2272
J. Sep. Sci. 2015, 38, 2272–2277
Yao Weijun
Research Article
Agilent Technologies (Shanghai) Co. Ltd., Shanghai 200131, China
Direct determination of acrylamide in food by gas chromatography with nitrogen chemiluminescence detection
Received January 19, 2015 Revised April 8, 2015 Accepted April 9, 2015
A method of gas chromatography with nitrogen chemiluminescence detection and using standard addition is described for the determination of acrylamide in heat-processed foods. Using a modified QuEChERS (quick, easy, cheap, effective, rugged, and safe) sample preparation method removes the acrylamide precursors completely, and the risk of overestimating acrylamide concentration due to additional analyte formation in the hot gas chromatograph inlet is also avoided. Sample preparation is rapid and inexpensive. A Deans switch device is utilized to heart-cut acrylamide and to prevent interferences from the solvent and matrix from reaching the detector. The pre-column is backflushed at high temperature to maintain a clean baseline and shorten the cycle time compared to baking out the column. Quantitation using standard addition is employed for compensation of potential variability in the acrylamide extraction efficiency in acetonitrile. The limit of detection and the limit of the quantification obtained for this method are 27 and 81 g/kg, respectively, in food samples (equivalent to 3.5 and 10.6 g/L in acetonitrile, respectively), and the linear range is 76– 9697 g/kg in food samples (equivalent to 10–1280 g/L in acetonitrile) with an R2 value of 0.9999. Keywords: Acrylamide / Gas chromatography / Heart cutting / Large-volume injection / Nitrogen chemiluminescence detection / Standard addition method DOI 10.1002/jssc.201500060
1 Introduction The 2002 discovery of acrylamide occurring in various heat processed starch-rich foods at concentrations exceeding those suggested for drinking water [1] led analytical chemists to explore reliable analytical methods for the determination of this potentially carcinogenic contaminant in a wide range of food matrices [2–6]. At present, the majority of methods are based on a mass spectrometer as a detection system coupled to LC [7–16] or GC [16–27]. Assays employing GC–MS techniques are either based on bromination derivatives or direct analysis without derivatization. The approach without derivatization is less laborious, but it is accompanied by problems of potential interfering co-extractives and the lower solubility of acrylamide in organic solvents in comparison to water during the sample preparation [2, 6, 28, 29]. LC–MS methods are attractive because of simple sample preparation, and the
Correspondence: Agilent Technologies (Shanghai) Co. Ltd., 412 Yinglun Road, Waigaoqiao Free Trade Zone, Shanghai, 200131, China Fax: +86 21 5048 2656 Email: [email protected]
Abbreviations: MMI, multi-mode inlet; NCD, nitrogen chemiluminescence detector; PCM, pressure control module; PSA, primary secondary amine
C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
performance of the LC–MS/MS method is found to be superior to that of the GC–MS method [10]. The issue in direct GC determination of free acrylamide arises from the same mechanism of the acrylamide formation in heated foods—the Maillard reaction. If the precursors (asparagine and reducing sugars) are not completely removed from the extract, additional acrylamide forms in the hot GC inlet, resulting in an overestimation of the native acrylamide content [29–31]. The novel sample preparation method developed by Mastovska and Lehotay overcomes this problem by the combination of MeCN extraction and primary secondary amine (PSA) cleanup. The usage of MeCN as an extraction solvent avoids the transfer of acrylamide precursors due to the low solubility of these compounds in this organic solvent. Additionally, the primary secondary amine sorbent strongly retains residual asparagine if it is present in the MeCN extract after the partition step. However, the overall extraction efficiency of acrylamide is reported to be over 60% [29]. The nitrogen chemiluminescence detector (NCD) has high sensitivity and selectivity to nitrogen-containing compounds such as acrylamide and MeCN. The nitrogencontaining compounds are oxidatively combusted at approximately 900⬚C for the complete conversion of chemically bound nitrogen into nitric oxide (R–N+O2 → CO2 + H2 O+NO), and a subsequent reaction with ozone to derive chemiluminescence (NO+O3 →NO2 * →NO + hv) [32].
www.jss-journal.com
J. Sep. Sci. 2015, 38, 2272–2277
At present, there is no report on the analysis of acrylamide in food matrices by nitrogen chemiluminescence detector.
2 Materials and methods 2.1 Materials Acrylamide (99.5%) was purchased from labor Dr. Ehrenstorfer (Augsburg, Germany). The stock solutions of acrylamide were prepared in MeCN and were stored at 4⬚C when not in use. The primary secondary amine sorbent was purchased from Agilent Technologies (Delaware, USA). The MeCN and n-hexane were supplied by Merck (Darmstadt, Germany); the deionized water was obtained from a DI apparatus (Millipore Corporation, MA, USA); the MgSO4 and NaCl were of analytical grade. The food samples (dry and unheated potato, French fries and potato chips) were purchased from local supermarkets.
