Article pubs.acs.org/JAFC
Analysis of Potential Migrants from Plastic Materials in Milk by Liquid Chromatography−Mass Spectrometry with Liquid−Liquid Extraction and Low-Temperature Purification Zsolt Bodai,† Bálint Sámuel Szabó,† Márton Novák,† Susanne Hámori,† Zoltán Nyiri,† Tamás Rikker,§ and Zsuzsanna Eke*,†,§ †
Joint Research and Training Laboratory on Separation Techniques (EKOL), Eötvös Loránd University, 1/A Pázmány Péter sétány, Budapest 1117, Hungary § Wessling International Research and Educational Center, 56 Fóti út, Budapest 1047, Hungary S Supporting Information *
ABSTRACT: A simple and fast analytical method was developed for the determination of six UV stabilizers (Cyasorb UV-1164, Tinuvin P, Tinuvin 234, Tinuvin 326, Tinuvin 327, and Tinuvin 1577) and five antioxidants (Irgafos 168, Irganox 1010, Irganox 3114, Irganox 3790, and Irganox 565) in milk. For sample preparation liquid−liquid extraction with low-temperature purification combined with centrifugation was used to remove fats, proteins, and sugars. After the cleanup step, the sample was analyzed with high-performance liquid chromatography−tandem mass spectrometry (LC-MS/MS). External standard and matrix calibrations were tested. External calibration proved to be acceptable for Tinuvin P, Tinuvin 234, Tinuvin 326, Tinuvin 327, Irganox 3114, and Irganox 3790. The method was successfully validated with matrix calibration for all compounds. Method detection limits were between 0.25 and 10 μg/kg. Accuracies ranged from 93 to 109%, and intraday precisions were 100%, there is a signal enhancement, and if ME is 50% for all compounds with every solvent. The lowest recoveries were observed with acetonitrile and 2propanol. Acetone, ethyl acetate, and tetrahydrofuran provided acceptable recoveries. Using acetone, the recoveries were >87% except for Irganox 565 (68%) and Tinuvin 1577 (77%). Extractions with ethyl acetate resulted 82% recoveries for all compounds. Thus, tetrahydrofuran was chosen for further studies.
advisable for some of the target compounds, mostly for the antioxidants, ESI was preferred in our method because it grants satisfying intensity for all compounds. The first tests of columns and eluents showed that 0.1 v/v % formic acid in Direct-Q water and acetonitrile cannot elute Irganox 565 and Cyasorb UV-1164 from the GL Sciences InertSustain C18, YMC-UltraHT Hydrosphere C18, and YMCTriart C18 columns. From the Kinetex C18 only Irganox 565 could not be eluted. Such problems did not occur with the Phenomenex Synergi 4u Polar-RP 80A, Thermo Hypersil Gold C18, and Kinetex PFP columns. On the Synergi 4u Polar-RP 80A and on the Thermo Hypersil Gold C18 the last peaks broadened severely compared to the Kinetex PFP. Kinetex PFP is made using core−shell technology that guaranteed high plate number and effective separation and therefore better signal-tonoise ratio compared to a fully porous column. Kintex PFP was chosen for further experiments. These initial experiments also revealed that with regard to sensitivity Tinuvin 327, Tinuvin 326, and Tinuvin P are the critical compounds. Using methanol−water instead of acetonitrile−water granted higher abundance for most compounds but also higher back pressure and longer retention time. Methanol is a weaker eluent compared to acetonitrile, but it increased the signal so methanol was used for further experiments. Effects of the pH of the eluent were studied with pH 2.80 ammonium formate (20 mmol/L) and pH 5.00 ammonium acetate (20 mmol/L) in water. Lower pH granted higher abundance for all of the compounds. The increment of the signal of Tinuvin P was the lowest: it gave only a 1.14 times bigger peak at pH 2.8 compared to that at pH 5.00. The highest 10032
dx.doi.org/10.1021/jf503110v | J. Agric. Food Chem. 2014, 62, 10028−10037
Journal of Agricultural and Food Chemistry
Article
Figure 3. Recoveries for liquid−liquid extraction with low-temperature purification at different pH values using tetrahydrofuran on 1.5% fat milk. Error bars represent the mean ± standard deviation of recoveries calculated using average peak areas of the standard in the sample spiked after the extraction, n = 3.
Figure 4. Matrix effects using liquid−liquid extraction with low-temperature purification on 1.5% fat milk with different injection volumes. Error bars represent the mean ± standard deviation of matrix effects calculated using average peak areas of the standard solution, n = 3.
