Critical factors and pitfalls affecting the extraction of acrylamide from foods

Critical factors and pitfalls affecting the extraction of acrylamide from foods, Critical factors, and pitfalls affecting the extraction of acrylamide from foods, Critical factors and pitfalls affecting the extraction of acrylamide from foods

Analytica Chimica Acta 557 (2006) 287–295

Critical factors and pitfalls affecting the extraction of acrylamide from foods: An optimisation study

Erik V Peterssona,b,∗, Johan Ros´ena, Charlotta Turnerb,

Rolf Danielssonb, Karl-Erik Hellen¨asa

aNational Food Administration, P.O Box 622, 751 26 Uppsala, Sweden

bUppsala University, Institute of Chemistry, Department of Analytical Chemistry, P.O Box 599, 751 24 Uppsala, Sweden

Received 9 August 2005; received in revised form 5 October 2005; accepted 9 October 2005

Available online 29 November 2005

Abstract

A stepwise study of common factors for the extraction of acrylamide (AA) from relevant food matrices was performed The investigated extraction factors were sample particle size (fine or coarse), defatting (yes or no), extraction solvent (water or water/methanol), homogenisation

by Ultra Turrax (yes or no), extraction temperature (25 or 60◦C) and extraction time (5 min to 17 h) An optimised method comprised the use of fine particles (<1000␮m), water as the extraction solvent and shaking of the sample for 45 min at 25◦C This extraction method was suitable for

all tested matrices (coffee, crispbread, mashed potatoes, milk chocolate and potato crisps) The analytical results (from LC-MS/MS analysis after SPE clean-up) correlated well with those obtained by the original, more labour-intensive, extraction procedure and there was excellent agreement with the assigned AA levels of several proficiency test samples analysed for evaluation Defatting and additional homogenisation by Ultra Turrax did not increase the AA yield when other factors were set to appropriate levels In general, the study indicates that incomplete extraction is the most likely cause of erroneous results This might occur when the food is not sufficiently macerated, when water/methanol is used as the extraction solvent, and when using a short extraction time or low extraction temperature, especially when these conditions are combined Formation of AA during the extraction procedure is another possible error source, though it was not seen under any of the experimental conditions employed when

a fructose-enriched blank potato tissue was extracted

© 2005 Elsevier B.V All rights reserved

Keywords: Acrylamide; Foods; Analysis; LC-MS/MS; Extraction; Optimisation; Method optimisation

1 Introduction

Since the initial discovery of acrylamide (AA) as a

heat-induced food toxicant in fried rat feed in 2000 [1], and then

in heated foods in 2002[2], several methods for the analysis

of AA in foods have been developed The most common

meth-ods include chromatography/detection with LC-MS/MS (liquid

chromatography tandem mass spectrometry) or GC-MS (gas

chromatography mass spectrometry; with or without

derivati-sation of the analyte) In a review from 2003, special attention

was given to sample preparation, which represents the

great-est differences between the methods[3] It was concluded that

several steps of the extraction procedure had not been

thor-∗Corresponding author Tel.: +46 18 175762; fax: +46 18 105848.

E-mail address: erik.petersson@kemi.uu.se (E.V Petersson).

oughly investigated, including the influence of extractant com-position, extraction temperature/time, defatting and mechanical treatment

An evaluation of the results from a proficiency test organ-ised by the German Federal Institute for Risk Assessment (BfR)

in 2002 also showed great inter-laboratory differences, suppos-edly related to different extraction procedures[4] In particular, there was a significant difference between aqueous and non-aqueous extraction of AA from cocoa powder Also, in another proficiency test, jointly organised by BfR and the European Commission’s Joint Research Centre (JRC), a cocoa sample caused large variations, which made it impossible to establish

an assigned value[5]

In 2004, an evaluation of an inter-laboratory comparison study arranged by JRC showed that a large number of the mea-sured AA levels were outside the acceptable range, especially for a crispbread sample[6] The choice of analysis technique

0003-2670/$ – see front matter © 2005 Elsevier B.V All rights reserved.

