromatograph coupled to a Kodiak 800 quadropole tandem mass spectrometer (Chromtech, Idstein, Germany). The extracts were injected splitless, injector temperature at 230 ◦ C, onto a DB-WAX capillary column (30 m × 0.25 mm I.D., 0.25 m film thickness) (J & W Scientific, Folsom, CA, USA). The temperature program was as follows: isothermal for 1 min at 70 ◦ C, increased at a rate of 20 ◦ C min−1 to 230 ◦ C, isothermal for 10 min. Detection was performed using positive chemical ionisation with ammonia as reactant gas. Acrylamide was identified by single reaction monitoring (SRM) set to records at m/z 89 > 72 and 89 > 55 for acrylamide and 92 > 75 for d3 -acrylamide. Acrylamide concentrations were determined using the internal standard, d3 -acrylamide, and applying a calibrated response factor.
209
3. Results and discussion 3.1. Routine analyses of acrylamide 3.1.1. Extraction of acrylamide The greatest differences between the analytical methods commonly used were found for acrylamide extraction and clean-up [11]. Most commonly water extraction is used because acrylamide is highly soluble in water and less soluble in organic solvents. Extraction was performed after defatting because swelling properties were improved when fat was removed. Variations in the extraction time (15, 30 and 60 min) did not affect the amount of acrylamide determined in a potato chips sample using an ultrasonic bath at 60 ◦ C. In all cases 1000 g kg−1 acrylamide were analysed independent of the extraction time. However, to ensure effective extraction even for very complex matrices this step was performed for 30 min. The amount of added water (10 times the sample weight) was adjusted such that a homogeneous suspension was obtained even for strong swelling samples like potato chips. The use of pressurised liquid extraction (PLE; Dionex trade name ASE for accelerated solvent extraction) as a simple and fast method for the extraction of acrylamide from food was described by Höfler et al. [16]. Since this method seems to be most practicable for routine analysis, experiments were carried out to check its applicability. For extraction a Dionex ASE 200 (Darmstadt, Germany) system was used. Extraction was performed according to Höfler et al. using both water and 1% formic acid as extraction solvent. However, in both cases a blockage of the cell followed by a termination of the extraction process was observed. This problem was solved after the use of extraction thimbles but the extracts remain opaque. Extraction with ethyl acetate at 80 ◦ C and 7 mPa pressure yielded clear extracts but the extraction process was neither reproducible nor applicable to
2.5. Recovery tests and blank values Recovery tests for controlling the analytical method and quantification were repeatedly performed by quantification of acrylamide in different foodstuffs before and after the addition of acrylamide (1 mg kg−1 ). A blank value is determined by performing the whole analysis without addition of sample.
Fig. 1. Acrylamide levels of a raw sugar sample analysed after routine extraction and PLE (ASE) at different temperatures.
210
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
all kinds of matrices. A significant lower yield of extraction was found for e.g. cacao and milk powder. On the other hand, extraction of raw sugar yielded higher amounts of acrylamide compared to extraction with water (routine extraction) (Fig. 1). It is not clear why higher amounts were found in these cases. Since the concentration of acrylamide analysed in the sugar extract increases proportional to the extraction temperature (50–80 ◦ C) it was assumed that the higher amounts might be attributable to a formation of acrylamide during PLE. However, PLE at 80 ◦ C of raw mashed potatoes (which have a high potential of acrylamide formation at temperatures above 120 ◦ C) did not result in the formation of acrylamide. Pedersen and Olsson developed an improved extraction method for acrylamide in potato chips [17]. Using Soxhlet extraction with methanol lasting for 10 days, they found nearly seven times higher acrylamide levels than by conventional extraction methods with water. In order to re-evaluate their method, raw, non-heated potato slices were extracted using Soxhlet extraction. After 7 days 18,500 g kg−1 acrylamide were analysed in the methanol extract. It was shown that acrylamide was formed in the methanol extract as
an artefact after extraction of its precursors, asparagine and reducing sugars [18]. It was therefore concluded that Soxhlet extraction is not suitable for extraction of acrylamide in food. In conclusion, water extraction proved to be most reliable for various matrices. A sufficient extraction was achieved and a formation of acrylamide as shown for Soxhlet extraction could be excluded. 3.1.2. Clean-up The main goal of our research was to develop a simple and fast method for routine analyses of acrylamide at concentrations above 30 g kg−1 . Consequently, the clean-up procedure should be as easy and robust as possible and applicable to different types of matrices in order to achieve a high sample throughput. Most HPLC–MS–MS methods published so far carry out a clean-up step that consists of a solid phase extraction (SPE) [2,3,5]. Our experiments have shown that no significant positive effects were observed for matrices like potato chips, French fries, bakery ware, breakfast cereals, or roasted coffee using different SPE columns reported in the literature (data not shown). It was therefore decided to avoid lengthy purification steps since a simple
Fig. 2. MRM chromatograms of a potato chips sample obtained after routine extraction and HPLC–MS–MS analysis (concentration of acrylamide in the sample was 1049 g kg−1 ).
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
211
clean-up consisting of Carrez precipitation turned out to be sufficient. Using this method, up to 60 samples could be prepared per technician and day. Fig. 2 shows the typical MRM chromatograms obtained after routine sample extraction of a potato chips sample with acrylamide concentrations of about 1000 g kg−1 . 3.2. Acrylamide analysis in complex matrices For some matrices the simple clean-up step with Carrez applied in the routine method proved insufficient. As example, some molasses, namely those derived from beet sugar showed a peak with a retention time similar to acrylamide on both the m/z 72 > 72 and 72 > 55 transition which mimics acrylamide (Fig. 3). However, after application of an additional clean-up step which consists of a re-extraction with ethyl acetate, interfering substances were removed and acrylamide concentrations below 30 g kg−1 were analysed. This was not the case for cane molasses. In contrast to beet molasses, cane molasses contain significant amounts of acrylamide. Consequently, the presence of interference must be taken into account when interpreting the chromatographic data after application of the routine method.
Fig. 4. HPLC–MS–MS analysis of acrylamide in soluble coffee (concentration of acrylamide was 630 g kg−1 ). MRM chromatograms of the transitions at m/z 72 > 55 (acrylamide) and 75 > 58 (d3 -acrylamide) obtained after (A) routine extraction and (B) alternative extraction.