10033
dx.doi.org/10.1021/jf503110v | J. Agric. Food Chem. 2014, 62, 10028−10037
Journal of Agricultural and Food Chemistry
Article
Table 3. Validation Results Using External Calibration accuracy (RSD%) 1.5% fat
a
3.5% fat
analyte
500 μg/kg
5 mg/kg
500 μg/kg
5 mg/kg
Cyasorb UV-1164 Irgafos 168 Irganox 1010 Irganox 3114 Irganox 3790 Irganox 565 Tinuvin 1577 Tinuvin 234 Tinuvin 326 Tinuvin 327 Tinuvin P
125 (8) 64 (4) 100 (9) 98 (11) 100 (5) 133(6) 129 (7) 90 (8) 97 (3) 80 (12) 97 (1)
148 67 114 97 95 127 136 91 92 75 97
117 67 113 96 96 129 135 89 98 81 94
152 68 126 94 96 127 118 94 95 78 95
(16) (5) (12) (8) (5) (9) (11) (8) (7) (9) (5)
(9) (7) (7) (12) (4) (12) (14) (13) (5) (12) (3)
(17) (12) (7) (10) (4) (10) (16) (7) (9) (10) (5)
determination coefficient
MDLa (μg/kg)
0.9972 0.9974 0.9926 0.9920 0.9982 0.9904 0.9942 0.9928 0.9972 0.9966 0.9986
2.5 5 5 2.5 2.5 2.5 5 2.5 25 25 25
MDL, method detection limits.
Matrix Effect. Measurement error originating from ion suppression/enhancement can be eliminated by using internal standards or matrix calibration. Isotope-labeled internal standards offer an expensive way to decrease the matrix effects; moreover, for our target compounds we could not find any in commercial trade. Nonlabeled internal standards ought to resemble the corresponding target compound regarding ionization efficiency, functional groups, polarity, and retention time. Finding suitable nonlabeled internal standards that meet all of the mentioned criteria is almost impossible. On the other hand, matrix calibration is a time-consuming technique, so after minimization of the matrix effect, it is worthwhile to test external calibration first. Instead of dilution lower matrix effect can be obtained by applying lower injection volumes; 1, 2, and 5 μL of injected volume were tested (Figure 4). Significant ion enhancement was experienced in the cases of Irganox 565, Cyasorb UV-1164, and Tinuvin 1577. By contrast, ion suppression took place in the case of Tinuvin 234. With 1 μL injection volume, the matrix effect could be decreased to 116% in the case of Tinuvin 1577, to 135% in the case of Irganox 565, and to 111% in the case of Cyasorb UV-1164. For the remaining compounds decreasing the matrix amount (i.e., injection volume) did not result in remarkable changes. One microliter injection volume was chosen. Method Validation. Specificity was found to be satisfactory without any chromatographic interference at the retention time of the compounds. MRM chromatograms for negative milk and spiked negative milk samples are shown in the Supporting Information (Figures S3−S13). External Calibration. Validation data obtained with external calibration are summarized in Table 3. Calibration curves ranged from 125 μg/kg to 25 mg/kg. The lowest determination coefficient was 0.9904 (Irganox 565). Parameters of the calibration curves are given in the Supporting Information (Table S1). The MDLs were in the range from 2.5 to 25 μg/kg. These detection limits are at least 1 order of magnitude lower than the corresponding SMLs. Using external calibration, no significant effect of the fat content of the milk samples on accuracy and precision was observed. Good accuracy and precision values were obtained for Tinuvin P, Tinuvin 326, Irganox 3114, Irganox 3790, and Tinuvin 234 with both types of milk samples (1.5 and 3.5% fat contents). In this group the lowest accuracy was 89%, whereas