2.2 Sample preparation The food samples were homogenized by a laboratory blender. Two grams of a representative sample were weighed into a 50 mL centrifuge tube, and 5 mL of hexane, 10 mL of deionized water and 10 mL of MeCN were added by using solvent dispensers. After the addition of 4 g of anhydrous MgSO4 and 0.5 g of NaCl, the tube was immediately sealed and was vigorously shaken for 1 min to avoid the formation of crystalline agglomerates and to ensure the sufficient solvent interaction with the sample. Water facilitated the extraction of acrylamide; hexane was served for sample defatting, and the salt combination induced the separation of the water and acetonitrile layers and forced the majority of acrylamide into the acetonitrile layer [29]. The sample tube was centrifuged for 5 min at 4000 rpm using a T.D.L-5000bR centrifuge (Anke, Shanghai, China), and the three layers were clearly separated. The fat was extracted into the upper hexane phase. The solids and salts were left in the lower water layer. A 1 mL volume of the middle MeCN phase was transferred into a centrifuge tube containing 50 mg of primary secondary amine and 150 mg of anhydrous MgSO4 . This suspension was mixed using a Vortex mixer for 30 s and centrifuged at 4000 rpm for 1 min. With no other treatment, the supernatant was placed into an autosampler vial for GC–NCD analysis. To determine the acrylamide present using the method of standard addition, two additional 2 g subsamples of every sample were also analyzed as above, but first each subsample was spiked with acrylamide using either 70 or 140 L of the 31 ng/L acrylamide standard.
2.3. Instrumentation The GC–NCD analyses were performed on an Agilent 7890A gas chromatograph, equipped with a multi-mode inlet (MMI), C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Gas Chromatography
2273
a Deans switch device (CFT), and electronic pressure control module (PCM), coupled to an Agilent 255 model nitrogen chemiluminescence detector. Fig. 1 shows a typical configuration. The retention gap is a 0.5 m × 0.53 mm deactivated fused-silica column. Analytical separation was performed in a 5 m × 0.53 mm × 1 m DBWaxEtr pre-column and a 15 m × 0.53 mm x 1 m DBWaxEtr analytical column from Agilent. The sample introduction was performed by an Agilent automatic liquid injector (model 7683B).
2.4 GC–NCD operation conditions The sample injection volume was 20 L (5 L × 4 times, normal injection speed, total injection time was around 30 s, splitless mode, 3 min purge time). The MMI inlet temperature was 70⬚C (1 min) and was increased by 600⬚C/min to 300⬚C. The oven temperature started at 60⬚C (1 min), was increased by 25⬚C/min to 150⬚C, and then held for 2.5 min (a total of 6.5 min). The column was configured with dimensions of 20 m × 0.53 mm × 1 m with the MMI inlet as the column inlet pressure source and the other (atmospheric pressure) as the outlet. The flow rate was set to 20 mL/min of He in constant flow mode. The initial source pressure was read at 9.1 psig. The switching flow from pressure control module started at 45 mL/min (1 min) and was increased by 2 mL/min to 40 mL/min. The Deans switch valve was cycled from its initial state of OFF to ON at 4.05 min, and then OFF again at 4.18 min per the elution time of a known standard as explained below. The post run was 3 min, with the oven at 240⬚C, the switching flow from the pressure control module at 60 mL/min, and the inlet pressure 2 psi. The nitrogen chemiluminescence detector conditions were as follows: 930⬚C burner temperature, 5 mL/min of hydrogen and 10 mL/min of oxygen.
3 Results and discussion 3.1 Method development As previously described, the combination of MeCN extraction and primary secondary amine cleanup is an efficient sample preparation method for this approach of direct GC injection because it removes the acrylamide precursors completely (acrylamide was not detected in the blank matrix of dry and unheated potato samples, Fig. 2). However, MeCN contains nitrogen and we observed that a large amount of MeCN solvent would compromise the nitrogen chemiluminescence detector stability and cause solvent peak tailing, which would interfere with the acrylamide determination. The solvent exchange to methanol by evaporation was attempted, but the data showed that the recovery was not reproducible. Therefore, the approach using a Deans switching device to heart-cut the target acrylamide and a much smaller amount of co-eluted www.jss-journal.com
2274
Y. Weijun
J. Sep. Sci. 2015, 38, 2272–2277
Figure 1. Schematic diagram of the GC–NCD system equipped with Deans switching. (A) bypass mode: elutes from the pre-column are swept to the vent; (B) heart-cutting mode: the target acrylamide is heart-cut to the analytical column for further separation; (C) backflush mode: the P1 pressure is reduced to 2 psig which is much lower than P2, and then the flow direction in pre-column is reversed; the residue in the pre-column is backflushed to the split vent under a high column temperature.
