This article was downloaded by: [University of Tennessee, Knoxville] On: 22 August 2014, At: 01:52 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Food Additives & Contaminants: Part A Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/tfac20
Determination of caramel colourants by-products in liquid foods by ultrahigh- performance liquid chromatography- tandem mass spectrometry (UPLCMS/MS) a
a
b
Séverine Goscinny , Vincent Hanot , Hasna Trabelsi & Joris Van Loco
a
a
Department of Food, Medicines and Consumer Safety, Scientific Institute of Public Health, Brussels, Belgium b
Department of UFR Biomedical, University of Paris Descartes, Paris, France Accepted author version posted online: 25 Jul 2014.
To cite this article: Séverine Goscinny, Vincent Hanot, Hasna Trabelsi & Joris Van Loco (2014): Determination of caramel colourants by-products in liquid foods by ultrahigh- performance liquid chromatography- tandem mass spectrometry (UPLC-MS/MS), Food Additives & Contaminants: Part A, DOI: 10.1080/19440049.2014.940609 To link to this article: http://dx.doi.org/10.1080/19440049.2014.940609
Disclaimer: This is a version of an unedited manuscript that has been accepted for publication. As a service to authors and researchers we are providing this version of the accepted manuscript (AM). Copyediting, typesetting, and review of the resulting proof will be undertaken on this manuscript before final publication of the Version of Record (VoR). During production and pre-press, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal relate to this version also.
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Publisher: Taylor & Francis Journal: Food Additives & Contaminants: Part A DOI: 10.1080/19440049.2014.940609
Determination of caramel colourants by-products in liquid foods by Ultrahigh- Performance Liquid Chromatography- Tandem Mass
ip
1
sc r
Séverine Goscinny1*, Vincent Hanot1, Hasna Trabelsi2, and Joris Van Loco1
Department of Food, Medicines and Consumer Safety, Scientific Institute of
an u
Public Health, 1050 Brussels, Belgium. 2Department of UFR Biomedical, University of Paris Descartes, 75006 Paris, France
M
*Corresponding author: [email protected]
d
Abstract
te
2-Methylimidazole, 4-Methylimidazole (2-MI and 4-MI), 2-Acetyl-4-(1, 2, 3, 4 tetrahydroxybutyl) imidazole (THI) and 5-hydroxymethylfurfural (5-HMF) are neo-
ep
formed compounds generated during the manufacture of caramel colours and are transferred to the processed food. These contaminants are known to have a toxicological profile that may pose health risks. Hence, to characterize THI, 2-and 4-MI and 5-HMF levels in liquid foods an UPLC-MS/MS method was developed and sample preparation
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
t
Spectrometry (UPLC-MS/MS)
was divided in two analytical strategies depending on the concentration range expected in the type of foods targeted. For the determination of the imidazole substitutes (THI, 2- MI and 4-MI), a sample enrichment and clean-up step by strong cation solid-phase extraction was developed. This method is capable of quantifying over a range of 5 ng/ml (LOQ) to 500 ng/ml with recoveries of 75.4% to 112.4% and RSDs of 1.5 % to 15 %. While, for the determination of 5-HMF, a standard addition method was applied covering the linear range of 0.25 to 30 µg/ml with RSDs from 2.8 % (for intraday precision) to 9.2 % (for intermediate precision). The validated analytical methods were applied to 28 liquid food samples purchased from local markets. THI was found only in the beer samples at levels
up to 141.2 ng/ml. For 2-MI, non-quantifiable traces were observed for all the samples, while 4-MI was observed in all the samples with large concentration variations (from 0.99). Then, native content of 5-HMF is back calculated utilizing the calibration function at the intercept of the curve with the
an u
concentration axis.
Method validation
M
The compounds were defined in two distinct groups because two levels of concentration range were expected: Group 1 (from 0 to 500 ng /ml) gathered THI, 2-MI
d
and 4-MI and Group 2 (from 0 to 30 µg/ml of matrix diluted 50 fold) consisted of 5-
te
HMF. For Group 1, the validation parameters measured to evaluate the method’s
ep
performance were specificity, linearity, matrix effect, limit of quantification (LOQ), accuracy (recovery), and precision (repeatability and intermediate precision). For group 2 quantification was done using the standard addition method which inherently
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
curve. A plot of the response area against the added concentration (ng/mL) of 5-HMF is
compensates for recovery and matrix effect. As this approach requires true linearity between the two variables (e.g. instrument response, concentration), the linear range of HFM response was assessed. For both groups, dark beer (purchased in a local supermarket) was used as test material due to the complexity of the sample matrix compared to other types of beers, cola drinks and energy drinks tested during the method development.