doi:10.1016/j.aca.2005.10.014

(LC-MS/MS or GC-MS) had a significant influence on the

results GC-MS without derivatisation was found to lead to

over-estimation of the AA content Furthermore, the composition of

the extraction solvent had a significant influence on the results,

especially for crispbread samples It was concluded that

addi-tional work is necessary in order to identify problematic steps

in the analytical procedures

Certain pitfalls affecting the extraction of AA have been

proposed Three authors suggested in situ formation of AA

during soxhlet extraction of potato crips at 60◦C in methanol

[7–9] This was a method published by Pedersen and Olsson

extrac-tion (ASE) for extracextrac-tion of food samples with ethyl acetate

at elevated temperature and pressure, found incomplete

extrac-tion of AA from cacao and milk powder compared to the use

of water as an extraction solvent Under the same conditions,

formation of AA in raw sugar samples was suspected, but not

both incomplete extraction and formation as possible pitfalls

Furthermore, formation has also been suggested in the

injec-tion port during direct GC-MS analysis (if AA precursors are

present)[12] Other possible pitfalls include contamination of

thermal degradation of AA In a very recent publication, it has

been proposed that up to 4.1 times higher amounts of AA can

be extracted from certain foods at strongly alkaline conditions

(pH 12) compared to extraction at more normal conditions (pH

6)[14] It is not clear whether this is physically or chemically

released AA, or if it is bioavailable Therefore, the relevance

of this seemingly higher AA content for human exposure is

unknown

Commonly used methods for the determination of AA

in foods include extraction, clean-up (in some methods also

derivatisation) and chromatography/detection It has been

con-cluded that investigations concerning the extraction step are

investigation of common extraction factors and their resulting

AA yield from relevant food matrices An optimised

extrac-tion method is presented together with a discussion on possible

pitfalls

2 Experimental

2.1 Apparatus

MS was performed using a triple quadrupole: Micromass

Quattro Ultima with electrospray (Micromass UK Ltd.,

Altrin-cham, Cheshire, UK) The HPLC system used was a Waters

Alliance 2695 (Waters Ltd., Watford, Hertfordshire, UK)

material (Hypersil-Keystone, Runcorn, UK) The Ultra Turrax

homogeniser was a DISP 25 (InterMed, Roskilde, Denmark)

The incubated shaker was a ROSI 1000 (Thermolyne, Dubuque,

IA, USA) The stack-sieving device was a B¨uhler-Miag model

300/B (Milano, Italy)

2.2 Reagents and chemicals

Acrylamide, cyclohexane, methanol, 2-propanol and ethyl acetate were supplied by Merck (Darmstadt, Germany), ethanol was supplied by Kemetyl AB (Haninge, Sweden), triple-labelled

supplied by CDN Isotopes (Pointe-Claire, Que., Canada) The solid phase extraction (SPE) columns used were ENV+ 1 g and MultiMode 1 g from IST (Mid Glamorgan, UK) HPLC-grade acetonitrile was supplied by Lab-Scan (Dublin, Ireland) and pure water was produced from a Millipore purification system

2.3 Food samples

Food samples were of Swedish produce and purchased off-the-shelf from a retail store in Uppsala, Sweden After

samples were classified in two groups: (1a) samples containing only fine particles in the retail package and (1b) samples further homogenised into two portions giving coarse and fine particles, respectively

(1a) Samples containing only fine particles:

Raw potatoes (blank) were prepared from 1000 g fresh

potatoes mashed together with 20 g fructose

Coffee (500 g) medium-roasted powder.

(1b) Samples homogenised into coarse and fine particles:

Crispbread: Half a package (250 g of 500 g) of crispbread

was divided into two portions and homogenised with a mortar and a pestle into “coarse” and “fine” particles, respectively The size-distribution of the particles were characterised by stack-sieving for 30 min using 1610,

1110, 670, 460, 219 and 150␮m sieves This

characteri-sation showed that the main fraction of the coarse portion was in the 1610␮m mesh and the main fraction of the fine

Milk chocolate bars: Two packages (2× 100 g) of milk

chocolate bars containing around 30% cocoa powder were homogenised by grating into “coarse” and “fine” particles, respectively Characterisation by stack-sieving (performed as for crispbread) showed the main fraction

of the coarse portion particles in the 1110␮m mesh and

of the fine portion particles in the 670␮m mesh

Potato crisps: A package (200 g) of potato crisps with

onion-taste was divided into two portions, which were homogenised with a mortar and a pestle into “fine” and

“coarse” particles, respectively The particle size distribu-tion was not possible to characterise by stack-sieving due

to the fatty and sticky consistency However, there was a clearly visible difference between the portions of coarse and fine particles, and the main fraction of the particles was estimated to be >1000 and <1000␮m, respectively

Proficiency test samples: Food Analysis Performance

Assess-ment Scheme (FAPAS®) breakfast cereal test material was round

10, series 30 obtained from FAPAS, CSL (York, UK)

E.V Petersson et al / Analytica Chimica Acta 557 (2006) 287–295 289

Four different samples from “the third proficiency test, jointly

organised by BfR and DG-JRC IRMM, on the analysis of

acry-lamide from coffee and cocoa samples” consisting of cocoa

powder, coffee, coffee extract and coffee surrogate

2.4 Procedures

2.4.1 AA determination procedure

The sample extraction and clean-up were performed with a

modified version of the established in-house method[13,15]