The extended extraction procedure was also found to solve background problems in some complex matrices like malt products, aroma extracts, soluble coffee, or cacao. Fig. 4 shows the MRM chromatograms of soluble coffee obtained after application of the routine and alternative extraction, respectively. Regarding the chromatograms received after routine extraction interferences by coeluting substances were detected both on the m/z 72 > 55 (acrylamide) and 75 > 58 (d3 -acrylamide) transition, which were used for quantification. However, applying the alternative extraction method chromatograms without interference were obtained which allow reliable acrylamide quantification. Since the alternative method is approximately three times more time consuming it is recommended using this method only in the case of expected interference or intense analytical noise. 3.3. Acrylamide analysis at low levels
Fig. 3. HPLC–MS–MS analysis of acrylamide in beet and cane molasses. MRM chromatograms of the transitions at m/z 72 > 55 (acrylamide) and 75 > 58 (d3 -acrylamide) obtained (A) after routine extraction of beet and cane molasses and (B) after alternative extraction of beet molasses.
The extended extraction procedure does not only solve background problems in some matrices but offers a concentration step by re-extraction of acrylamide in ethyl acetate and evaporation to 1 ml. Additionally, the second extraction step has the advantage that the amount of water used in the first extraction step does not affect the end volume of the prepared extract, e.g. the LOQ. The sample quantity used for routine extraction is limited to 2 g. An higher amount requires an higher volume of water for extraction because of the swelling properties of most samples. Therefore, the LOQ
212
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
well suited for GC–MS–MS analysis. GC–MS–MS with positive chemical ionisation and ammonia as reactant gas was proven to be sensitive and selective by offering two transitions for acrylamide (m/z 89 > 72 and 89 > 55) even at this low level (Fig. 5B). It is therefore well suited for a reliable quantification of acrylamide at levels up to 5 g kg−1 . However, for quantification of acrylamide at low levels care must be taken to avoid acrylamide contamination. Verification of solvent purity is recommended. In order to eliminate carry-over heating of all used glass equipment for 16 h is essential. Nevertheless, a blind value of 1 g kg−1 was often detected and could not be excluded. 3.4. Detection limit, recovery, and linearity
Fig. 5. MRM chromatograms of a baby food sample (acrylamide content of 10 g kg−1 ) obtained after alternative extraction using (A) HPLC–MS–MS analysis (monitored transitions at m/z 72 > 55 for acrylamide and 75 > 58 for d3 -acrylamide) and (B) GC–MS–MS analysis (monitored transitions at m/z 89 > 55 and 89 > 72 for acrylamide and 92 > 75 for d3 -acrylamide).
of the routine method could be not decreased by increasing the weight of the sample. However, using the alternative extraction method 50 ml of water were used. This allows to increase the sample weight and to further increase the concentration level. Consequently, a lower LOQ is achievable. Using 10 g of sample concentrations up to 10 g kg−1 were easily detected in the MRM mode using the transition at m/z 72 > 55. Fig. 5A shows typical MRM chromatograms of a baby food sample with acrylamide concentrations of 10 g kg−1 . Nevertheless, HPLC–MS–MS offers only one characteristic transition at this low level. Since ethyl acetate is used as solvent for re-extraction the extended extraction method offers extracts which are also
Regarding the routine extraction method estimations of the LOQ (signal-to-noise ratio of 9:1) and the limit of detection (LOD) (signal-to-noise ratio of 3:1) for the different matrices did not exceed 30 and 10 g kg−1 , respectively. Recoveries were determined by the addition of 100, 200, 500, 1000, and 1500 g kg−1 of acrylamide to very light coloured potato chips (acrylamide < 30 g kg−1 ) and cookies (acrylamide = 60 g kg−1 ) followed by triplicate analysis. Furthermore, wheat flour (acrylamide < 30 g kg−1 ) was spiked with 30, 50, and 100 g kg−1 of acrylamide in order to validate the method at the lower concentration level. For potato chips the recoveries averaged 90% at the lower concentration range and 116% at the higher concentration. For cookies and flour recoveries averaged 87–97 and 94–100% of the acrylamide added. The results are summarised in Table 1. Daily performed recovery tests (addition of 1 mg kg−1 acrylamide) for different food matrices showed a recovery between 80 and 110% independent of the kind of food matrix. Coefficients of variation (CVs) for selected food categories (potato chips, crisp bread, cookies, roasted coffee) were determined by 10-fold analysis of the respective food and were found to be between 3 and 10% depending on the food matrix. The results are summarised in Table 2. Regarding the alternative extraction method only 20% of acrylamide and the internal standard, d3 -acrylamide, were transferred by re-extraction from the aqueous into the ethyl acetate phase, respectively. Due to the four times lower injection volume (a higher injection volume resulted in a peak broadening), no significantly lower LOQ was achieved in spite of the concentration step. Recoveries were determined by the addition of 100, 200, and 500 g kg−1 of acrylamide to cacao powder (acrylamide = 190 g kg−1 ) followed by triplicate analysis. The recoveries averaged between 84 and 91% of the acrylamide added (Table 1). The CVs determined by 10-fold analysis of a cacao powder and a soluble coffee sample were found to be 9 and 4%, respectively (Table 2). Using GC–MS–MS analysis an LOQ (signal-to-noise ratio of 9:1) of 5 g kg−1 and an LOD (signal-to-noise ratio of 3:1) of 1.5 g kg−1 was achieved after fivefold increasing the sample weight. Recoveries were determined by the addition
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
213
Table 1 Recovery (%) and coefficient of variation (CV) (%) of acrylamide for the different methods analysed in various food matrices at different spiking levels (n = 3) Spiking level
5 20 30 50 100 200 500 1000 1500 a b c
Potato chipsa
Cookiesa
Floura
Cacaob
Baby foodc
Recovery (%)
CV (%)
Recovery (%)
CV (%)
Recovery (%)
CV (%)
Recovery (%)
CV (%)
Recovery (%)
CV (%)
– – – – 87 90 90 93 116
– – – – 4 4 1 3 1
– – – – 87 91 97 96 94
– – – – 5 5 5 6 8
– – 94 94 100 – – – –
– – 5 4 4 – – – –
– – – – 84 87 91 – –
– – – – 9 7 5 – –
88 90 87 84 86 – – – –
20 19 18 12 9 – – – –
Routine extraction/HPLC–MS–MS. Alternative extraction/HPLC–MS–MS. Alternative extraction/GC–MS–MS.