Validation study for group 1 Matrix effect was assessed by performing a t-test between a calibration curve done in solvent and a calibration curve in matrix obtained by end spikes after SPE of a dark beer. Then, linearity was tested using the same injections results. As the test material was not free from the target compounds, the precision and recovery of the
ip
t
method was determined by comparing a sample spiked after SPE and the same sample
sc r
exercise 8 times the same day, the repeatability of the method was determined
expressed as RSD. Intermediate precision was estimated by processing one sample per
an u
level, on four non-consecutive days.
Validation study for Group 2
M
Calibration curve was prepared at various expected concentrations 0.25, 0.5, 1, 1.5, 2 and 30 µg/mL in solvent (n=3). The corresponding response area was plotted
d
against concentration level and submitted to the Mendel’s test to evaluate linearity
te
(Mendel 1964). Then, it was possible to build a protocol (dilution factors and estimation
ep
of spike levels) for the establishment of the standard addition protocol. Intraday precision was tested by performing 6 analysis (standard addition curves) of the same test sample (dark beer), and intermediate precision was evaluated following the same
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
spiked before SPE at the same level, for two levels 20 ppb and 200 ppb. Performing this
procedure performed once on 4 non consecutive days.
Real samples
Samples of beverages and sauces (13 beers, 7 cola, 6 energy drinks, 1 sherry
vinegar and 1 bouillon sauce) were purchased from local supermarkets and tested for all four analytes.
Results and discussion UPLC-MS/MS method optimization For chromatographic separation, the variability in polarity of the targeted compounds (from very polar THI to moderately polar 5-HMF) has driven the choice of column towards stationary phases offering a larger range of selectivity than
ip
t
conventional C18 reverse phase. Preliminary tests were carried out using three different
sc r
from Thermo, and Acquity HSS T3 (1.8 µm, 100 × 2.1 mm) from Waters, with mobile phases composed of water (A) and methanol with ammonia solution (B) because high
an u
pH gave best results in terms of retention and resolution. The elution time was 15 min with gradient starting at 95% of A to 5 % in 11 min, then kept for 2 min before going
M
back to initial conditions in 2 min. Hypercarb column gave the best retention for THI, nonetheless it was co-eluting with 5-HMF and long tailing was observed for 2- and 4-
d
MI peaks. HSS T3 column gave the same tailing effect for 2- and 4-MI, furthermore the
te
THI peak was distorted resulting in poor area reproducibility. YMC-TriArt column gave the best results in terms of peak symmetry, resolving power between 2- and 4-MI.
ep
Hence, further LC optimization was done with the YMC-TriArt column and the chromatographic method developed in this study allowed good separation of all the
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
columns; TriArt (1.9 µm, 100 × 2.0 mm) from YMC, Hypercarb ( 5 µm, 100 × 2.1 mm)
compounds in less than 6 min (figure 1). The optimization of the different MS/MS parameters was performed by direct
infusion of individual solutions at 1 µg/mL in water. Most intense product ion was kept for quantification and the second most abundant product ion was used for confirmation. For THI, 2- and 4-MI only two intense fragments were obtained during the optimization of the collision energy whereas for 5-HMF three intense fragments were obtained. With the use of time windows, 5-HMF was isolated which easily permitted to keep the three
transitions without sacrificing the number of points per peak nor sensitivity. During the analysis of real samples, the 5-HMF peak had a shoulder in the case of dark beers. This distortion was present for the 3 transitions which led to believe it could be a case of column overload. Therefore, the same sample was injected but the injection volume was lowered to 1 µL. The chromatographic peak response was less intense but the shoulder
ip
t
was still present. Another hypothesis for this distortion was the possible presence of an
sc r
resolve 5-HMF from the suspected interference. The new chromatographic separation
showed a peak at 3.95 min (corresponding to 5-HMF) and two other non-resolved peaks
an u
at 4.16 and 4.25 min respectively (suspected interference). The same chromatographic profile was found for the 3 transitions. Although, the triple quadrupole is not performing best in full scan, we decided to analyse the same sample and also a standard solution of
M
5-HMF in full scan mode (acquisition done in continuum mode from m/z 50 to 500)
d
with the 15 min LC run and cone voltage at 25 V to induce in source collision. Then, we
te
extracted the mass of 5-HMF (formula C6H6O3, m/z = 127) for the standard solution and the dark beer total ion chromatograms. The summed mass spectrum over the peak at
ep
3.95 min of the standard solution shows the expected profile of 5-HMF with the molecular ion (m/z=127.3) and three fragments m/z 109, 81 and in very small amount 60 (see figure 2). For the dark beer sample, the mass spectrum obtained for 5-HMF
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
interference with the same MRM transitions. The gradient was stretched up to 15 min to