solvent into a second analytical column was a preferred solution [33, 34]. The valveless technique of the Deans switch performs flow switching by setting the appropriate pressure levels in a network of flow resistors, and the pressures and flows are controlled by valves and flow sources outside of the analytical flow path. The ability to switch the flow from the pressure control model needs to be set up appropriately so that the pre-column effluent could be completely purged to vent the analytical column by changing the three-way valve’s state. The low switching flow rate leads to an incomplete transfer and acts as a splitter [35], while the high switching flow rate raises the pre-column’s outlet pressure, causing the pre-column’s actual flow rate to become low and the peaks to become flattened. The GC manufacturer recommends the value of 2–3 times the primary flow, so a switching flow rate C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
of 45 mL/min was chosen. The restrictor’s dimension was calculated from the Agilent Dean Switch Calculator (Version A.01.01). The experimental results showed that a low level of solvent worked well under the software recommended conditions, but a large amount of solvent partially entered into the analytical column while the valve was in the OFF state. A practical approach is to cut the restrictor’s length step by step until the solvent is purged to vent completely in the OFF state. Careful timing of the Deans switching valve resulted in a stable baseline and the complete separation of real samples (Figs. 2 and 3). The correct timing of the valve switching time was determined by iteration as follows. First, a wide heart-cutting time between 3.0 min (ON) and 5.0 min (OFF) was chosen to ensure that the acrylamide standard was completely eluted and its response area was noted (as A0 ). The www.jss-journal.com
Gas Chromatography
J. Sep. Sci. 2015, 38, 2272–2277
2275
Table 1. Linearity results of using the calibration standards y = 6.746x R2 = 0.9999
Figure 2. The overlaid chromatograms of unheated potato, French fries, potato chips and spiked potato chips in ascending order of the acrylamide concentration.
Figure 3. The overlaid chromatograms of the 10, 20 and 40 g/L acrylamide standards in MeCN.
observed chromatographic peak width was less than 0.2 min, so a narrow heart-cutting window of 0.2 min was applied to the later tuning steps. Next, a second heart-cutting timing of 3.0 min (ON) and 3.2 min (OFF) was tried and compared with the earlier acrylamide area response (A0). By repeatedly increasing the heart-cut time by 0.2 min and comparing the acrylamide response area against the original value (A0 ), the best cutting window was obtained. The final heart-cut time window was optimized with 0.01 min increments to further maximize the selection for acrylamide against other interferences (such as residual solvent). It is worth mentioning that the presence of an additional detector (FID, NPD, etc.) where the vent is indicated in Figure 1 would have eliminated the need for iterating the collection window. A large-volume injection (LVI) shows a great advantage in modern GC analysis [36–38]. The increased sample injection volume (20 L) improved the acrylamide detection limits, and it eliminated the need for extra concentration steps that can otherwise cause analyte loss. The MMI was operated in splitless mode, rather than the commonly used solvent vent
C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Acrylamide in MeCN (g/l)
Relative response factor to 1280 g/l standard
RSD% (n = 6)
10 20 40 80 160 320 640 1280
0.98 0.96 0.98 0.97 1.00 0.97 1.00 1.00
4.3 3.6 1.7 1.2 1.2 1.2 1.2 0.6
mode. We observed that the response was approximately 10% less in solvent vent mode. This could be due to a loss of acrylamide from the split vent due to solvent evaporation. The sample injection ended at 30 s. The solvent was vaporized in the inlet liner and retention gap immediately, and its vapor passed to pre-column and vented. The majority of acrylamide was held in the inlet liner. Still, a minority of acrylamide was carried out of the liner by the solvent and was trapped in retention gap under the initial oven temperature. The halfmeter retention gap minimized the band broadening caused by the solvent flooding into the column head [36]. The later eluting contaminants within a matrix could cause several problems, such as ghost peaks, baseline perturbations, high background, and shifting retention times. It is common to extend the temperature program and bakeout period to remove the retained sample components and column contaminants. A significant reduction in baking time is readily achieved when a valve or Tee is used to reverse the column flow [39]. This feature is combined in Deans switch and made backflushing the pre-column a routine practice. In the post run stage, the MMI inlet’s head pressure P1 was reduced to a low value (2 psi), while the exit pressure of the pre-column P2 was increased to a high value (approximately 20 psi). The flow direction in the pre-column was reversed. The residue in the pre-column was backflushed to a split vent at a high column temperature (240⬚C). Approximately, a 30% reduction in the total cycle time was achieved versus the usual approach of an extended high-temperature temperature bakeout period to remove the retained extraneous sample components. Here, the column configuration was different from the common Tee setting, where the mid-point flow source was set as the outlet of the pre-column and inlet of the analytical column. Unlike the negligible resistance between the pressure control module and mid-point of the Tee, the resistance of the CFT plate’s inner channel and narrow tube connecting the pressure control module and CFT plate were variable, and the actual mid-point pressure was lower than the pressure control module value and was dependent on the hardware. The current setting of the MMI inlet as the column inlet pressure source and the other as the outlet showed a high consistency of retention times.