of 5, 20, 30, 50, and 100 g kg−1 of acrylamide to a vegetable stew (baby food) (acrylamide < 5 g kg−1 ) followed by triplicate analysis. The recoveries averaged between 84 and 90% of the acrylamide added (Table 1). The CV determined by 10-fold analysis of skim-milk powder (baby food) was found to be 12% (Table 1). Altogether, more than 20 different product categories, including several different matrices were screened. Fig. 6A summarises the different food categories and the number of samples analysed using the different methods, respectively. As can be seen, the routine method was applicable to 90% of the analysed samples. The extended alternative extraction was essential for only 6% of the tested samples. GC–MS–MS analysis was mainly applied to baby food. As expected potato products and bakery ware presented the most frequently analysed food. However, the figure reflects the complexity of the acrylamide problem since acrylamide can be found in all kind of heat-treated foods and therefore more or less all types of food products have to be analysed. The mean values of the acrylamide concentration analysed in the respective food categories are given in Fig. 6B. Comparison of the number of analysed samples and the mean value of the respective food category reveals that in
some cases the low number of analysed samples did not correlate with an expected low acrylamide concentration. Product groups with remarkable acrylamide values are for example some instant products, roasted vegetables (e.g. onion, chicory) and special herbal teas. This reflects a possible underestimation of some product groups that can
Table 2 Mean, coefficient of variation (CV), and limit of quantification (LOQ) of acrylamide (n = 10) for different food matrices determined using the respective extraction method and HPLC–MS–MS or GC–MS–MS analysis Matrix
Mean (g kg−1 )
CV (%)
LOQ (g kg−1 )
Potato chipsa Crisp breada Butter biscuita Cacaob Coffee, roasteda Coffee, solubleb Baby foodc
620 439 546 190 282 816 8
9.6 3.6 2.7 9.1 9.2 4.1 11.7
30 30 30 30 30 30 5
a b c
Routine extraction/HPLC–MS–MS. Alternative extraction/HPLC–MS–MS. Alternative extraction/GC–MS–MS.
Fig. 6. (A) Number of analysed samples and distribution of the methods used for analysis. (B) Mean values of acrylamide concentrations analysed in the respective food categories.
214
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
contribute to the daily uptake of acrylamide. However, it has been estimated so far that 38% of the calories of the daily intake come from foods containing acrylamide [19]. 3.5. Causes for variability of acrylamide levels Recurrent analysis of acrylamide in the same food product (same recipe, same manufacturing process) revealed a wide variability of acrylamide levels which was mostly higher than the 10% found for analytical variations. Moreover, acrylamide levels vary widely in different product samples of one batch. It was shown for potato chips that acrylamide concentrations analysed in different packages of one batch might range between plus and minus 50% from the mean. However, it has been assumed that the inhomogeneity of the sample matrix or the stability of acrylamide in the food may contribute to the variability of acrylamide levels, as well.
Fig. 8. Decrease of acrylamide concentrations in a homogenised sample of potato chips over 100 days (n = 28).
3.5.1. Inhomogeneity of samples Potato chips were proven to be not homogeneous since its precursors were not uniformly distributed in the potato material. Analyses of a part of a homogenised sample of potato chips which contains predominantly dark particles resulted in one fourth higher acrylamide levels than analyses of the blend (Fig. 7). This finding shows that acrylamide is not evenly distributed—rather it could be concentrated in special particles of the food matrix. 3.5.2. Stability of acrylamide in food Analysis of acrylamide in potato chips over 100 days gave a first hint that acrylamide concentrations are not stable over time (Fig. 8). Since the fine-ground sample was stored in a closed package at room temperature peroxide formation occurred. In the cold extracted fat of the sample a peroxide value of 3.8 mVal O2 kg−1 was determined which was assumed to be the cause of acrylamide degradation. The comparative analysis of beet and cane molasses has shown that beet molasses contains much less acrylamide than cane molasses (approximately 10 times lower). Commonly, this fact was associated with the presence of higher
Fig. 9. Decrease of acrylamide concentrations in a sulphurised 60◦ Brix sugar solution spiked with 1000 g kg−1 acrylamide (䉱) and control sample without addition of sulphite (䊏).
amounts of asparagine in products derived from cane. However, sulphite is used as a preservative agent in beet sugar processing. It was therefore assumed that an oxidative degradation of acrylamide triggered by sulphite could be responsible for the lower acrylamide concentrations determined in beet molasses. In order to simulate the effect of sulphite in sugar processing juices and molasses a 60◦ Brix solution of sugar was added with sodium sulphite (500 mg kg−1 ) and spiked with 1000 g kg−1 acrylamide. For control a 60◦ Brix solution of sugar was spiked with 1000 g kg−1 acrylamide and analysed without addition of sulphite. Acrylamide levels were analysed after 7, 24, 48, and 168 h. As shown in Fig. 9 acrylamide concentrations in the sulphurised sample decrease rapidly and drop down to only 40 g kg−1 after 7 days. On the other hand, acrylamide concentrations remain stable over 168 h without addition of sulphite. These observations gave reasons to suggest that acrylamide concentrations in some foods can vary depending on the storage time since the levels can be affected by special food constituents and/or reaction products. 4. Conclusion
Fig. 7. Analysis of a homogenised sample of potato chips (䉱) and a part of the same blend which contains predominantly dark particles (䊏).
Methods used in routine analysis have to be simple, fast and reliable. The occurrence of acrylamide in various food
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
products makes it very difficult to develop one method which is suitable for all purposes and matrices. Therefore, two different methods were used, one which is very simple, fast and applicable to most samples and a second one which gives reliable results even for difficult matrices like soluble coffee or cacao. For the analysis of acrylamide at low levels GC–MS–MS turned out to be a suitable method since levels up to 5 g kg−1 can be quantified by two mass transitions without derivatisation. However, measurement uncertainties are not only caused by analytical variations but rather by inhomogeneity of the sample matrix or by the instability of acrylamide in the food during storage. The variability of acrylamide levels in food depends on the nature, the manufacturing process, and probably on the storage conditions of the respective food. Altogether, this can cause variations of up to 50% which exceed the analytical variability by nearly a factor of 5. Acknowledgements We like to thank Wolfgang Meins and Marco Gangnus for acrylamide analysis and Bert Pöpping for proof-reading and editing this document. References [1] Swedish National Food Administration, Information about Acrylamide in Food. http://www.slv.se/engdefault.asp.