2.4.1.1 Typical sample extraction Different amounts of

sam-ple were used depending on the sensitivity needed or if the

matrix was known to contain interfering substances Two grams

of sample (4 g for crispbread and raw potatoes) was weighed

into a 50 ml falcon-tube If applicable, fine or coarse

parti-cles were chosen Some samples were defatted with 10 ml

of cyclohexane for 30 min (short, gentle shaking by hand

every 10th minute) Then 40 ml of extraction solvent (water

stan-dard (acrylamide-2H3; 1000 ng ml−1) were added to the tube

(giving a two-phase extraction if cyclohexane was added

pre-viously) In some cases, the sample was further homogenised

with an Ultra Turrax homogeniser (2 min, 9500 rpm) The

sam-ple was always extracted in an incubated horizontal shaker

(100 rpm) at different times and temperatures (no pre-warming

of extraction solvent was made) according to the experimental

conditions, followed by centrifugation in a cooling centrifuge

(10◦C, 4000 rpm, 20 min) Afterwards a portion of the

super-natant was removed (10 ml from 4 g samples; 2 ml from 2 g

samples)

2.4.1.2 Sample clean-up Supernatants containing methanol

were concentrated under vacuum in a vortex evaporator (40◦C,

30 min) in order to evaporate the methanol, and then the volume

was made up to the original with water An SPE column

(mul-timode, 1 g) was pretreated with acetonitrile (3 ml) and water

(2× 6 ml) in a vacuum manifold The supernatant was passed

through the column (for 2 g samples: additional rinsing

after-wards with 3 ml water, which was also collected) giving 10 or

5 ml eluate collected The eluate was loaded onto a second SPE

column (ENV+, 1 g) that had been pretreated with methanol (5 ml) and water (5 ml) in a vacuum manifold The column was

methanol in water The first fraction of the eluate was discarded (because it was free from AA) and only the fraction at 1.7–3.7 ml was collected The collected extract was evaporated by means of

of approximately 0.5–1 ml

2.4.1.3 LC-MS/MS analysis Three 10␮l injections of the final

extract were analysed using LC-MS/MS according to the in-house method[16], the only modification being that 0.1% (v/v) acetic acid in water was used as a mobile phase For

quan-tification of AA the area ratio of m/z 55/58 (analyte/AA-2H3) was used, and concentrations were calculated against a stan-dard curve (4, 8, 40, 400 and 4000 ng ml−1) Ion-suppression

AA-13C1to 200␮l of the extract just before LC-MS/MS

analy-sis The ion-suppression was calculated as the relative difference

com-pared with those obtained for standard solutions

For calculations of the AA losses, the area ratio of m/z 56/58

was compared with the expected ratio

Naturally occurring13C in native AA contributes to the mea-sured level of the second internal standard (13C1-AA) For the samples with the highest AA levels (giving the greatest potential contribution of13C from native AA), in this work potato crisp samples (∼1500 ␮g kg−1AA), this contribution was always less than 8%

2.4.2 Stepwise study of extraction parameters

The experiments were performed in a stepwise study con-sisting of four different steps The software package Minitab® 12.22, State College, PA, USA, was used to create experimental designs and to evaluate the results

2.4.2.1 Screening experiment (step I) Two different

experi-mental designs were used, one with four factors (design (1a) for foods with only fine particles,Table 1) and one with six factors (design (1b) for foods homogenised into fine or coarse particles,

Table 2) The experiments were performed without replicates

Table 1

Factors, levels and acrylamide results for design (1a)

Defatting Extracion solvent Extraction time (min) Extraction temperature ( ◦C) AA results (␮g kg −1)

Coffee Potatoes

Coffee and raw fructose-enriched mashed potatoes were the tested foods, consisting of only fine particles A fractional factorial design (resolution IV) without replicates was used, which resulted in 8 experiments per foodstuff.

Table 2

Factors, levels and acrylamide results for design 1b

Particle size Defatting Extraction

solvent

Ultra Turrax Extraction

temp ( ◦C) Extractiontime (min) Block AA results (␮g kg −1)

Crispbread Milk chocolate Potato crisps

Crispbread, milk chocolate and potato crisps were the tested foods, homogenised into fine or coarse particles A fractional factorial design resolution IV was used, which resulted in 16 experiments per foodstuff The design was blocked into two blocks.

and central points To determine the experimental errors,

sepa-rate experiments with replicates were utilized For design (1a),

five replicates were made with defatting, 40% MeOH as an

extraction solvent, 32.5 min extraction time and 42.5◦C

extrac-tion temperature The foods with design (1b) were further

inves-tigated in step II with a design that included replicates (Table 3),

from which the experimental error was estimated In both cases,

the standard deviation (S.D.) for replicates was obtained with

four degrees of freedom The influence of the experimental error

on the effects is given by S.D.eff=√

[(4× S.D.2)/n], where n is

the number of experiments in the design (8 and 16, respectively,

for designs (1a) and (1b)) From S.D.eff the significance level

(p-value) and least significant value for the effects were

calcu-lated

2.4.2.2 The interaction between particle size and Ultra

Tur-rax homogenisation (step II) A full factorial replicated design

(design 2 for foods homogenised into fine or coarse particles,

Table 3

Factors, levels and acrylamide results for design 2, without defatting and using

water as extraction solvent at 42.5 ◦C for 32.5 min

Design 2 AA results ( ␮g kg −1)

Particle size Ultra Turrax Crispbread Milk chocolate Potato

crisps

Crispbread, milk chocolate and potato crisps were the tested foods, homogenised

into fine or coarse particles A full factorial design with replicates was used,

resulting in eight experiments per foodstuff.