(summed over 3.95 min peak) reveals the presence of an additional m/z of 104 not present in the standard solution moreover, is the most intense ion observed in the mass spectrum of the interference. The interference mass spectrum is composed of the m/z 127, 109, 81 and 60 but at ratios entirely different from 5-HMF which explains its presence in all the 3 MRM defined for 5-HMF. From these observations we can conclude that there is still a small amount of the interference (identified by m/z 104) co-
eluting with 5-HMF, therefore the 15 min run (more than double the previous optimized run) has only fixed the problem visually but the contribution of ions 127, 109, 81, and 60 from the interference will impact on the ion ratios for 5-HMF. Interestingly, m/z 109 is produced in very small amount by the interference, therefore, taking advantage of this profile difference between 5-HMF and the interference, 5-HMF was infused again to
ip
t
maximise in source dissociation (CID) in favour of the fragment m/z 109, then collision
sc r
parameters were tested with the same dark beer diluted 10 fold using the elution total run of 6 min. No chromatographic shoulder was observed which demonstrates the
an u
efficiency of the new transitions monitored for 5-HMF to eliminate the impact of the interference on the detection of 5-HMF. Hence, these acquisition parameters were kept
M
onward.
Clean-up
d
THI eluted at 1.35 min with all the very polar compounds from the matrix,
te
hence, the chromatographic peaks of THI from diluted beer samples were very hilly
ep
which were seriously hampering quantification (data not shown). Different dilution factors were tested to bypass this problem but the limit of detection got outside the expected range. Therefore, a clean-up step was considered and tested for THI, 2- and 4-
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
energies were optimized for the production of fragments m/z 81 and 60. The new
MI. The choice of SPE was driven towards strong cation exchange sorbents offering an orthogonal selectivity to the chromatography separation (C18 end-capped). Strong cation exchange clean-up is highly beneficial for the method as it involved solvent exchange and the possibility to concentrate the cleaned extract (here a two fold factor). The development of the clean-up step was straightforward as THI and 2-and 4-MI exhibit at low pH values ionic forms (with pKa values of 3.5 and 10.6 for THI and 7.7 for 2-and 4-MI). Acidifying the sample before SPE process will allow ionic interactions
to retain the compounds. The washing step was first tried with acidified water to eliminate neutral polar compounds and facilitate THI quantification by reducing compounds eluting at the beginning of the chromatographic run. The recoveries obtained by washing the SPE with the acidified water yield 20% for THI but were above 95 % for 2-and 4-MI. To avoid delicate pH control of the washing solution due to
ip
t
the pKa of THI, the wash was done with pure methanol and yield recoveries above 85%
sc r
baseline for THI. Hence, this washing solution was kept. The elution of the analytes is
ensured using an organic solvent containing high concentration of cations (e.g. NH4+),
an u
in our case methanol containing 2% of ammonia solution was used, therefore breaking the retention mechanisms. Isolute®SCX and Strata™CX were tested and the later was selected because with standard solutions the recoveries were higher. Then, further tests
M
were performed for beer matrix and the only adaptation done was to lower the sample
d
intake from 1 ml to 0.5 ml as to achieve best recoveries. This phenomenon can be
te
explained by the overload of the cartridge capacity with 1 ml of matrix sample which causes analyte losses. This clean-up approach is not extensive and produces extracts
ep
showing much more stable chromatographic baselines in addition to achieved LOQs lower than 5 ppb (matrix equivalent).
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
and 98% for THI and 2- and 4-MI, respectively with improved chromatographic
Validation results for Group 1 (THI, 2- and 4-MI) The selectivity for THI, 2- and 4-MI was assessed by processing and analysing a
golden ales beer (blonde beer) that is considered as blank because no peaks were observed at the retention times of the target compounds. The matrix effect was evaluated by constructing two calibration curves; one in solvent (water) and the second in matrix extract and performing a t-test at 95% confidence level. The calibration curves expressed in ppb (matrix equivalent taking into
account a concentration factor of 2) were built at the following levels after SPE: 0, 5, 10, 15, 20, 50, 100, 250, 500 ppb (ng/mL matrix equivalent) using the standard solutions mix-I, mix-II, and mix-III. Significant differences between the slopes of the two curves were found for the three compounds with high signal suppression for THI and light signal enhancement for 2- and 4-MI. Examples of matrix effect for THI and 4-
ip
t
MI is given in figure 3. Therefore, calibration will have to be performed using matrix-
sc r
built previously and submitted to a Mandel’s Fitting test (ref). For THI, 2- and 4-MI the linear regression model is preferred and resulting determination coefficients (r2) were
an u
higher than 0.998.