The volume of tetrahydrofuran necessary for the extraction was tested at 2, 5, and 10 mL. Two milliliters of extraction solvent granted good recoveries for most of the compounds. (See Figure 2.) The recoveries were >80% except for Tinuvin 234. Increasing the extraction volume from 2 to 5 mL increased the recoveries, especially for Tinuvin 234, Cyanosorb-UV 1164, Irganox 3114, Irganox 1010, and Irganox 3790. With 5 mL of tetrahydrofuran the average recovery was 92%, and with 10 mL it was 94%. Anyhow, using 10 mL instead of 5 mL did not result in such a big increase in recoveries that it would be beneficial to use that much solvent with regard to environmental and financial reasons. Hence, 5 mL was used in the process of validation. Because the target compounds have various functional groups, optimization of the pH of the extraction was indispensable. At first extraction from water, 0.1 v/v % formic acid, and 0.1 v/v % ammonia solutions to acetonitrile as the organic phase were tested. Without pH modification lower recoveries were reached than with either formic acid or ammonia solution. Acidic condition provided the best recoveries for Irganox 3790, Irganox 3114, and Tinuvin 326. In this medium recoveries were 75% with ammonia solution for every compound. Because an alkaline environment gave acceptable recoveries for the biggest number of target compounds, milk samples were set to pH 9, 11, and 13 with NaOH (Figure 3). At pH 13 recovery for Tinuvin 234 was 77%. At pH 9 it was 73% as well as for Irgafos 168. pH 11 was found to be the most suitable for the extraction of most analytes: it provided >84% recoveries for all of the compounds. For further studies pH 11 was applied. On the basis of these results 50 μL of a 2 × 10−2 mol/L NaOH solution was added to milk samples to adjust the pH for extraction. To reach the highest recoveries, 5 mL of tetrahydrofuran was used for the extraction. The efficiency of low-temperature purification was tested by measuring the dry residue weight of the tetrahydrofuran extracts before and after the purification. In the case of 1.5% fat milk, the dry residue of a given amount of extract before lowtemperature purification is 1.67 mg; that of the same amount of extract after low-temperature purification is 1.39, so the decrease (0.28) is 16.7% of the original dry residue weight (1.67). 10034
dx.doi.org/10.1021/jf503110v | J. Agric. Food Chem. 2014, 62, 10028−10037
Journal of Agricultural and Food Chemistry
Article
Table 4. Validation Results Using Matrix Calibration accuracy (RSD%) 1.5% fat analyte Cyasorb UV-1164 Irgafos 168 Irganox 1010 Irganox 3114 Irganox 3790 Irganox 565 Tinuvin 1577 Tinuvin 234 Tinuvin 326 Tinuvin 327 Tinuvin P a
3.5% fat
25 μg/kg
100 μg/kg
500 μg/kg
1 mg/kg
93 95 100 109 108 107 94 95 97 103 95
98 103 96 106 109 104 100 102 101 101 97
102 99 100 95 94 96 102 98 100 99 98
99 100 100 102 102 101 99 101 100 100 101
(3) (3) (7) (3) (6) (1) (5) (2) (8) (2) (11)
(6) (2) (11) (8) (11) (2) (3) (8) (1) (4) (3)
(7) (8) (13) (11) (13) (5) (1) (2) (3) (5) (4)
(4) (2) (4) (10) (4) (4) (9) (3) (4) (4) (5)
2.5 mg/kg 100 100 100 100 100 100 100 100 100 100 100
(3) (2) (6) (9) (2) (3) (4) (3) (3) (2) (5)
100 μg/kg
1 mg/kg
82 92 98 108 92 93 91 100 87 107 94
93 100 99 92 87 95 101 100 101 102 101
(4) (0.4) (2) (12) (12) (2) (2) (4) (12) (2) (6)
determination coefficient
MDLa (μg/kg)
0.9988 0.9992 0.9958 0.9914 0.9980 0.9990 0.9978 0.9992 0.9990 0.9990 0.9980
0.25 2.5 0.25 0.5 0.5 0.25 0.25 0.25 2.5 10 2.5
(2) (1) (5) (12) (7) (3) (2) (3) (3) (1) (4)
MDL, method detection limits.
(Cyasorb UV-1164, Irgafos 168, Irganox 1010, Irganox 3114, Irganox 3790, Irganox 565, Tinuvin 1577, Tinuvin 234, Tinuvin 326, Tinuvin 327, and Tinuvin P) in milk samples. With external calibration it can be used for the determination of Irganox 3114, Irganox 3790, Tinuvin 234, Tinuvin 326, Tinuvin 327, and Tinuvin P, whereas it can be used for the screening of Irganox 565, Cyasorb UV-1164, Tinuvin 1577, Irganox 1010, and Irgafos 168.
the highest was 100%. For these components the highest RSD was 12% (for Irganox 3114). For Tinuvin 327 the accuracies were between 75 and 81% and the RSD values between 10 and 12%. On the basis of the goal of the measurement, these values can also be acceptable. For Irganox 565, Cyasorb UV-1164, Tinuvin 1577, and Irganox 1010 the accuracies were >110% due to matrix effects. For Irgafos 168 the accuracy was