www.jss-journal.com
2276
J. Sep. Sci. 2015, 38, 2272–2277
Y. Weijun
Table 2. The recovery yields of acrylamide obtained from the GC–NCD method
Acrylamide in MeCN (g/L)
Recovery (%)
Sample material
before
217 g/L
434 g/L
217 g/L
434 g/L
Acrylamide in sample
DI water (blank) dry potato French fries potato chips
spiking 0 0 58 93
spiked 143 144 201 243
spiked 278 281 336 374
spiked 66 66 66 69
spiked 64 65 64 65
(g/kg) < LOD < LOD 441 723
A common way of evaluating the recovery rate is by the spiking of internal standards. The isotope-labeled internal standards, such as deuterium-labeled (2 H3 -Acrylamide or d3 -Acrylamide) or carbon-labeled (13 C3 -Acrylamide), are the most ideal internal standards, but they can only be used in MS-based analysis and are inappropriate for nitrogen chemiluminescence detector. Non-isotope-labeled internal standards such as methylacrylamide, N,N-dimethylacrylamide, and acetamide are widely used also; however, satisfactory repeatability might not be achieved due to the differing stability and structure of the compounds [6]. A standard addition calibration is appropriate when there is no blank matrix to carry out the calibration [40–42]. Known amounts of analyte are introduced into the sample and are treated along with the sample. A one-point calibration was used for the quantification process [40], and the sample amount was obtained using the equation: Csample = Cadded × Ssample / Ssample+added − Ssample
(1)
where Csample was the initial concentration in the sample, Cadded was the added concentration to the sample, Ssample was the peak area of the sample, and Ssample+added was the peak area that corresponds to the sample with the added concentration. The recovery rate was calculated as follows: Recovery% = Ssample+added −Ssample /Sadded × 100 where, S’ added was the calculated peak area of standard added in MeCN from the external standard calibration curve.
3.2. Method assessment 3.2.1. Linearity The linearity of the method was tested by using standard solutions (eight levels and six replicates per level). A linear regression over the range of nominal concentrations of 10– 1280 g/L in MeCN had an R2 value ࣙ0.999 (Table 1).
C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
3.2.2 Recovery The method was validated for recovery rates of the acrylamide in DI water (as blank), dry potato, French fries and potato chips at 0, 2.17 and 4.34 g of acrylamide for every 2 g of sample, and 0, 70 and 140 L of the 31 ng/L of acrylamide in MeCN solution was added to samples, respectively, before extraction. The samples were treated as described before. The acrylamide recovery rate in MeCN from the spiked solutions varied between 64 and 69% (Table 2). 3.2.3 LOD and LOQ The LOD was established using an LOD = 3.3×(s/S) and an LOQ was established using LOQ = 10×(s/S), where s was the SD of the 5 g/L standard with a S/N of approximately 3, and S was the slope of the calibration curve. The LOD and the LOQ obtained by this method were 3.5 and 10.6 g/L in MeCN (equivalent to 27 and 81 g/kg in the food samples, with an average 66% recovery rate), respectively. The lower LOQ could be determined by increasing the injection volume. The EFSA monitoring data from 2007–2012 (2013/647/EU) showed that the indicative acrylamide values were 50 g/kg to 4000 g/kg in fried food; therefore, GC–NCD could be used to screen those samples. 3.2.4 Comparison of GC–NCD with other techniques It was found that the typical LOD of LC–MS was 2 –10 g/kg [7–14]; for direct analysis, it was 40 g/kg by GC–MS [19] and 150 g/kg by GC with nitrogen phosphorous detection [41]; it was 2–10 g/kg with derivatization GC–MS [16–27]; and it was 0.1–0.6 g/kg by GC with electrochemical detection (ECD) [42–44]. The analysis time for LC–MS was 8–20 min [16] and for GC–MS was 24.6–30 min [16, 17]. Although the LOD of GC– nitrogen chemiluminescence detector NCD was higher than that for LC–MS or GC–MS, the system’s low cost, the method’s simplicity, and nitrogen chemiluminescence detector’s high selectivity were still attractive advantages. The GC total cycle time was approximately 12 min (6.5 min of run time, 3 min of post-run time and approximately 2 min of oven cooling time); therefore, this method could be used as a screening step before a final LC–MS or GC–MS analysis.
www.jss-journal.com
J. Sep. Sci. 2015, 38, 2272–2277
4 Conclusions Through proper sample preparation and configuration of the gas chromatograph, GC–NCD analysis can be used for the determination of acrylamide in starchy foods. The key aspects of the approach are the following: a rapid and simple sample preparation through modified QuEChERS, a largevolume injection to improve the detection limits to below 100 g/kg, heart-cutting to eliminate interferences and the acetonitrile solvent that is incompatible with the detector, and backflushing to reduce the cycle time and carryover. Combining these advantages enabled this first reported approach of GC with nitrogen chemiluminescence detector to determine acrylamide in food matrices. The authors have declared no conflict of interest.