Table 3) was used to evaluate the interaction between the two factors

2.4.2.3 Optimisation of quantitative factors (step III) A full

factorial design with replicates was used to optimise extraction temperature and time Two different temperatures and four dif-ferent times resulted in eight difdif-ferent combinations

2.4.2.4 In-house comparison between optimised and estab-lished in-house method (step IV) The estabestab-lished in-house

method was similar to the typical sample extraction method described in the procedures section, with a few exceptions: no defatting was used, water was used as an extraction solvent and the sample was only extracted with an Ultra Turrax homogeniser (2 min, 9500 rpm), not shaken in an incubated shaker[13,15] The optimised method was also similar to the typical sample extraction method, but comprised water as an extraction sol-vent, small sample particles and horizontal shaking for 45 min

at 100 rpm and 25◦C, omitting defatting.

3 Results and discussion

3.1 Stepwise study of extraction parameters

The study comprised four consecutive steps Step I was an initial screening experiment to examine the importance of six typical extraction factors (chosen by considering commonly used methods[3,4,6,17]): particle size, defatting, extraction sol-vent, Ultra Turrax homogenisation, extraction temperature and extraction time With this experiment the optimal level for the qualitative factors (i.e all factors except temperature and time) was established In step II, a study on the interaction between particle size and Ultra Turrax homogenisation was investigated

to find out if the use of Ultra Turrax was necessary for small particles The quantitative factors (i.e temperature and time)

E.V Petersson et al / Analytica Chimica Acta 557 (2006) 287–295 291

were optimised in step III Finally, in step IV the new optimised

extraction method was compared with the established in-house

method The results were also compared with the assigned

val-ues of proficiency test samples

The AA concentrations were measured with LC-MS/MS

(using an internal standard) after SPE clean-up In addition, AA

losses and ion suppression were determined in steps I and II by

the use of a second internal standard added to the sample extract

prior to LC-MS/MS analysis In all the steps, attention was paid

not only to finding optimal settings, but also to identifying

pos-sible extraction pitfalls that lead to incomplete extraction, for

example, or undesired formation of AA during the extraction

The foods tested were coffee, crispbread, milk chocolate and

potato crisps, all relevant foods giving large contributions to the

coffee and milk chocolate have been identified as problematic

(asparagine) and reducing sugars at temperatures above 100◦C

has been suggested as a major formation pathway for AA in

heated foods[21–23] Sugar is the limiting factor for AA

forma-tion in potatoes, whereas asparagine is in excess[24] Therefore,

raw mashed potatoes, containing 2% added fructose, was used

as a blank to reveal any formation of AA during the extraction

Fructose was selected as it has been shown to be the most

effec-tive sugar for AA formation[25,26]

3.1.1 (I) Screening step

3.1.1.1 Formation The raw, sweetened mashed potatoes used

as a blank showed no quantifiable (i.e <5␮g kg−1) response,

not even when using 80% methanol in combination with a

high temperature (60◦C) and a long extraction time (60 min).

This specific combination could be interesting because reports

of AA formation during extraction so far involve organic

sol-vents [7,11] In order to further examine possible formation,

some additional experiments were carried out with seven

dif-ferent organic extraction media: 80% methanol, 90% methanol

and 100% methanol, 2-propanol, ethanol, acetonitrile and ethyl

acetate under similar conditions (no defatting, no Ultra Turrax

homogenisation, extraction time 60 min, extraction

tempera-ture 60◦C) However, these experiments showed no quantifiable

signs of formation, not even when injecting three times the

nor-mal amount of extract into the LC-MS/MS system in order to

increase the sensitivity Possible formation of AA when using

even longer extraction times combined with a high extraction

temperature was further investigated in step III, but then only

with water as the extraction solvent

Since formation of AA under the conditions used in step I

could not be shown, all low AA levels obtained in step I were

considered as incomplete extraction (as discussed in the

follow-ing section)

3.1.1.2 Impact of the extraction factors The levels of the

extraction factors were chosen by considering commonly used

methods[3,4,6,17] Since only two levels for each factor were

used, these had to be chosen realistically and at the same

time had to be “extreme” enough to give decisive effects

The factors and levels for designs (1a) and (1b) are shown

Fig 1 Relative effects of extraction factors’ (effect divided by mean value) on

AA yield A positive value corresponds to increased AA yield for the factor level indicated in the label To be significant at the 1% level, the size of an effect must exceed the significance limits shown to the right.