The limit of quantification was first estimated by spiking the blonde beer with decreasing amounts of analytes, then the level of quantification was set when the
M
resulting signal to noise yield 6. In these experiments LOQ for THI was fond at 1 ppb
d
and for 2- and 4-MI at 2 ppb. Nonetheless, for practical reason the LOQs were set to 5
te
ppb.
The extraction recoveries (accuracy) were determined by analysing spiked
ep
samples of dark beer at two levels: 20 and 200 ppb (n=8 on the same day and n=4 on different days per level) and comparing the result to the corresponding sample spiked after the clean-up. The recoveries ranged from 91.1 to 101.1 % for THI, 107.8 to
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
matched calibration curves. The linearity was assessed using the matrix-match curve
112.4% for 2-MI and 75.4 to 92.5% for 4-MI. The lower recoveries for 4-MI could be explained by the use of the same mixture solutions over a period of a month with multiple re-freezing process probably causing tautomerisation of 4-MI to 2-MI. As this phenomenon was not reported in the literature to confirm this hypothesis, fresh mixtures were prepared and recoveries were determined at 200 ppb with the same dark beer
sample. The resulting recoveries were 84 % for THI, 93% for 2- and 4-MI, hence mixture solutions will have to be prepared fresh every week to overcome this problem. These data were also used to determine the precision (repeatability and intermediate precision) of the method expressed as relative standard deviation (RSD) which were below 20% for all the analytes for both levels (detailed results are presented
ip
t
in table 2). The values obtained during validation, demonstrates that the method gives
an u
Validation results for Group 2 (5-HMF)
sc r
measurement for dietary intake studies.
The Mandel’s test preferred a linear over the quadratic model for the defined concentration range which enables the usage of standard addition methodology if the
M
response falls within this response range. The precision expressed as RSD’s of the concentration found was 2.8% for intraday (n=6) and 9.2% for intraday (n=4). These
te
d
data suggest that the standard addition procedure is properly set-up.
Samples analysis
ep
Twenty-eight samples were processed following the developed methods and the
amounts found are reported in table 3. The results for 2-MI were not reported because for all the samples had a response below LOQ. THI was found only in the beer samples
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
quantitative results with limits of quantification sufficiently low to enable accurate
with varying amounts ranging from 195.3a
17
10
0.150
231.2 > 213.3
17
10
0.150
Time window 1: from 0.72 to 2.00
an u
from 2.00 to 3.50 min
3.03
4-MeI
83.2 > 42.2
25
18
0.100
83.2 > 56.2 a
25
15
0.100
83.2 > 42.2 a
25
18
0.100
83.2 > 56.2
25
15
0.100
127.3 > 53.3 b
25
20
0.150
127.3 > 81.1 b
25
15
0.150
127.3 > 109.4 b
25
10
0.150
109.4 > 53.3
36
14
0.150
109.4 > 81.1 a
36
9
0.150
M
2.16
2-MeI
sc r
Time window 2:
d
Time window 3:
min
3.8
ep
5-HMF
te
from 3.50 to 4.50
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
THI
ip
t
min
3.80
a
MRM transition used for quantification,
b
These MRM transitions were first optimized for 5-HMF and were no longer used in
favour of the two new MRM transitions to overcome interferences present in dark beer.
Table 2. Validation results for Group 1 with test material dark beer
Analyte
Intraday precision
Intermediate precision
(n=8)
(n=4)
Level
RSD (average recoveries )
5-HMF
200 ppb
5.3% (96.7%)
3.0% (99.1%)
20 ppb
11.7% (112.4%)
5.9% (110.4%)
200 ppb
6.7% (107.8%)
1.5% (103.2%)
20 ppb
4.5% (79.2%)
200 ppb
13.3% (82.1%)
Standard addition
2.8%
ip
t
15% (91.1%)
sc r
4-MeI
6.9% (101.1%)
9.0% (75.4%)
10.9% (92.5%)
9.2%
Table 3. Levels of THI, 4-MI and 5-HMF found in tested samples
M
Label Sample type
THI (ng/ml)
4-MI (ng/ml)
5-HMF (µg/ml)
Food Additives & Contaminants: Part A Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/tfac20
Determination of caramel colourants by-products in liquid foods by ultrahigh- performance liquid chromatography- tandem mass spectrometry (UPLCMS/MS) a
a
b
Séverine Goscinny , Vincent Hanot , Hasna Trabelsi & Joris Van Loco
a
a
Department of Food, Medicines and Consumer Safety, Scientific Institute of Public Health, Brussels, Belgium b
Department of UFR Biomedical, University of Paris Descartes, Paris, France Accepted author version posted online: 25 Jul 2014.