References [1] Swedish National Food Administration, Information About Acrylamide in Food, April 24, 2002. [2] Taeymans, D., Wood, J., Ashby, P., Blank, I., Studer, A., Stadler, R. H., Gonde, P., VanEijck, P., Lalljie, S., Lingnert, H., Lindblom, M., Matissek, R., Muller, D., Tallmadge, D., O’Brien, J., Thompson, S., Silvani, D., Whitmore, T., Crit. Rev. Food Sci. Nutr. 2004, 44, 323–347. [3] Zhang, Y., Zhang, G., Zhang, Y., J. Chromatogr. A 2005, 1075, 1–21. [4] Keramat, J., LeBail, A., Prost, C., Soltanizadeh, N., Food Bioprocess Technol. 2011, 4, 340–363. [5] Tekkeli, S. E. K., Onal, C., Onal, A., Food Anal. Methods 2012, 5, 29–39. [6] Elbashir, A. A., Omar, M. M. A., Ibrahim, W. A. W., Schmitz, O. J., Aboul-Enein, H. Y., Crit. Rev. Anal. Chem. 2014, 44(2), 107–141. [7] Jezussek, M., Schieberle, P., J. Agric. Food Chem. 2003, 51, 7866–7871. [8] Delatour, T., Perisset, A., Goldmann, T., Riediker, S., Stadler, R. H., J. Agric. Food Chem. 2004, 52, 4625–4631. [9] Aguas, P. C., Fitzhenry, M. J., Giannikopoulos, G., Varelis, P., Anal. Bioanal. Chem. 2006, 385, 1526–1531. [10] Wenzl, T., Karasek, L., Rosen, J., Hellenaes, K.-E., Crews, C., Castle, L., Anklama, E., J. Chromatogr. A 2006, 1132, 211–218. ´ K.-E., J. Chromatogr. A. [11] Rosen, J., Nyman, A., Hellenas, 2007, 1172, 19–24. [12] Karasek, L., Wenzl, T., Anklam, E., Food Chem. 2009, 114, 1555–1558. [13] Bortolomeazzi, R., Munari, M., Anese, M., Verardo, G., Food Chem. 2012, 135, 2687–2693. [14] Ozer, M. S., Kola, O., Altan, A., Duran, H., Zorlugenc, B., J. Food Agric. Environ. 2012, 10, 74–77. [15] Yang, F., Li, Z., Bian, Z., Tang, G., Fan, Z., Wang, Y., Liu, S., Zhang, H., J. Sep. Sci. 2014, 37, 3625–3631. [16] DIN 10785:2013 Analysis of coffee and coffee products – Determination of acrylamide – Methods using HPLC– MSMS and GC–MS after derivatization
C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Gas Chromatography
2277
[17] Rothweiler, B., Kuhn, E. Prest, H., GC–MS approaches to the analysis of acrylamide. PittCon 2003 Poster 2003. [18] Pittet, A., Perisset, A., Oberson, J. M., J. Chromatogr. A. 2004, 1035, 123–130. [19] Becalski, A., Lau, B. P.-Y., Lewis, D., Seaman, S. W., Sun, W. F., Friedman, M., Mottram, D., Eds., Chemistry and Safety of Acrylamide in Food, Springer: New York, 2005, pp 271–284. [20] Tateo, F., Bononi, M., Andreoli, G., J. Food Compos. Anal. 2007, 20, 232–235. [21] Fernandes, J. O., Soares, C., J. Chromatogr. A. 2007, 1175, 1–6. [22] Palazoglu, T. K., Gokmen, V., J. Agric. Food Chem. 2008, 56, 6162–6166. [23] Soares, C. M. D., Alves, R. C., Casal, S., Oliveira, M. B. P., Fernandes, J. O., J. Food Sci. 2010, 75, T57–T63. [24] Bent, G.-A., Maragh, P., Dasgupta, T., Food Chem. 2012, 133, 451–457. [25] Komthong, P., Suriyaphan, O., Charoenpanich, J., Food Addit Contam Part B Surveill 2012, 5, 20–25. [26] Lim, H. H., Shin, H. S., J. Sep. Sci. 2013, 36, 3059– 3066. [27] Blanch, G. P., Morales, F. J., Moreno, Fde L., del Castillo, M. L., J. Sep. Sci. 2013, 36, 320–324. [28] Castle, L., Eriksson, S., J. AOAC Int. 2005, 88, 274–284. [29] Mastovska, K., Lehotay, S. J., J. Agric. Food Chem. 2006, 54(19): 7001–7008. [30] Dunovska, L., Cajka, T., Hajslova, J., Holadova, K., Anal. Chim. Acta. 2006, 578, 234–240. [31] Weisshaar, R., Routes. Eur. J. Lipid Sci. Technol. 2004, 106, 786–792. [32] Yan, X., J. Chromatogr. A 2002, 976, 3–10. [33] Deans, D. R., Chromatographia 1968, I, 187–194. [34] Tranchida, P. Q., Sciarrone, D., Dugo, P., Mondello L., Anal. Chim. Acta 2012, 716, 66–75. [35] Boeker, P., Leppert, J., Mysliwietz, B., Lammers, P. S., Anal Chem. 2013, 85, 9021–30. [36] Grob K., On-column Injection in Capillary Gas Chromatography, Huthig, Heidelberg, 1987, pp. 512, 110. [37] Klee, M. S., Wylie, P., Wilson B., Agilent Application Note 5965–7770E, 1997. [38] Hoh, E., Mastovska, K., J. Chromatogr. A 2008, 1186, 2– 15. [39] Klee, M. S., J. Sep. Sci. 2009, 32, 88–98. [40] Garrido Frenich, A.; Martinez Vidal, J. L.; Fernandez Moreno, J. L.; Romero-Gonzalez, R. J., Chromatogr. A 2009, 1216, 4798–4808. [41] Kim, S. H., Hwang, J.-H., Lee, K.-G., Food Sci. Biotechnol. 2011, 20, 835–839. [42] Zhang, Y., Dong, Y., Ren, Y., Zhang, Y., J. Chromatogr. A 2006, 1116, 209–216. [43] Zhu, Y., Li, G., Duan, Y., Chen, S., Zhang, C., Li, Y., Food Chem. 2008, 109, 899–908. [44] Qu, Y., Liu, C., Luo, F., Qiu, B., Chen, X., J. Sep. Sci. 2013, 36, 3889–3895.