inTables 1 and 2, respectively, together with the correspond-ing individual AA results The AA levels obtained were in the

for potato crisps The lowest AA results were obtained with 80% methanol as the extraction solvent in combination with other unfavourable conditions, e.g short extraction time and coarse particles Results from the statistical evaluation of the designed study are shown inFig 1, showing the relative effects (effect divided by mean value) on the AA yield for the different extrac-tion factors The overall pattern is fairly similar for the different foods, indicating the possibility of a generic extraction proce-dure for a majority of food types The 1% significance level for each factor (the least significant effect) is also indicated, based

on analytical variability estimates obtained with four degrees of freedom A 1% significance level was found to be more relevant than a 5% significance level, as the replicate experiments were made under repeatability conditions and were separated from the experimental design The RSD values were: 4% (coffee), 2% (crispbread), 13% (milk chocolate) and 6% (potato crisps) The absence of significant effects with chocolate is partly explained

by the higher analytical variability for this item, probably due

to its low AA content

3.1.1.2.1 Defatting. The use of defatting with cyclohex-ane had in most cases no significant effect on the AA yield, not even for fatty matrices It showed, however, significance

rea-son for this negative effect is unknown Defatting was accom-plished in the same tube as the extraction and the cyclohexane was left in the tube throughout the whole extraction (giving a two-phase extraction step) The risk for loss of sample prior

to extraction that could otherwise occur when the solvent was removed was thereby minimised Defatting did not show any benefits for the MS detection in this work Nevertheless, it might have other advantages, for example, longer lifetime of the HPLC-columns and prevention of clogging/overloading the SPE-columns Defatting of samples prior to AA analysis is com-mon practice; for example, 37 out of 72 laboratories used it in

a proficiency test evaluated in 2004[6], but no systematic study has so far been performed to demonstrate its feasibility

3.1.1.2.2 Sample particle size. This factor had a

signifi-cant effect for the samples with a big difference between the fine

and coarse particle portions, e.g potato crisps and crispbread

Therefore, it was concluded that it is important to use sufficiently

small particles (<1000␮m) in order to obtain adequate

extrac-tion yield However, samples containing very fine particles can

be difficult to disperse and might also clog the SPE-columns In

our laboratory, sample disintegration is routinely accomplished

by using a food-processor Other possibilities might include

grinding in a laboratory mill or freeze-drying followed by

pul-verisation

3.1.1.2.3 Homogenisation by Ultra Turrax. Wet

homo-genisation and vigorous mixing of the sample in extraction

sol-vent for 2 min with an Ultra Turrax prior to the general extraction

step gave significant positive effects with potato crisps and

crisp-bread The need for Ultra Turrax homogenisation was further

investigated in step II, as discussed later

3.1.1.2.4 Extraction solvent. This factor showed

signif-icance for all the foods except milk chocolate In general,

pure water enabled higher AA recovery compared to 80%

methanol The use of pure water gives effective extraction of

AA, but also coextraction of other unwanted compounds in the

matrix (e.g salts, proteins, carbohydrates), which might

inter-fere with the detection and degenerate the chromatographic

system, if not removed by clean-up Varying results

concern-ing the choice of extraction solvents have been presented by

other authors One group tested accelerated solvent extraction

led to incomplete extraction of AA from cacao and milk

Another group tried soxhlet extraction of potato crips at 60◦C

BfR, various methods using organic solvents during extraction

of cocoa powder generally led to higher AA results compared

organ-ised by JRC, extraction of crispbread and butter cookies with

mixtures of water and organic solvents seemed to give higher

results than extraction with pure water or pure organic

sol-vents [6] However, it might be difficult to draw conclusions

about the influence of single factors in proficiency tests, as

other factors at the same time may differ considerably between

laboratories (e.g one laboratory using water extraction and

LC-MS and another laboratory using organic solvent extraction

and direct GC-MS) Nonetheless, pure water seems to be the

most popular extraction solvent, for example, it was used by

more than 50% of the laboratories in the two proficiency tests

[4,6]

3.1.1.2.5 Extraction temperature. This factor showed

sig-nificance for all the food samples except milk chocolate

Extrac-tion at 60◦C gave higher extraction recovery compared to 25◦C.

This issue is further discussed in step III

3.1.1.2.6 Extraction time. By using a long extraction time

of 60 min, higher extraction recovery was obtained than when

using 5-min extraction The effect was significant for all the food

samples except milk chocolate This issue is further discussed

in step III

The screening experiment, step I, was set up to get “clean” main factors with resolution IV (i.e not to be confounded with two-factor interactions) From the results of step I, the most important main factors could be identified, and the qualitative part of them could be set to their optimal level The quantitative factors were then later optimised in step III

As a consequence of the rather low resolution, the two-factor interactions in step I were confounded with at least one other two-factor interaction It was not possible to resolve the inter-actions, and therefore no further conclusions were drawn about them