215
[2] J. Rosén, K.E. Hellenäs, Analyst 127 (2002) 880–882. [3] E. Tareke, P. Rydberg, P. Karlsson, S. Eriksson, M. Törnquist, J. Agric. Food Chem. 50 (2002) 498–506. [4] J.S. Ahn, L. Castle, D.B. Clark, A.S. Loyd, M.R. Philo, D.R. Speck, Food Addit. Contam. 19 (2002) 1116–1124. [5] B. Gutsche, R. Weißhaar, J. Buhlert, Deut. Lebensmi.-Rundsch. 98 (2002) 437–443. [6] A. Becalski, B.P.Y. Lau, D. Lewis, S.W. Seaman, J. Agric. Food Chem. 51 (2003) 802–808. [7] D.S. Mottram, B. Wedzicha, A.T. Dodson, Nature 419 (2002) 448. [8] R.H. Stadler, I. Blank, N. Varga, F. Robert, J. Hau, P.A. Guy, M.C. Robert, S. Riediker, Nature 419 (2002) 449. [9] R. Weißhaar, B. Gutsche, Deut. Lebensm.-Rundsch. 98 (2002) 397– 399. [10] D.V. Zyzak, R.A. Sanders, M. Stojanovic, D.H. Tallmadge, B.L. Eberhart, D.K. Ewald, D.C. Gruber, T.R. Morsch, M.A. Strothers, G.P. Rizzi, M.D. Villagran, J. Agric. Food Chem. 51 (2003) 4782– 4787. [11] T. Wenzl, B. de la Calle, E. Anklam, Food Addit. Contam. 20 (2003) 885–902. [12] L. Castle, J. Agric. Food Chem. 41 (1993) 1261–1263. [13] M. Biedermann, S. Biedermann-Brem, A. Noti, K. Grob, O. Egli, H. Mändli, Mitt. Lebensm. Hyg. 93 (2002) 638–652. [14] C. Fauhl, H. Klaffke, W. Mathar, R. Palvinskas, R. Wittkowski. http:// www.bfr.bund.de/cmc/detail.php?template=internet de index js. [15] T. Wenzl, B. de la Calle, R. Gatermann, K. Hoenicke, F. Ulberth, E. Anklam, Anal. Bioanal. Chem., in press. [16] F. Höfler, R. Maurer, S. Cavalli, GIT 9 (2002) 619–621. [17] J.R. Pedersen, J.O. Olsson, Analyst 128 (2003) 332–334. [18] K. Grob, M. Biedermann, K. Hoenicke, R. Gatermann, Analyst 129 (2004) 92. [19] Milling and Baking News, November 25, 2003. http://www.worldgrain.com/feature stories.asp?ArticleID=68697.
Analysis of acrylamide in different foodstuffs using liquid chromatography–tandem mass spectrometry and gas chromatography–tandem mass spectrometry K. Hoenicke∗ , R. Gatermann, W. Harder, L. Hartig Eurofins/Wiertz-Eggert-Jörissen GmbH, Stenzelring 14 b, D-2107 Hamburg, Germany Received 16 December 2003; received in revised form 12 March 2004; accepted 12 March 2004
Abstract Acrylamide levels over a wide range of different food products were analysed using both liquid chromatography–tandem mass spectrometry (HPLC–MS–MS) and gas chromatography–tandem mass spectrometry (GC–MS–MS). Two different sample preparation methods for HPLC–MS–MS analysis were developed and optimised with respect to a high sample throughput on the one hand, and a robust and reliable analysis of difficult matrices on the other hand. The first method is applicable to various foods like potato chips, French fries, cereals, bread, and roasted coffee, allowing the analysis of up to 60 samples per technician and day. The second preparation method is not as simple and fast but enables analysis of difficult matrices like cacao, soluble coffee, molasses, or malt. In addition, this method produces extracts which are also well suited for GC–MS–MS analysis. GC–MS–MS has proven to be a sensitive and selective method offering two transitions for acrylamide even at low levels up to 1 g kg−1 . For the respective methods the repeatability (n = 10), given as coefficient of variation, ranged from 3% (acrylamide content of 550 g kg−1 ) to 12% (acrylamide content of 8 g kg−1 ) depending on the food matrix. The repeatability (n = 3) for different food samples spiked with acrylamide (5–1500 g kg−1 ) ranged from 1 to 20% depending on the spiking level and the food matrix. The limit of quantification (referred to a signal-to-noise ratio of 9:1) was 30 g kg−1 for HPLC–MS–MS and 5 g kg−1 for GC–MS–MS. It could be demonstrated that measurement uncertainties were not only a result of analytical variability but also of inhomogeneity and stability of the acrylamide in food. © 2004 Elsevier B.V. All rights reserved. Keywords: Acrylamide; HPLC–MS–MS; GC–MS–MS; Sample preparation; Complex matrices; Inhomogeneity; Stability
1. Introduction In April 2002 the Swedish National Food Administration reported the finding of alarmingly high levels of acrylamide in heat-treated potato products and other baked goods [1]. Many researchers have confirmed the presence of acrylamide in different processed foods and it was shown that its concentration might reach levels as high as several mg kg−1 depending on the composition and the way of processing [2–6]. Numerous paths of formation have been discussed, predominantly the thermal degradation of the free amino ∗ Corresponding author. Tel.: +49-40-752709-479; fax: +49-40-752709-35. E-mail address: [email protected] (K. Hoenicke).
0003-2670/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2004.03.086
acid asparagine via an Amadori product presupposing the interaction of reducing sugars like fructose and glucose [7–10]. Consequently, the Maillard reaction, responsible for characteristic browning and flavour is considered to be responsible for the elevated levels of acrylamide in baked, fried, or roasted food. Since acrylamide is considered a potential carcinogen, worldwide monitoring of this substance in various food products has started. Many data have been published in several journals and magazines of diverse scientific quality, but only a small number of articles were concerned with the technical aspects of the actual measurements. There are primarily two approaches for the analysis of acrylamide in food, based either on gas chromatography with mass spectrometric detectionc (GC–MS) or liquid chromatography with tandem mass spectrometric detection (HPLC–MS–MS) [11].
208
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
The analyses of acrylamide was first performed by GC–MS following bromination to 2,3-dibromopropionamide [12]. Although this method is very sensitive, the derivatisation is laborious and time consuming. Recently, some methods were developed which omit the time consuming derivatisation step and measure acrylamide directly after extraction and clean-up using either GC–MS [13] or HPLC–MS–MS [2,3]. The HPLC method has the advantage that the transfer of acrylamide into an organic solvent or the removal of water after aqueous extraction is not necessary. An interlaboratory comparison study organised in September 2002 by the Federal Institute for Risk Assessment (BfR, previously BgVV) in Germany revealed that the quality of acrylamide analysis depends on the respective food matrix. Consistent results were found for crisp bread or butter cookies, while the analysis of cacao powder caused problems [14]. In a recent proficiency study, only approximately 50% of the participating laboratories reported satisfactory analysis values. Many did not report data or reported data below the limit of quantification (LOQ), others were outside of the accepted range. Comparable results were obtained in an interlaboratory comparison study, organised by the Institute for Reference Materials and Measurement (IRMM) of the Directorate General Joint Research Centre (DG JRC), for crisp bread with acrylamide levels close to the LOQ [15]. The outcome of both proficiency tests suggest that there is not only a lack of analytical methods for the analysis in complex matrices and at low levels but also a lack of communication between the laboratories identifying the pros and cons of used methods. This paper presents a simple and fast method for the detection of acrylamide applicable to a wide range of food category groups like potato chips, French fries, breakfast cereals, bakery products, confectioneries, nuts, and roasted coffee. The method is suitable for routine analysis and allows for a high sample throughput. Additionally, an alternative method for the reliable quantification of acrylamide in complex matrices like molasses, malt, partially cacao or soluble coffee or in lower concentration (up to 5 g kg−1 ), e.g. baby food, is described.