To cite this article: Séverine Goscinny, Vincent Hanot, Hasna Trabelsi & Joris Van Loco (2014): Determination of caramel colourants by-products in liquid foods by ultrahigh- performance liquid chromatography- tandem mass spectrometry (UPLC-MS/MS), Food Additives & Contaminants: Part A, DOI: 10.1080/19440049.2014.940609 To link to this article: http://dx.doi.org/10.1080/19440049.2014.940609
Disclaimer: This is a version of an unedited manuscript that has been accepted for publication. As a service to authors and researchers we are providing this version of the accepted manuscript (AM). Copyediting, typesetting, and review of the resulting proof will be undertaken on this manuscript before final publication of the Version of Record (VoR). During production and pre-press, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal relate to this version also.
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Publisher: Taylor & Francis Journal: Food Additives & Contaminants: Part A DOI: 10.1080/19440049.2014.940609
Determination of caramel colourants by-products in liquid foods by Ultrahigh- Performance Liquid Chromatography- Tandem Mass
ip
1
sc r
Séverine Goscinny1*, Vincent Hanot1, Hasna Trabelsi2, and Joris Van Loco1
Department of Food, Medicines and Consumer Safety, Scientific Institute of
an u
Public Health, 1050 Brussels, Belgium. 2Department of UFR Biomedical, University of Paris Descartes, 75006 Paris, France
M
*Corresponding author: [email protected]
d
Abstract
te
2-Methylimidazole, 4-Methylimidazole (2-MI and 4-MI), 2-Acetyl-4-(1, 2, 3, 4 tetrahydroxybutyl) imidazole (THI) and 5-hydroxymethylfurfural (5-HMF) are neo-
ep
formed compounds generated during the manufacture of caramel colours and are transferred to the processed food. These contaminants are known to have a toxicological profile that may pose health risks. Hence, to characterize THI, 2-and 4-MI and 5-HMF levels in liquid foods an UPLC-MS/MS method was developed and sample preparation
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
t
Spectrometry (UPLC-MS/MS)
was divided in two analytical strategies depending on the concentration range expected in the type of foods targeted. For the determination of the imidazole substitutes (THI, 2- MI and 4-MI), a sample enrichment and clean-up step by strong cation solid-phase extraction was developed. This method is capable of quantifying over a range of 5 ng/ml (LOQ) to 500 ng/ml with recoveries of 75.4% to 112.4% and RSDs of 1.5 % to 15 %. While, for the determination of 5-HMF, a standard addition method was applied covering the linear range of 0.25 to 30 µg/ml with RSDs from 2.8 % (for intraday precision) to 9.2 % (for intermediate precision). The validated analytical methods were applied to 28 liquid food samples purchased from local markets. THI was found only in the beer samples at levels
up to 141.2 ng/ml. For 2-MI, non-quantifiable traces were observed for all the samples, while 4-MI was observed in all the samples with large concentration variations (from 0.99). Then, native content of 5-HMF is back calculated utilizing the calibration function at the intercept of the curve with the
an u
concentration axis.
Method validation
M
The compounds were defined in two distinct groups because two levels of concentration range were expected: Group 1 (from 0 to 500 ng /ml) gathered THI, 2-MI
d
and 4-MI and Group 2 (from 0 to 30 µg/ml of matrix diluted 50 fold) consisted of 5-
te
HMF. For Group 1, the validation parameters measured to evaluate the method’s
ep
performance were specificity, linearity, matrix effect, limit of quantification (LOQ), accuracy (recovery), and precision (repeatability and intermediate precision). For group 2 quantification was done using the standard addition method which inherently
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
curve. A plot of the response area against the added concentration (ng/mL) of 5-HMF is
compensates for recovery and matrix effect. As this approach requires true linearity between the two variables (e.g. instrument response, concentration), the linear range of HFM response was assessed. For both groups, dark beer (purchased in a local supermarket) was used as test material due to the complexity of the sample matrix compared to other types of beers, cola drinks and energy drinks tested during the method development.