www.jss-journal.com
J. Sep. Sci. 2015, 38, 2272–2277
Yao Weijun
Research Article
Agilent Technologies (Shanghai) Co. Ltd., Shanghai 200131, China
Direct determination of acrylamide in food by gas chromatography with nitrogen chemiluminescence detection
Received January 19, 2015 Revised April 8, 2015 Accepted April 9, 2015
A method of gas chromatography with nitrogen chemiluminescence detection and using standard addition is described for the determination of acrylamide in heat-processed foods. Using a modified QuEChERS (quick, easy, cheap, effective, rugged, and safe) sample preparation method removes the acrylamide precursors completely, and the risk of overestimating acrylamide concentration due to additional analyte formation in the hot gas chromatograph inlet is also avoided. Sample preparation is rapid and inexpensive. A Deans switch device is utilized to heart-cut acrylamide and to prevent interferences from the solvent and matrix from reaching the detector. The pre-column is backflushed at high temperature to maintain a clean baseline and shorten the cycle time compared to baking out the column. Quantitation using standard addition is employed for compensation of potential variability in the acrylamide extraction efficiency in acetonitrile. The limit of detection and the limit of the quantification obtained for this method are 27 and 81 g/kg, respectively, in food samples (equivalent to 3.5 and 10.6 g/L in acetonitrile, respectively), and the linear range is 76– 9697 g/kg in food samples (equivalent to 10–1280 g/L in acetonitrile) with an R2 value of 0.9999. Keywords: Acrylamide / Gas chromatography / Heart cutting / Large-volume injection / Nitrogen chemiluminescence detection / Standard addition method DOI 10.1002/jssc.201500060
1 Introduction The 2002 discovery of acrylamide occurring in various heat processed starch-rich foods at concentrations exceeding those suggested for drinking water [1] led analytical chemists to explore reliable analytical methods for the determination of this potentially carcinogenic contaminant in a wide range of food matrices [2–6]. At present, the majority of methods are based on a mass spectrometer as a detection system coupled to LC [7–16] or GC [16–27]. Assays employing GC–MS techniques are either based on bromination derivatives or direct analysis without derivatization. The approach without derivatization is less laborious, but it is accompanied by problems of potential interfering co-extractives and the lower solubility of acrylamide in organic solvents in comparison to water during the sample preparation [2, 6, 28, 29]. LC–MS methods are attractive because of simple sample preparation, and the
Correspondence: Agilent Technologies (Shanghai) Co. Ltd., 412 Yinglun Road, Waigaoqiao Free Trade Zone, Shanghai, 200131, China Fax: +86 21 5048 2656 Email: [email protected]
Abbreviations: MMI, multi-mode inlet; NCD, nitrogen chemiluminescence detector; PCM, pressure control module; PSA, primary secondary amine
C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
performance of the LC–MS/MS method is found to be superior to that of the GC–MS method [10]. The issue in direct GC determination of free acrylamide arises from the same mechanism of the acrylamide formation in heated foods—the Maillard reaction. If the precursors (asparagine and reducing sugars) are not completely removed from the extract, additional acrylamide forms in the hot GC inlet, resulting in an overestimation of the native acrylamide content [29–31]. The novel sample preparation method developed by Mastovska and Lehotay overcomes this problem by the combination of MeCN extraction and primary secondary amine (PSA) cleanup. The usage of MeCN as an extraction solvent avoids the transfer of acrylamide precursors due to the low solubility of these compounds in this organic solvent. Additionally, the primary secondary amine sorbent strongly retains residual asparagine if it is present in the MeCN extract after the partition step. However, the overall extraction efficiency of acrylamide is reported to be over 60% [29]. The nitrogen chemiluminescence detector (NCD) has high sensitivity and selectivity to nitrogen-containing compounds such as acrylamide and MeCN. The nitrogencontaining compounds are oxidatively combusted at approximately 900⬚C for the complete conversion of chemically bound nitrogen into nitric oxide (R–N+O2 → CO2 + H2 O+NO), and a subsequent reaction with ozone to derive chemiluminescence (NO+O3 →NO2 * →NO + hv) [32].
www.jss-journal.com
J. Sep. Sci. 2015, 38, 2272–2277
At present, there is no report on the analysis of acrylamide in food matrices by nitrogen chemiluminescence detector.
2 Materials and methods 2.1 Materials Acrylamide (99.5%) was purchased from labor Dr. Ehrenstorfer (Augsburg, Germany). The stock solutions of acrylamide were prepared in MeCN and were stored at 4⬚C when not in use. The primary secondary amine sorbent was purchased from Agilent Technologies (Delaware, USA). The MeCN and n-hexane were supplied by Merck (Darmstadt, Germany); the deionized water was obtained from a DI apparatus (Millipore Corporation, MA, USA); the MgSO4 and NaCl were of analytical grade. The food samples (dry and unheated potato, French fries and potato chips) were purchased from local supermarkets.