2.2. Sample preparation Depending on the sample matrix, the sample was dried and ground. Dry samples like potato chips, crisp bread, cereals and cookies were ground by an Alexanderwerk mill, mesh size 2 mm (Alexanderwerk, Horsham, PA, USA). Wet samples like bread were dried at 50 ◦ C before grinding. Meat, French fries and other potato products were reduced in size using a mincer. 2.2.1. Routine extraction method The 2 g of the homogenised sample was weighed into a filter, placed on a Witt’scher pot, equipped with a vacuum pump, and defatted by adding 80 ml iso-hexane (40–80 ml min−1 ). The 200 l of internal standard, d3 -acrylamide (10 g ml−1 ), were added to the defatted sample and after an incubation time of 30 min, 20 ml of water was added. In case of high fat matrices like chocolate, marzipan or peanut butter 20 ml of water and 200 l of internal standard, d3 -acrylamide (10 g ml−1 ), were added followed by liquid/liquid extraction with 40 ml iso-hexane. Acrylamide was extracted in an ultrasonic bath (Sonorex RK 510H, Bandelin Elecronic, Berlin, Germany) at 60 ◦ C for 30 min. The sample was purified by adding 20 ml of acetonitrile and 500 l of Carrez I (potassium hexacyanoferrate, c = 150 g l−1 ) and Carrez II (zinc sulphate, c = 300 g l−1 ), respectively. The sample was subsequently centrifuged at 4500 rpm for 10 min (Hettich EBA 8 S, Hettich Zentrifugen, Tuttlingen, Germany) and the supernatant was filtered through a membrane filter (OPTI-Flow, 0.45 m NYL, 25 mm, Wicom, Heppenheim, Germany).
2. Experimental
2.2.2. Alternative extraction method The 50 ml of water and 200 l of internal standard, d3 -acrylamide (10 g ml−1 ), was added to 2 g of the homogenised sample. For baby food 10 g of homogenised sample and 1 ml of internal standard, d3 -acrylamide (1 g ml−1 ), was used. After extraction of acrylamide in a ultrasonic bath at 60 ◦ C for 30 min, 30 ml of iso-hexane and 5 ml of Carrez I (potassium hexacyanoferrate, c = 150 g l−1 ) and Carrez II (zinc sulphate, c = 300 g l−1 ) was added. The sample was centrifuged at 4500 rpm for 10 min. The aqueous phase was saturated with sodium chloride and extracted twice with 50 ml ethyl acetate. The combined organic phases were concentrated to 1 ml using a Turbo Vap 500 (Zyrmark, Idstein, Germany).
2.1. Chemicals
2.3. HPLC–MS–MS analysis
Acrylamide was supplied by Sigma (Deisenhofen, Germany), and deuterium-labelled [2 H3 ]acrylamide (d3 -acrylamide) by CIL (Cambridge Isotope Labs, Andover, MA, USA). All other solvents and chemicals used were of analytical grade. Stock solution of standards (1 mg ml−1 ) and working standard solutions (10 and 1 g ml−1 ) were prepared in methanol/water (20:80 v/v).
The quantification of acrylamide was performed by HPLC–MS–MS with positive electrospray ionisation using an Applied Biosystem API 2000 triple quadrupole mass spectrometer (Darmstadt, Germany) coupled to an Agilent 1100 HPLC system (Waldbronn, Germany) consisting of a 1100 binary pump, a 1100 thermostated autosampler, and a 1100 thermostated column compartment. The
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
column used was a Merck LiChrospher 100 CN column (250 mm × 4 mm I.D., 5 m) (Darmstadt, Germany) together with a Merck LiChrospher 100 RP-18 guard column (4 mm × 4 mm I.D.). The operating conditions were as follows: mobile phase, 50% acetonitrile in 1% acetic acid isocratic for 5 min, following rinsing with 100% acetonitrile for 5 min and reconditioning with 50% acetonitrile in 1% acetic acid for 10 min; flow rate, 0.7 ml min−1 , split 1:5; injection volume, 40 l (aqueous extracts) or 10 l (ethyl acetate extracts); column temperature, 25 ◦ C; autosampler temperature, 20 ◦ C. Acrylamide was identified using multiple reaction monitoring (MRM) set to records m/z 72 > 72, 72 > 55, and 72 > 44, utilising a collision energy of 18 eV. Monitored transitions for the internal standard, d3 -acrylamide, were m/z 75 > 75, 75 > 58, and 75 > 44. The dwell time for each MRM transition was 200 ms. The collisionally activated dissociation (CAD) operated with 4 mTorr pressure of nitrogen as collision gas. The electrospray voltage was set to 5500 V, the source temperature was 350 ◦ C. Quantification was performed by comparison of the peak area ratio of acrylamide with the internal standard, d3 -acrylamide (50 ng ml−1 ), monitored using the MRM transitions at m/z 72 > 55 (acrylamide) and 75 > 58 (d3 -acrylamide). In the case of dried samples (e.g. bread) the loss on drying was included in the acrylamide calculation. 2.4. GC–MS–MS analysis Alternative quantification was performed on an Agilent 6890N gas chromatograph coupled to a Kodiak 800 quadropole tandem mass spectrometer (Chromtech, Idstein, Germany). The extracts were injected splitless, injector temperature at 230 ◦ C, onto a DB-WAX capillary column (30 m × 0.25 mm I.D., 0.25 m film thickness) (J & W Scientific, Folsom, CA, USA). The temperature program was as follows: isothermal for 1 min at 70 ◦ C, increased at a rate of 20 ◦ C min−1 to 230 ◦ C, isothermal for 10 min. Detection was performed using positive chemical ionisation with ammonia as reactant gas. Acrylamide was identified by single reaction monitoring (SRM) set to records at m/z 89 > 72 and 89 > 55 for acrylamide and 92 > 75 for d3 -acrylamide. Acrylamide concentrations were determined using the internal standard, d3 -acrylamide, and applying a calibrated response factor.