Validation study for group 1 Matrix effect was assessed by performing a t-test between a calibration curve done in solvent and a calibration curve in matrix obtained by end spikes after SPE of a dark beer. Then, linearity was tested using the same injections results. As the test material was not free from the target compounds, the precision and recovery of the
ip
t
method was determined by comparing a sample spiked after SPE and the same sample
sc r
exercise 8 times the same day, the repeatability of the method was determined
expressed as RSD. Intermediate precision was estimated by processing one sample per
an u
level, on four non-consecutive days.
Validation study for Group 2
M
Calibration curve was prepared at various expected concentrations 0.25, 0.5, 1, 1.5, 2 and 30 µg/mL in solvent (n=3). The corresponding response area was plotted
d
against concentration level and submitted to the Mendel’s test to evaluate linearity
te
(Mendel 1964). Then, it was possible to build a protocol (dilution factors and estimation
ep
of spike levels) for the establishment of the standard addition protocol. Intraday precision was tested by performing 6 analysis (standard addition curves) of the same test sample (dark beer), and intermediate precision was evaluated following the same
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
spiked before SPE at the same level, for two levels 20 ppb and 200 ppb. Performing this
procedure performed once on 4 non consecutive days.
Real samples
Samples of beverages and sauces (13 beers, 7 cola, 6 energy drinks, 1 sherry
vinegar and 1 bouillon sauce) were purchased from local supermarkets and tested for all four analytes.
Results and discussion UPLC-MS/MS method optimization For chromatographic separation, the variability in polarity of the targeted compounds (from very polar THI to moderately polar 5-HMF) has driven the choice of column towards stationary phases offering a larger range of selectivity than
ip
t
conventional C18 reverse phase. Preliminary tests were carried out using three different
sc r
from Thermo, and Acquity HSS T3 (1.8 µm, 100 × 2.1 mm) from Waters, with mobile phases composed of water (A) and methanol with ammonia solution (B) because high
an u
pH gave best results in terms of retention and resolution. The elution time was 15 min with gradient starting at 95% of A to 5 % in 11 min, then kept for 2 min before going
M
back to initial conditions in 2 min. Hypercarb column gave the best retention for THI, nonetheless it was co-eluting with 5-HMF and long tailing was observed for 2- and 4-
d
MI peaks. HSS T3 column gave the same tailing effect for 2- and 4-MI, furthermore the
te
THI peak was distorted resulting in poor area reproducibility. YMC-TriArt column gave the best results in terms of peak symmetry, resolving power between 2- and 4-MI.
ep
Hence, further LC optimization was done with the YMC-TriArt column and the chromatographic method developed in this study allowed good separation of all the
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
columns; TriArt (1.9 µm, 100 × 2.0 mm) from YMC, Hypercarb ( 5 µm, 100 × 2.1 mm)
compounds in less than 6 min (figure 1). The optimization of the different MS/MS parameters was performed by direct
infusion of individual solutions at 1 µg/mL in water. Most intense product ion was kept for quantification and the second most abundant product ion was used for confirmation. For THI, 2- and 4-MI only two intense fragments were obtained during the optimization of the collision energy whereas for 5-HMF three intense fragments were obtained. With the use of time windows, 5-HMF was isolated which easily permitted to keep the three
transitions without sacrificing the number of points per peak nor sensitivity. During the analysis of real samples, the 5-HMF peak had a shoulder in the case of dark beers. This distortion was present for the 3 transitions which led to believe it could be a case of column overload. Therefore, the same sample was injected but the injection volume was lowered to 1 µL. The chromatographic peak response was less intense but the shoulder
ip
t
was still present. Another hypothesis for this distortion was the possible presence of an
sc r
resolve 5-HMF from the suspected interference. The new chromatographic separation
showed a peak at 3.95 min (corresponding to 5-HMF) and two other non-resolved peaks
an u
at 4.16 and 4.25 min respectively (suspected interference). The same chromatographic profile was found for the 3 transitions. Although, the triple quadrupole is not performing best in full scan, we decided to analyse the same sample and also a standard solution of
M
5-HMF in full scan mode (acquisition done in continuum mode from m/z 50 to 500)
d
with the 15 min LC run and cone voltage at 25 V to induce in source collision. Then, we
te
extracted the mass of 5-HMF (formula C6H6O3, m/z = 127) for the standard solution and the dark beer total ion chromatograms. The summed mass spectrum over the peak at
ep
3.95 min of the standard solution shows the expected profile of 5-HMF with the molecular ion (m/z=127.3) and three fragments m/z 109, 81 and in very small amount 60 (see figure 2). For the dark beer sample, the mass spectrum obtained for 5-HMF
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
interference with the same MRM transitions. The gradient was stretched up to 15 min to