2.2 Sample preparation The food samples were homogenized by a laboratory blender. Two grams of a representative sample were weighed into a 50 mL centrifuge tube, and 5 mL of hexane, 10 mL of deionized water and 10 mL of MeCN were added by using solvent dispensers. After the addition of 4 g of anhydrous MgSO4 and 0.5 g of NaCl, the tube was immediately sealed and was vigorously shaken for 1 min to avoid the formation of crystalline agglomerates and to ensure the sufficient solvent interaction with the sample. Water facilitated the extraction of acrylamide; hexane was served for sample defatting, and the salt combination induced the separation of the water and acetonitrile layers and forced the majority of acrylamide into the acetonitrile layer [29]. The sample tube was centrifuged for 5 min at 4000 rpm using a T.D.L-5000bR centrifuge (Anke, Shanghai, China), and the three layers were clearly separated. The fat was extracted into the upper hexane phase. The solids and salts were left in the lower water layer. A 1 mL volume of the middle MeCN phase was transferred into a centrifuge tube containing 50 mg of primary secondary amine and 150 mg of anhydrous MgSO4 . This suspension was mixed using a Vortex mixer for 30 s and centrifuged at 4000 rpm for 1 min. With no other treatment, the supernatant was placed into an autosampler vial for GC–NCD analysis. To determine the acrylamide present using the method of standard addition, two additional 2 g subsamples of every sample were also analyzed as above, but first each subsample was spiked with acrylamide using either 70 or 140 L of the 31 ng/L acrylamide standard.
2.3. Instrumentation The GC–NCD analyses were performed on an Agilent 7890A gas chromatograph, equipped with a multi-mode inlet (MMI), C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Gas Chromatography
2273
a Deans switch device (CFT), and electronic pressure control module (PCM), coupled to an Agilent 255 model nitrogen chemiluminescence detector. Fig. 1 shows a typical configuration. The retention gap is a 0.5 m × 0.53 mm deactivated fused-silica column. Analytical separation was performed in a 5 m × 0.53 mm × 1 m DBWaxEtr pre-column and a 15 m × 0.53 mm x 1 m DBWaxEtr analytical column from Agilent. The sample introduction was performed by an Agilent automatic liquid injector (model 7683B).
2.4 GC–NCD operation conditions The sample injection volume was 20 L (5 L × 4 times, normal injection speed, total injection time was around 30 s, splitless mode, 3 min purge time). The MMI inlet temperature was 70⬚C (1 min) and was increased by 600⬚C/min to 300⬚C. The oven temperature started at 60⬚C (1 min), was increased by 25⬚C/min to 150⬚C, and then held for 2.5 min (a total of 6.5 min). The column was configured with dimensions of 20 m × 0.53 mm × 1 m with the MMI inlet as the column inlet pressure source and the other (atmospheric pressure) as the outlet. The flow rate was set to 20 mL/min of He in constant flow mode. The initial source pressure was read at 9.1 psig. The switching flow from pressure control module started at 45 mL/min (1 min) and was increased by 2 mL/min to 40 mL/min. The Deans switch valve was cycled from its initial state of OFF to ON at 4.05 min, and then OFF again at 4.18 min per the elution time of a known standard as explained below. The post run was 3 min, with the oven at 240⬚C, the switching flow from the pressure control module at 60 mL/min, and the inlet pressure 2 psi. The nitrogen chemiluminescence detector conditions were as follows: 930⬚C burner temperature, 5 mL/min of hydrogen and 10 mL/min of oxygen.
3 Results and discussion 3.1 Method development As previously described, the combination of MeCN extraction and primary secondary amine cleanup is an efficient sample preparation method for this approach of direct GC injection because it removes the acrylamide precursors completely (acrylamide was not detected in the blank matrix of dry and unheated potato samples, Fig. 2). However, MeCN contains nitrogen and we observed that a large amount of MeCN solvent would compromise the nitrogen chemiluminescence detector stability and cause solvent peak tailing, which would interfere with the acrylamide determination. The solvent exchange to methanol by evaporation was attempted, but the data showed that the recovery was not reproducible. Therefore, the approach using a Deans switching device to heart-cut the target acrylamide and a much smaller amount of co-eluted www.jss-journal.com
2274
Y. Weijun
J. Sep. Sci. 2015, 38, 2272–2277
Figure 1. Schematic diagram of the GC–NCD system equipped with Deans switching. (A) bypass mode: elutes from the pre-column are swept to the vent; (B) heart-cutting mode: the target acrylamide is heart-cut to the analytical column for further separation; (C) backflush mode: the P1 pressure is reduced to 2 psig which is much lower than P2, and then the flow direction in pre-column is reversed; the residue in the pre-column is backflushed to the split vent under a high column temperature.