209
3. Results and discussion 3.1. Routine analyses of acrylamide 3.1.1. Extraction of acrylamide The greatest differences between the analytical methods commonly used were found for acrylamide extraction and clean-up [11]. Most commonly water extraction is used because acrylamide is highly soluble in water and less soluble in organic solvents. Extraction was performed after defatting because swelling properties were improved when fat was removed. Variations in the extraction time (15, 30 and 60 min) did not affect the amount of acrylamide determined in a potato chips sample using an ultrasonic bath at 60 ◦ C. In all cases 1000 g kg−1 acrylamide were analysed independent of the extraction time. However, to ensure effective extraction even for very complex matrices this step was performed for 30 min. The amount of added water (10 times the sample weight) was adjusted such that a homogeneous suspension was obtained even for strong swelling samples like potato chips. The use of pressurised liquid extraction (PLE; Dionex trade name ASE for accelerated solvent extraction) as a simple and fast method for the extraction of acrylamide from food was described by Höfler et al. [16]. Since this method seems to be most practicable for routine analysis, experiments were carried out to check its applicability. For extraction a Dionex ASE 200 (Darmstadt, Germany) system was used. Extraction was performed according to Höfler et al. using both water and 1% formic acid as extraction solvent. However, in both cases a blockage of the cell followed by a termination of the extraction process was observed. This problem was solved after the use of extraction thimbles but the extracts remain opaque. Extraction with ethyl acetate at 80 ◦ C and 7 mPa pressure yielded clear extracts but the extraction process was neither reproducible nor applicable to
2.5. Recovery tests and blank values Recovery tests for controlling the analytical method and quantification were repeatedly performed by quantification of acrylamide in different foodstuffs before and after the addition of acrylamide (1 mg kg−1 ). A blank value is determined by performing the whole analysis without addition of sample.
Fig. 1. Acrylamide levels of a raw sugar sample analysed after routine extraction and PLE (ASE) at different temperatures.
210
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
all kinds of matrices. A significant lower yield of extraction was found for e.g. cacao and milk powder. On the other hand, extraction of raw sugar yielded higher amounts of acrylamide compared to extraction with water (routine extraction) (Fig. 1). It is not clear why higher amounts were found in these cases. Since the concentration of acrylamide analysed in the sugar extract increases proportional to the extraction temperature (50–80 ◦ C) it was assumed that the higher amounts might be attributable to a formation of acrylamide during PLE. However, PLE at 80 ◦ C of raw mashed potatoes (which have a high potential of acrylamide formation at temperatures above 120 ◦ C) did not result in the formation of acrylamide. Pedersen and Olsson developed an improved extraction method for acrylamide in potato chips [17]. Using Soxhlet extraction with methanol lasting for 10 days, they found nearly seven times higher acrylamide levels than by conventional extraction methods with water. In order to re-evaluate their method, raw, non-heated potato slices were extracted using Soxhlet extraction. After 7 days 18,500 g kg−1 acrylamide were analysed in the methanol extract. It was shown that acrylamide was formed in the methanol extract as
an artefact after extraction of its precursors, asparagine and reducing sugars [18]. It was therefore concluded that Soxhlet extraction is not suitable for extraction of acrylamide in food. In conclusion, water extraction proved to be most reliable for various matrices. A sufficient extraction was achieved and a formation of acrylamide as shown for Soxhlet extraction could be excluded. 3.1.2. Clean-up The main goal of our research was to develop a simple and fast method for routine analyses of acrylamide at concentrations above 30 g kg−1 . Consequently, the clean-up procedure should be as easy and robust as possible and applicable to different types of matrices in order to achieve a high sample throughput. Most HPLC–MS–MS methods published so far carry out a clean-up step that consists of a solid phase extraction (SPE) [2,3,5]. Our experiments have shown that no significant positive effects were observed for matrices like potato chips, French fries, bakery ware, breakfast cereals, or roasted coffee using different SPE columns reported in the literature (data not shown). It was therefore decided to avoid lengthy purification steps since a simple
Fig. 2. MRM chromatograms of a potato chips sample obtained after routine extraction and HPLC–MS–MS analysis (concentration of acrylamide in the sample was 1049 g kg−1 ).
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
211
clean-up consisting of Carrez precipitation turned out to be sufficient. Using this method, up to 60 samples could be prepared per technician and day. Fig. 2 shows the typical MRM chromatograms obtained after routine sample extraction of a potato chips sample with acrylamide concentrations of about 1000 g kg−1 . 3.2. Acrylamide analysis in complex matrices For some matrices the simple clean-up step with Carrez applied in the routine method proved insufficient. As example, some molasses, namely those derived from beet sugar showed a peak with a retention time similar to acrylamide on both the m/z 72 > 72 and 72 > 55 transition which mimics acrylamide (Fig. 3). However, after application of an additional clean-up step which consists of a re-extraction with ethyl acetate, interfering substances were removed and acrylamide concentrations below 30 g kg−1 were analysed. This was not the case for cane molasses. In contrast to beet molasses, cane molasses contain significant amounts of acrylamide. Consequently, the presence of interference must be taken into account when interpreting the chromatographic data after application of the routine method.
Fig. 4. HPLC–MS–MS analysis of acrylamide in soluble coffee (concentration of acrylamide was 630 g kg−1 ). MRM chromatograms of the transitions at m/z 72 > 55 (acrylamide) and 75 > 58 (d3 -acrylamide) obtained after (A) routine extraction and (B) alternative extraction.
The extended extraction procedure was also found to solve background problems in some complex matrices like malt products, aroma extracts, soluble coffee, or cacao. Fig. 4 shows the MRM chromatograms of soluble coffee obtained after application of the routine and alternative extraction, respectively. Regarding the chromatograms received after routine extraction interferences by coeluting substances were detected both on the m/z 72 > 55 (acrylamide) and 75 > 58 (d3 -acrylamide) transition, which were used for quantification. However, applying the alternative extraction method chromatograms without interference were obtained which allow reliable acrylamide quantification. Since the alternative method is approximately three times more time consuming it is recommended using this method only in the case of expected interference or intense analytical noise. 3.3. Acrylamide analysis at low levels
Fig. 3. HPLC–MS–MS analysis of acrylamide in beet and cane molasses. MRM chromatograms of the transitions at m/z 72 > 55 (acrylamide) and 75 > 58 (d3 -acrylamide) obtained (A) after routine extraction of beet and cane molasses and (B) after alternative extraction of beet molasses.
The extended extraction procedure does not only solve background problems in some matrices but offers a concentration step by re-extraction of acrylamide in ethyl acetate and evaporation to 1 ml. Additionally, the second extraction step has the advantage that the amount of water used in the first extraction step does not affect the end volume of the prepared extract, e.g. the LOQ. The sample quantity used for routine extraction is limited to 2 g. An higher amount requires an higher volume of water for extraction because of the swelling properties of most samples. Therefore, the LOQ
212
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
well suited for GC–MS–MS analysis. GC–MS–MS with positive chemical ionisation and ammonia as reactant gas was proven to be sensitive and selective by offering two transitions for acrylamide (m/z 89 > 72 and 89 > 55) even at this low level (Fig. 5B). It is therefore well suited for a reliable quantification of acrylamide at levels up to 5 g kg−1 . However, for quantification of acrylamide at low levels care must be taken to avoid acrylamide contamination. Verification of solvent purity is recommended. In order to eliminate carry-over heating of all used glass equipment for 16 h is essential. Nevertheless, a blind value of 1 g kg−1 was often detected and could not be excluded. 3.4. Detection limit, recovery, and linearity
Fig. 5. MRM chromatograms of a baby food sample (acrylamide content of 10 g kg−1 ) obtained after alternative extraction using (A) HPLC–MS–MS analysis (monitored transitions at m/z 72 > 55 for acrylamide and 75 > 58 for d3 -acrylamide) and (B) GC–MS–MS analysis (monitored transitions at m/z 89 > 55 and 89 > 72 for acrylamide and 92 > 75 for d3 -acrylamide).