(summed over 3.95 min peak) reveals the presence of an additional m/z of 104 not present in the standard solution moreover, is the most intense ion observed in the mass spectrum of the interference. The interference mass spectrum is composed of the m/z 127, 109, 81 and 60 but at ratios entirely different from 5-HMF which explains its presence in all the 3 MRM defined for 5-HMF. From these observations we can conclude that there is still a small amount of the interference (identified by m/z 104) co-
eluting with 5-HMF, therefore the 15 min run (more than double the previous optimized run) has only fixed the problem visually but the contribution of ions 127, 109, 81, and 60 from the interference will impact on the ion ratios for 5-HMF. Interestingly, m/z 109 is produced in very small amount by the interference, therefore, taking advantage of this profile difference between 5-HMF and the interference, 5-HMF was infused again to
ip
t
maximise in source dissociation (CID) in favour of the fragment m/z 109, then collision
sc r
parameters were tested with the same dark beer diluted 10 fold using the elution total run of 6 min. No chromatographic shoulder was observed which demonstrates the
an u
efficiency of the new transitions monitored for 5-HMF to eliminate the impact of the interference on the detection of 5-HMF. Hence, these acquisition parameters were kept
M
onward.
Clean-up
d
THI eluted at 1.35 min with all the very polar compounds from the matrix,
te
hence, the chromatographic peaks of THI from diluted beer samples were very hilly
ep
which were seriously hampering quantification (data not shown). Different dilution factors were tested to bypass this problem but the limit of detection got outside the expected range. Therefore, a clean-up step was considered and tested for THI, 2- and 4-
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
energies were optimized for the production of fragments m/z 81 and 60. The new
MI. The choice of SPE was driven towards strong cation exchange sorbents offering an orthogonal selectivity to the chromatography separation (C18 end-capped). Strong cation exchange clean-up is highly beneficial for the method as it involved solvent exchange and the possibility to concentrate the cleaned extract (here a two fold factor). The development of the clean-up step was straightforward as THI and 2-and 4-MI exhibit at low pH values ionic forms (with pKa values of 3.5 and 10.6 for THI and 7.7 for 2-and 4-MI). Acidifying the sample before SPE process will allow ionic interactions
to retain the compounds. The washing step was first tried with acidified water to eliminate neutral polar compounds and facilitate THI quantification by reducing compounds eluting at the beginning of the chromatographic run. The recoveries obtained by washing the SPE with the acidified water yield 20% for THI but were above 95 % for 2-and 4-MI. To avoid delicate pH control of the washing solution due to
ip
t
the pKa of THI, the wash was done with pure methanol and yield recoveries above 85%
sc r
baseline for THI. Hence, this washing solution was kept. The elution of the analytes is
ensured using an organic solvent containing high concentration of cations (e.g. NH4+),
an u
in our case methanol containing 2% of ammonia solution was used, therefore breaking the retention mechanisms. Isolute®SCX and Strata™CX were tested and the later was selected because with standard solutions the recoveries were higher. Then, further tests
M
were performed for beer matrix and the only adaptation done was to lower the sample
d
intake from 1 ml to 0.5 ml as to achieve best recoveries. This phenomenon can be
te
explained by the overload of the cartridge capacity with 1 ml of matrix sample which causes analyte losses. This clean-up approach is not extensive and produces extracts
ep
showing much more stable chromatographic baselines in addition to achieved LOQs lower than 5 ppb (matrix equivalent).
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
and 98% for THI and 2- and 4-MI, respectively with improved chromatographic
Validation results for Group 1 (THI, 2- and 4-MI) The selectivity for THI, 2- and 4-MI was assessed by processing and analysing a
golden ales beer (blonde beer) that is considered as blank because no peaks were observed at the retention times of the target compounds. The matrix effect was evaluated by constructing two calibration curves; one in solvent (water) and the second in matrix extract and performing a t-test at 95% confidence level. The calibration curves expressed in ppb (matrix equivalent taking into
account a concentration factor of 2) were built at the following levels after SPE: 0, 5, 10, 15, 20, 50, 100, 250, 500 ppb (ng/mL matrix equivalent) using the standard solutions mix-I, mix-II, and mix-III. Significant differences between the slopes of the two curves were found for the three compounds with high signal suppression for THI and light signal enhancement for 2- and 4-MI. Examples of matrix effect for THI and 4-
ip
t
MI is given in figure 3. Therefore, calibration will have to be performed using matrix-
sc r
built previously and submitted to a Mandel’s Fitting test (ref). For THI, 2- and 4-MI the linear regression model is preferred and resulting determination coefficients (r2) were
an u
higher than 0.998.