3.1.2 (II) The interaction between particle size and Ultra Turrax homogenisation

This study was performed to find out if there was an interac-tion between Ultra Turrax homogenisainterac-tion and particle size, or more specifically if the Ultra Turrax effect would appear only with coarse sample particles This possible interaction could not

be evaluated in design (1b), as discussed above

The results from step II (raw data shown inTable 3), evalu-ated with a 1% significance-level, showed significantly higher extraction yield of AA from crispbread and potato crisps when using fine particle size For milk chocolate, there was no signif-icant effect, probably because of the small difference between the fine and coarse particle portions However, under the experi-mental conditions employed in this step, further homogenisation

by Ultra Turrax gave no significant effect either with coarse or fine particles for any of the food items The explanation for this apparent inconsistency with the results obtained in step I is probably related to the fact that in step II the other qualitative factors were set to their optimal level, e.g water was used as the only extraction solvent, and the quantitative factors were set to their central point value Thus, if there actually was an interac-tion between Ultra Turrax and some other factor(s) than particle size, that effect could not be detected in step II

3.1.2.1 suppression and AA losses in steps I and II

Ion-suppression, i.e a lower detector signal from samples than from standard solutions, caused by matrix effects in the MS interface,

is a common problem in LC-MS The use of a second isotope labelled AA, in addition to the normal internal standard, made it possible to estimate the degree of MS ion-suppression for the dif-ferent samples and extraction conditions This second standard, AA-13C1, was added to the final extract just prior to the LC-MS/MS step The results showed that the ion-suppression was matrix-dependent, while the different extraction factors, includ-ing defattinclud-ing, showed no significant effect (data not shown) Mashed potatoes and potato crisps showed no ion-suppression while crispbread and coffee gave 20–50% and 30–40% suppres-sion, respectively Milk chocolate surprisingly demonstrated a higher signal (10%) than the standard solutions Ion-suppression

is, amongst other things, dependent on the concentration of the extract and the type of matrix In the sample preparation meth-ods used in this paper, the 4 g samples (crispbread and mashed raw potato) were concentrated about 10–15 times and the 2 g samples (coffee, milk chocolate and potato crisps) about two to three times during SPE clean-up and evaporation Stronger

ion-E.V Petersson et al / Analytica Chimica Acta 557 (2006) 287–295 293

suppression was, as expected, found in the more concentrated

samples (e.g crispbread), but also in known difficult matrices

like cocoa[19]and coffee[19,20]

The two differently labelled AA internal standards were also

employed to estimate the losses of AA during extraction and

work-up The losses for coffee, milk chocolate and potato crisps

were between 0 and 20%, while crispbread and raw potatoes

showed somewhat higher losses of 30–35% The reason for the

difference between these two groups of matrices was

proba-bly that the SPE clean-up step differed slightly for them (see

Section2) The experiment did not directly enable

differenti-ation between losses occurring during the extraction step (e.g

by degradation or reaction of AA with matrix components) and

losses due to incomplete recovery during the subsequent SPE

work-up of the extracts However, since the losses measured

were similar to those observed during the earlier method

devel-opment of the SPE step, it was concluded that deterioration of

AA during extraction was negligible under the conditions used

in the present study

Since an internal standard is normally used, ion suppression

and/or varying extraction yields will have little impact on the

quantification of AA, although both might negatively affect the

detection limit and the precision at the lower end of the working

range The lack of any special problem caused by extra extraction

of salts, proteins, carbohydrates, etc., is probably due to the use

of an effective two-stage SPE clean-up No interfering peaks

were found, and ion-suppression was never greater than 50%

for problematic matrices like coffee

3.1.3 (III) Optimisation of quantitative factors

This step was performed to optimise the quantitative factors

used in steps I and II, which were extraction time and extraction

temperature All qualitative factors were set to their optimal

levels (i.e fine particles were extracted with water, omitting

defatting and Ultra Turrax.) All food samples from groups (1a)

and (1b) were studied The temperature was set to 25 and 60◦C

and the time was set to 30, 60, 120 min and, as an extreme value,

overnight (17 h) A set-up corresponding to a full factorial design

with replicates was used, and the results are shown inFigs 2–5

3.1.3.1 Raw potato No AA formation (i.e <5␮g kg−1) could

Fig 2 Extracted amount of AA from coffee depending on time and temperature.

The error bars indicate 5% significance level.

Fig 3 Extracted amount of AA from crispbread depending on time and tem-perature The error bars indicate 5% significance level.

Fig 4 Extracted amount of AA from potato crisps depending on time and temperature The error bars indicate 5% significance level.

overnight and injecting triple the normal amount of extract into the LC-MS/MS system to increase the sensitivity

3.1.3.2 Coffee, crispbread and potato crisps Raising the

recovery, as can be seen inFigs 2–4 Moreover, a plateau value was reached already at the shortest time studied, i.e after 30 min This result is in agreement with another study, where no increase

in AA yield was found for potato crisps extracted with water in

a 60◦C ultrasonic bath for 15, 30 and 60 min[11].