of the routine method could be not decreased by increasing the weight of the sample. However, using the alternative extraction method 50 ml of water were used. This allows to increase the sample weight and to further increase the concentration level. Consequently, a lower LOQ is achievable. Using 10 g of sample concentrations up to 10 g kg−1 were easily detected in the MRM mode using the transition at m/z 72 > 55. Fig. 5A shows typical MRM chromatograms of a baby food sample with acrylamide concentrations of 10 g kg−1 . Nevertheless, HPLC–MS–MS offers only one characteristic transition at this low level. Since ethyl acetate is used as solvent for re-extraction the extended extraction method offers extracts which are also
Regarding the routine extraction method estimations of the LOQ (signal-to-noise ratio of 9:1) and the limit of detection (LOD) (signal-to-noise ratio of 3:1) for the different matrices did not exceed 30 and 10 g kg−1 , respectively. Recoveries were determined by the addition of 100, 200, 500, 1000, and 1500 g kg−1 of acrylamide to very light coloured potato chips (acrylamide < 30 g kg−1 ) and cookies (acrylamide = 60 g kg−1 ) followed by triplicate analysis. Furthermore, wheat flour (acrylamide < 30 g kg−1 ) was spiked with 30, 50, and 100 g kg−1 of acrylamide in order to validate the method at the lower concentration level. For potato chips the recoveries averaged 90% at the lower concentration range and 116% at the higher concentration. For cookies and flour recoveries averaged 87–97 and 94–100% of the acrylamide added. The results are summarised in Table 1. Daily performed recovery tests (addition of 1 mg kg−1 acrylamide) for different food matrices showed a recovery between 80 and 110% independent of the kind of food matrix. Coefficients of variation (CVs) for selected food categories (potato chips, crisp bread, cookies, roasted coffee) were determined by 10-fold analysis of the respective food and were found to be between 3 and 10% depending on the food matrix. The results are summarised in Table 2. Regarding the alternative extraction method only 20% of acrylamide and the internal standard, d3 -acrylamide, were transferred by re-extraction from the aqueous into the ethyl acetate phase, respectively. Due to the four times lower injection volume (a higher injection volume resulted in a peak broadening), no significantly lower LOQ was achieved in spite of the concentration step. Recoveries were determined by the addition of 100, 200, and 500 g kg−1 of acrylamide to cacao powder (acrylamide = 190 g kg−1 ) followed by triplicate analysis. The recoveries averaged between 84 and 91% of the acrylamide added (Table 1). The CVs determined by 10-fold analysis of a cacao powder and a soluble coffee sample were found to be 9 and 4%, respectively (Table 2). Using GC–MS–MS analysis an LOQ (signal-to-noise ratio of 9:1) of 5 g kg−1 and an LOD (signal-to-noise ratio of 3:1) of 1.5 g kg−1 was achieved after fivefold increasing the sample weight. Recoveries were determined by the addition
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
213
Table 1 Recovery (%) and coefficient of variation (CV) (%) of acrylamide for the different methods analysed in various food matrices at different spiking levels (n = 3) Spiking level
5 20 30 50 100 200 500 1000 1500 a b c
Potato chipsa
Cookiesa
Floura
Cacaob
Baby foodc
Recovery (%)
CV (%)
Recovery (%)
CV (%)
Recovery (%)
CV (%)
Recovery (%)
CV (%)
Recovery (%)
CV (%)
– – – – 87 90 90 93 116
– – – – 4 4 1 3 1
– – – – 87 91 97 96 94
– – – – 5 5 5 6 8
– – 94 94 100 – – – –
– – 5 4 4 – – – –
– – – – 84 87 91 – –
– – – – 9 7 5 – –
88 90 87 84 86 – – – –
20 19 18 12 9 – – – –
Routine extraction/HPLC–MS–MS. Alternative extraction/HPLC–MS–MS. Alternative extraction/GC–MS–MS.
of 5, 20, 30, 50, and 100 g kg−1 of acrylamide to a vegetable stew (baby food) (acrylamide < 5 g kg−1 ) followed by triplicate analysis. The recoveries averaged between 84 and 90% of the acrylamide added (Table 1). The CV determined by 10-fold analysis of skim-milk powder (baby food) was found to be 12% (Table 1). Altogether, more than 20 different product categories, including several different matrices were screened. Fig. 6A summarises the different food categories and the number of samples analysed using the different methods, respectively. As can be seen, the routine method was applicable to 90% of the analysed samples. The extended alternative extraction was essential for only 6% of the tested samples. GC–MS–MS analysis was mainly applied to baby food. As expected potato products and bakery ware presented the most frequently analysed food. However, the figure reflects the complexity of the acrylamide problem since acrylamide can be found in all kind of heat-treated foods and therefore more or less all types of food products have to be analysed. The mean values of the acrylamide concentration analysed in the respective food categories are given in Fig. 6B. Comparison of the number of analysed samples and the mean value of the respective food category reveals that in
some cases the low number of analysed samples did not correlate with an expected low acrylamide concentration. Product groups with remarkable acrylamide values are for example some instant products, roasted vegetables (e.g. onion, chicory) and special herbal teas. This reflects a possible underestimation of some product groups that can
Table 2 Mean, coefficient of variation (CV), and limit of quantification (LOQ) of acrylamide (n = 10) for different food matrices determined using the respective extraction method and HPLC–MS–MS or GC–MS–MS analysis Matrix
Mean (g kg−1 )
CV (%)
LOQ (g kg−1 )
Potato chipsa Crisp breada Butter biscuita Cacaob Coffee, roasteda Coffee, solubleb Baby foodc
620 439 546 190 282 816 8
9.6 3.6 2.7 9.1 9.2 4.1 11.7
30 30 30 30 30 30 5
a b c
Routine extraction/HPLC–MS–MS. Alternative extraction/HPLC–MS–MS. Alternative extraction/GC–MS–MS.
Fig. 6. (A) Number of analysed samples and distribution of the methods used for analysis. (B) Mean values of acrylamide concentrations analysed in the respective food categories.