The limit of quantification was first estimated by spiking the blonde beer with decreasing amounts of analytes, then the level of quantification was set when the
M
resulting signal to noise yield 6. In these experiments LOQ for THI was fond at 1 ppb
d
and for 2- and 4-MI at 2 ppb. Nonetheless, for practical reason the LOQs were set to 5
te
ppb.
The extraction recoveries (accuracy) were determined by analysing spiked
ep
samples of dark beer at two levels: 20 and 200 ppb (n=8 on the same day and n=4 on different days per level) and comparing the result to the corresponding sample spiked after the clean-up. The recoveries ranged from 91.1 to 101.1 % for THI, 107.8 to
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
matched calibration curves. The linearity was assessed using the matrix-match curve
112.4% for 2-MI and 75.4 to 92.5% for 4-MI. The lower recoveries for 4-MI could be explained by the use of the same mixture solutions over a period of a month with multiple re-freezing process probably causing tautomerisation of 4-MI to 2-MI. As this phenomenon was not reported in the literature to confirm this hypothesis, fresh mixtures were prepared and recoveries were determined at 200 ppb with the same dark beer
sample. The resulting recoveries were 84 % for THI, 93% for 2- and 4-MI, hence mixture solutions will have to be prepared fresh every week to overcome this problem. These data were also used to determine the precision (repeatability and intermediate precision) of the method expressed as relative standard deviation (RSD) which were below 20% for all the analytes for both levels (detailed results are presented
ip
t
in table 2). The values obtained during validation, demonstrates that the method gives
an u
Validation results for Group 2 (5-HMF)
sc r
measurement for dietary intake studies.
The Mandel’s test preferred a linear over the quadratic model for the defined concentration range which enables the usage of standard addition methodology if the
M
response falls within this response range. The precision expressed as RSD’s of the concentration found was 2.8% for intraday (n=6) and 9.2% for intraday (n=4). These
te
d
data suggest that the standard addition procedure is properly set-up.
Samples analysis
ep
Twenty-eight samples were processed following the developed methods and the
amounts found are reported in table 3. The results for 2-MI were not reported because for all the samples had a response below LOQ. THI was found only in the beer samples
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
quantitative results with limits of quantification sufficiently low to enable accurate
with varying amounts ranging from 195.3a
17
10
0.150
231.2 > 213.3
17
10
0.150
Time window 1: from 0.72 to 2.00
an u
from 2.00 to 3.50 min
3.03
4-MeI
83.2 > 42.2
25
18
0.100
83.2 > 56.2 a
25
15
0.100
83.2 > 42.2 a
25
18
0.100
83.2 > 56.2
25
15
0.100
127.3 > 53.3 b
25
20
0.150
127.3 > 81.1 b
25
15
0.150
127.3 > 109.4 b
25
10
0.150
109.4 > 53.3
36
14
0.150
109.4 > 81.1 a
36
9
0.150
M
2.16
2-MeI
sc r
Time window 2:
d
Time window 3:
min
3.8
ep
5-HMF
te
from 3.50 to 4.50
Ac c
Downloaded by [University of Tennessee, Knoxville] at 01:52 22 August 2014
THI
ip
t
min
3.80
a
MRM transition used for quantification,
b
These MRM transitions were first optimized for 5-HMF and were no longer used in
favour of the two new MRM transitions to overcome interferences present in dark beer.
Table 2. Validation results for Group 1 with test material dark beer
Analyte
Intraday precision
Intermediate precision
(n=8)
(n=4)
Level
RSD (average recoveries )
5-HMF
200 ppb
5.3% (96.7%)
3.0% (99.1%)
20 ppb
11.7% (112.4%)
5.9% (110.4%)
200 ppb
6.7% (107.8%)
1.5% (103.2%)
20 ppb
4.5% (79.2%)
200 ppb
13.3% (82.1%)
Standard addition
2.8%
ip
t
15% (91.1%)
sc r
4-MeI
6.9% (101.1%)
9.0% (75.4%)
10.9% (92.5%)
9.2%
Table 3. Levels of THI, 4-MI and 5-HMF found in tested samples
M
Label Sample type
THI (ng/ml)
4-MI (ng/ml)
5-HMF (µg/ml)