3.1.3.3 Milk chocolate The only notable effect when

increas-ing the extraction temperature and time was found for milk

Fig 5 Extracted amount of AA from milk chocolate depending on time and temperature The error bars indicate 5% significance level.

a long extraction time at 60◦C (51% from 30 min to 17 h).

This increase could be the result of either AA formation or of

incomplete extraction Acrylamide might be entrapped in the

matrix and the extraction solvent needs both a long time and a

high temperature to penetrate the structure However, defatting

and two-phase extraction with cyclohexane did not increase the

extractability in the step I experiment Furthermore, no

signif-icant effect could be seen for milk chocolate in step I when

changing temperature or time, and nearly the same AA values

were reached already after 5 min in step I as after up to 2 h in step

III Not even extraction overnight at 25◦C increased the yield.

This suggests that extraction of AA is quite easy from this matrix,

and that the higher levels reached at 60◦C over night (17 h)

might implicate formation Although no AA was detected in the

fructose-enriched potato tissue in the present experiment,

forma-tion in chocolate might occur via a different reacforma-tion mechanism

enabling formation at 60◦C Formation at such a low

tempera-ture has previously been demonstrated, e.g in foods containing

excluded in the present experiment Therefore, 25◦C is

prefer-able as a generic extraction temperature The results from the

milk chocolate experiments can be used as base for further

inves-tigations of time/temperature effect on different brands of milk

chocolate along with pure cocoa powder Furthermore,

inves-tigations of the impact of high/low pH on the release of AA

from this matrix might be of use to study possible formation

mechanisms differing from those already known

3.1.4 (IV) In-house comparison between optimised and

previously used extraction method

Optimised method Based on the knowledge gained so far

in this study, a simple, optimised extraction method was

estab-lished It consisted of pre-grinding of samples to a particle size of

1000␮m or smaller followed by extraction with water at 25◦C

for 45 min using a horizontal shaker (100 rpm)

The performance of the optimised extraction method was

finally evaluated through analysis of a number of samples

from previous proficiency tests The new procedure was used

in parallel with the extraction procedure of an established

in-house analytical method, both utilising SPE clean-up of

and their AA levels (optimised method result and assigned

value, respectively) were: breakfast cereal (58 and 61␮g kg−1),

cocoa powder (101␮g kg−1; no assigned value), coffee (252 and

258␮g kg−1), coffee extract (867 and 858␮g kg−1) and

cof-fee surrogate (1356 and 1334␮g kg−1) Good linear correlation

(R2= 1; slope = 0.98; intercept = 8) was obtained when

compar-ing AA levels measured with the optimised extraction method

directly against the assigned values of the samples from the

pro-ficiency tests The results of the optimised extraction method

were also well correlated with those of the established method

(R2= 0.998; slope = 0.93; intercept = 17) A major advantage of

the optimised method is the higher sample throughput and less

manpower needed due to its automated nature The use of an

automated shaker allows the unattended extraction of more than

50 samples at the same time, and a time-consuming cleaning of

equipment between samples is not needed

4 Conclusions

The present study demonstrates that AA can be efficiently extracted from various food matrices using plain water The use

of aqueous alcohol as the extraction solvent, or removal of fat by

an organic solvent, gave negative or non-significant effects on the extraction yield Disintegration of the samples to small particles prior to extraction made additional homogenisation and vigor-ous mixing by Ultra Turrax unnecessary Quantitative extraction was achieved by gentle mixing at room temperature, and sta-ble results were obtained when the extraction time was varied between 30 min and 17 h

In general, the results from the optimisation experiments indicate that incomplete extraction is the most likely cause of erroneous results in the extraction of AA This might occur when the food is not finely divided enough, when using organic solvent extraction, a short extraction time and a low extraction temper-ature, in particular when using these factors in combination Other possible error sources might be the destruction or for-mation of AA during the extraction procedure Forfor-mation was not seen under any of the experimental conditions employed when a fructose-enriched blank potato tissue was extracted However, increased AA levels were noted at prolonged extrac-tion times when a chocolate sample was extracted at 60◦C for

up to 17 h It was concluded that the destruction of AA during extraction and work-up was small or absent for the investigated foods

Based on these results, an optimised extraction procedure was devised that should be suitable for a wide range of food matrices The method consisted of the use of disintegrated (particle size

<1000␮m) samples, water as the extraction solvent and with

shaking of the sample horizontally (100 rpm) at 25◦C for 45 min. This is a gentle and simple method without organic solvents It allows for a high sample throughput since multiple samples can

be extracted in parallel

In order to fully utilise this optimised extraction procedure,

it should ultimately be combined with optimised clean-up and chromatography/detection-steps In the present study, the new extraction procedure was employed within an established in-house LC-MS/MS method using an effective two-stage SPE clean-up of extracts Analytical results correlated well with those obtained by the original, more labour-intensive, extraction pro-cedure and there was excellent agreement with the assigned AA levels of several proficiency test samples analysed for evalua-tion