214
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
contribute to the daily uptake of acrylamide. However, it has been estimated so far that 38% of the calories of the daily intake come from foods containing acrylamide [19]. 3.5. Causes for variability of acrylamide levels Recurrent analysis of acrylamide in the same food product (same recipe, same manufacturing process) revealed a wide variability of acrylamide levels which was mostly higher than the 10% found for analytical variations. Moreover, acrylamide levels vary widely in different product samples of one batch. It was shown for potato chips that acrylamide concentrations analysed in different packages of one batch might range between plus and minus 50% from the mean. However, it has been assumed that the inhomogeneity of the sample matrix or the stability of acrylamide in the food may contribute to the variability of acrylamide levels, as well.
Fig. 8. Decrease of acrylamide concentrations in a homogenised sample of potato chips over 100 days (n = 28).
3.5.1. Inhomogeneity of samples Potato chips were proven to be not homogeneous since its precursors were not uniformly distributed in the potato material. Analyses of a part of a homogenised sample of potato chips which contains predominantly dark particles resulted in one fourth higher acrylamide levels than analyses of the blend (Fig. 7). This finding shows that acrylamide is not evenly distributed—rather it could be concentrated in special particles of the food matrix. 3.5.2. Stability of acrylamide in food Analysis of acrylamide in potato chips over 100 days gave a first hint that acrylamide concentrations are not stable over time (Fig. 8). Since the fine-ground sample was stored in a closed package at room temperature peroxide formation occurred. In the cold extracted fat of the sample a peroxide value of 3.8 mVal O2 kg−1 was determined which was assumed to be the cause of acrylamide degradation. The comparative analysis of beet and cane molasses has shown that beet molasses contains much less acrylamide than cane molasses (approximately 10 times lower). Commonly, this fact was associated with the presence of higher
Fig. 9. Decrease of acrylamide concentrations in a sulphurised 60◦ Brix sugar solution spiked with 1000 g kg−1 acrylamide (䉱) and control sample without addition of sulphite (䊏).
amounts of asparagine in products derived from cane. However, sulphite is used as a preservative agent in beet sugar processing. It was therefore assumed that an oxidative degradation of acrylamide triggered by sulphite could be responsible for the lower acrylamide concentrations determined in beet molasses. In order to simulate the effect of sulphite in sugar processing juices and molasses a 60◦ Brix solution of sugar was added with sodium sulphite (500 mg kg−1 ) and spiked with 1000 g kg−1 acrylamide. For control a 60◦ Brix solution of sugar was spiked with 1000 g kg−1 acrylamide and analysed without addition of sulphite. Acrylamide levels were analysed after 7, 24, 48, and 168 h. As shown in Fig. 9 acrylamide concentrations in the sulphurised sample decrease rapidly and drop down to only 40 g kg−1 after 7 days. On the other hand, acrylamide concentrations remain stable over 168 h without addition of sulphite. These observations gave reasons to suggest that acrylamide concentrations in some foods can vary depending on the storage time since the levels can be affected by special food constituents and/or reaction products. 4. Conclusion
Fig. 7. Analysis of a homogenised sample of potato chips (䉱) and a part of the same blend which contains predominantly dark particles (䊏).
Methods used in routine analysis have to be simple, fast and reliable. The occurrence of acrylamide in various food
K. Hoenicke et al. / Analytica Chimica Acta 520 (2004) 207–215
products makes it very difficult to develop one method which is suitable for all purposes and matrices. Therefore, two different methods were used, one which is very simple, fast and applicable to most samples and a second one which gives reliable results even for difficult matrices like soluble coffee or cacao. For the analysis of acrylamide at low levels GC–MS–MS turned out to be a suitable method since levels up to 5 g kg−1 can be quantified by two mass transitions without derivatisation. However, measurement uncertainties are not only caused by analytical variations but rather by inhomogeneity of the sample matrix or by the instability of acrylamide in the food during storage. The variability of acrylamide levels in food depends on the nature, the manufacturing process, and probably on the storage conditions of the respective food. Altogether, this can cause variations of up to 50% which exceed the analytical variability by nearly a factor of 5. Acknowledgements We like to thank Wolfgang Meins and Marco Gangnus for acrylamide analysis and Bert Pöpping for proof-reading and editing this document. References [1] Swedish National Food Administration, Information about Acrylamide in Food. http://www.slv.se/engdefault.asp.
215
[2] J. Rosén, K.E. Hellenäs, Analyst 127 (2002) 880–882. [3] E. Tareke, P. Rydberg, P. Karlsson, S. Eriksson, M. Törnquist, J. Agric. Food Chem. 50 (2002) 498–506. [4] J.S. Ahn, L. Castle, D.B. Clark, A.S. Loyd, M.R. Philo, D.R. Speck, Food Addit. Contam. 19 (2002) 1116–1124. [5] B. Gutsche, R. Weißhaar, J. Buhlert, Deut. Lebensmi.-Rundsch. 98 (2002) 437–443. [6] A. Becalski, B.P.Y. Lau, D. Lewis, S.W. Seaman, J. Agric. Food Chem. 51 (2003) 802–808. [7] D.S. Mottram, B. Wedzicha, A.T. Dodson, Nature 419 (2002) 448. [8] R.H. Stadler, I. Blank, N. Varga, F. Robert, J. Hau, P.A. Guy, M.C. Robert, S. Riediker, Nature 419 (2002) 449. [9] R. Weißhaar, B. Gutsche, Deut. Lebensm.-Rundsch. 98 (2002) 397– 399. [10] D.V. Zyzak, R.A. Sanders, M. Stojanovic, D.H. Tallmadge, B.L. Eberhart, D.K. Ewald, D.C. Gruber, T.R. Morsch, M.A. Strothers, G.P. Rizzi, M.D. Villagran, J. Agric. Food Chem. 51 (2003) 4782– 4787. [11] T. Wenzl, B. de la Calle, E. Anklam, Food Addit. Contam. 20 (2003) 885–902. [12] L. Castle, J. Agric. Food Chem. 41 (1993) 1261–1263. [13] M. Biedermann, S. Biedermann-Brem, A. Noti, K. Grob, O. Egli, H. Mändli, Mitt. Lebensm. Hyg. 93 (2002) 638–652. [14] C. Fauhl, H. Klaffke, W. Mathar, R. Palvinskas, R. Wittkowski. http:// www.bfr.bund.de/cmc/detail.php?template=internet de index js. [15] T. Wenzl, B. de la Calle, R. Gatermann, K. Hoenicke, F. Ulberth, E. Anklam, Anal. Bioanal. Chem., in press. [16] F. Höfler, R. Maurer, S. Cavalli, GIT 9 (2002) 619–621. [17] J.R. Pedersen, J.O. Olsson, Analyst 128 (2003) 332–334. [18] K. Grob, M. Biedermann, K. Hoenicke, R. Gatermann, Analyst 129 (2004) 92. [19] Milling and Baking News, November 25, 2003. http://www.worldgrain.com/feature stories.asp?ArticleID=68697.