Acknowledgements

Special thanks to colleagues at the Swedish National Food Administration: Birgitta Hellkvist for assistance with some of the SPE clean-up, and Carina Branzell for help during the homogenisation of the potatoes The authors would also like to thank Ann-Christine M˚artensson (Swedish University of Agri-cultural Sciences) for assistance during the stack-sieving This work has been carried out with support from the European Commission, Priority 5 on Food Quality and Safety (Contract

no FOOD-CT-2003-506820 Specific Targeted Project),

‘Heat-E.V Petersson et al / Analytica Chimica Acta 557 (2006) 287–295 295

generated food toxicants—identification, characterisation and

risk minimisation’ This publication reflects the author’s views

and not necessarily those of the EC The information in this

doc-ument is provided as it is and no guarantee or warranty is given

that the information is fit for any particular purpose The user

thereof uses the information at their sole risk and liability

References

[1] E Tareke, P Rydberg, P Karlsson, S Eriksson, M T¨ornqvist, Chem.

Res Toxicol 13 (2000) 517.

[2] E Tareke, P Rydberg, P Karlsson, S Eriksson, M T¨ornqvist, J Agric.

Food Chem 50 (2002) 4998.

[3] T Wenzl, M.B de la Calle, E Anklam, Food Addit Contam 20 (2003)

885.

[4] C Fauhl, H Klaffke, W Mathar, R Palvinskas, Acrylamide

interlabora-tory study, Federal Institute for Risk Assessment, Berlin, Germany, 2002,

http://www.bfr.bund.de/cm/208/acrylamide interlaboratory study 2002.

pdf.

[5] T Wenzl, Detailed report on the third European inter-laboratory

com-parison study on the determination of acrylamide in food, Joint

Research Centre – Institute for Reference Materials and Measurements,

Geel, Belgium, 2005, http://www.irmm.jrc.be/html/activities/acrylamide/

finalreport acrylamide PT3 JRC.pdf.

[6] T Wenzl, B de la Calle, R Gatermann, K Hoenicke, F Ulberth, E.

Anklam, Anal Bioanal Chem 379 (2004) 449.

[7] K Grob, M Biedermann, K Hoenicke, R Gatermann, Analyst 129

(2004) 92.

[8] J.W DeVries, B.E Post, Analyst 129 (2004) 93.

[9] M Tanaka, Y Yoneda, Y Terada, E Endo, T Yamada, Analyst 129 (2004) 96.

[10] J.R Pedersen, J.O Olsson, Analyst 128 (2003) 332.

[11] K Hoenicke, R Gatermann, W Harder, L Hartig, Anal Chim Acta

520 (2004) 207.

[12] L Castle, S Eriksson, J AOAC Int 88 (2005) 274.

[13] P Fohgelberg, J Ros´en, K.-E Hellen¨as, L Abramsson-Zetterberg, Food Chem Toxicol 43 (2005) 951.

[14] S Eriksson, P Karlsson, Lebensm.-Wiss Technol., (15 November 2005)

in press.

[15] H Fredriksson, J Tallving, J Ros´en, P ˚ Aman, Cereal Chem 81 (2004) 650.

[16] J Ros´en, K.-E Hellen¨as, Analyst 127 (2002) 880.

[17] D.B Clarke, J Kelly, L.A Wilson, J AOAC Int 85 (2002) 1370 [18] K Svensson, L Abramsson, W Becker, A Glynn, K.-E Hellen¨as, Y Lind, J Ros´en, Food Chem Toxicol 41 (2003) 1581.

[19] S Riediker, R.H Stadler, J Chromatogr A 1020 (2003) 121 [20] D Andrzejewski, J.A.G Roach, M.L Gay, S.M Musser, J Agric Food Chem 52 (2004) 1996.

[21] D.S Mottram, B.I Wedzicha, A.T Dodson, Nature 419 (2002) 448 [22] R.H Stadler, I Blank, N Varga, F Robert, J Hau, P.A Guy, M.-C Robert, S Riediker, Nature 419 (2002) 449.

[23] R Weisshaar, G Gutsche, Dtsch Lebensm -Rundsch 98 (2002) 397 [24] A Becalski, B.P.-Y Lau, D Lewis, S.W Seaman, S Hayward, M Sahagian, M Ramesh, Y Leclerc, J Agric Food Chem 52 (2004) 3801.

[25] M Biedermann, A Noti, S Biedermann-Brem, V Mozzetti, K Grob, Mitt Lebensm Hyg 93 (2002) 668.

[26] P Pollien, C Lindinger, C Yeretzian, I Blank, Anal Chem 75 (2003) 5488.

[27] M Biedermann, K Grob, Mitt Lebensm Hyg 94 (2003) 406.