solvent into a second analytical column was a preferred solution [33, 34]. The valveless technique of the Deans switch performs flow switching by setting the appropriate pressure levels in a network of flow resistors, and the pressures and flows are controlled by valves and flow sources outside of the analytical flow path. The ability to switch the flow from the pressure control model needs to be set up appropriately so that the pre-column effluent could be completely purged to vent the analytical column by changing the three-way valve’s state. The low switching flow rate leads to an incomplete transfer and acts as a splitter [35], while the high switching flow rate raises the pre-column’s outlet pressure, causing the pre-column’s actual flow rate to become low and the peaks to become flattened. The GC manufacturer recommends the value of 2–3 times the primary flow, so a switching flow rate C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
of 45 mL/min was chosen. The restrictor’s dimension was calculated from the Agilent Dean Switch Calculator (Version A.01.01). The experimental results showed that a low level of solvent worked well under the software recommended conditions, but a large amount of solvent partially entered into the analytical column while the valve was in the OFF state. A practical approach is to cut the restrictor’s length step by step until the solvent is purged to vent completely in the OFF state. Careful timing of the Deans switching valve resulted in a stable baseline and the complete separation of real samples (Figs. 2 and 3). The correct timing of the valve switching time was determined by iteration as follows. First, a wide heart-cutting time between 3.0 min (ON) and 5.0 min (OFF) was chosen to ensure that the acrylamide standard was completely eluted and its response area was noted (as A0 ). The www.jss-journal.com
Gas Chromatography
J. Sep. Sci. 2015, 38, 2272–2277
2275
Table 1. Linearity results of using the calibration standards y = 6.746x R2 = 0.9999
Figure 2. The overlaid chromatograms of unheated potato, French fries, potato chips and spiked potato chips in ascending order of the acrylamide concentration.
Figure 3. The overlaid chromatograms of the 10, 20 and 40 g/L acrylamide standards in MeCN.
observed chromatographic peak width was less than 0.2 min, so a narrow heart-cutting window of 0.2 min was applied to the later tuning steps. Next, a second heart-cutting timing of 3.0 min (ON) and 3.2 min (OFF) was tried and compared with the earlier acrylamide area response (A0). By repeatedly increasing the heart-cut time by 0.2 min and comparing the acrylamide response area against the original value (A0 ), the best cutting window was obtained. The final heart-cut time window was optimized with 0.01 min increments to further maximize the selection for acrylamide against other interferences (such as residual solvent). It is worth mentioning that the presence of an additional detector (FID, NPD, etc.) where the vent is indicated in Figure 1 would have eliminated the need for iterating the collection window. A large-volume injection (LVI) shows a great advantage in modern GC analysis [36–38]. The increased sample injection volume (20 L) improved the acrylamide detection limits, and it eliminated the need for extra concentration steps that can otherwise cause analyte loss. The MMI was operated in splitless mode, rather than the commonly used solvent vent
C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Acrylamide in MeCN (g/l)
Relative response factor to 1280 g/l standard
RSD% (n = 6)
10 20 40 80 160 320 640 1280
0.98 0.96 0.98 0.97 1.00 0.97 1.00 1.00
4.3 3.6 1.7 1.2 1.2 1.2 1.2 0.6
mode. We observed that the response was approximately 10% less in solvent vent mode. This could be due to a loss of acrylamide from the split vent due to solvent evaporation. The sample injection ended at 30 s. The solvent was vaporized in the inlet liner and retention gap immediately, and its vapor passed to pre-column and vented. The majority of acrylamide was held in the inlet liner. Still, a minority of acrylamide was carried out of the liner by the solvent and was trapped in retention gap under the initial oven temperature. The halfmeter retention gap minimized the band broadening caused by the solvent flooding into the column head [36]. The later eluting contaminants within a matrix could cause several problems, such as ghost peaks, baseline perturbations, high background, and shifting retention times. It is common to extend the temperature program and bakeout period to remove the retained sample components and column contaminants. A significant reduction in baking time is readily achieved when a valve or Tee is used to reverse the column flow [39]. This feature is combined in Deans switch and made backflushing the pre-column a routine practice. In the post run stage, the MMI inlet’s head pressure P1 was reduced to a low value (2 psi), while the exit pressure of the pre-column P2 was increased to a high value (approximately 20 psi). The flow direction in the pre-column was reversed. The residue in the pre-column was backflushed to a split vent at a high column temperature (240⬚C). Approximately, a 30% reduction in the total cycle time was achieved versus the usual approach of an extended high-temperature temperature bakeout period to remove the retained extraneous sample components. Here, the column configuration was different from the common Tee setting, where the mid-point flow source was set as the outlet of the pre-column and inlet of the analytical column. Unlike the negligible resistance between the pressure control module and mid-point of the Tee, the resistance of the CFT plate’s inner channel and narrow tube connecting the pressure control module and CFT plate were variable, and the actual mid-point pressure was lower than the pressure control module value and was dependent on the hardware. The current setting of the MMI inlet as the column inlet pressure source and the other as the outlet showed a high consistency of retention times.
www.jss-journal.com
2276
J. Sep. Sci. 2015, 38, 2272–2277
Y. Weijun
Table 2. The recovery yields of acrylamide obtained from the GC–NCD method
Acrylamide in MeCN (g/L)
Recovery (%)
Sample material
before
217 g/L
434 g/L
217 g/L
434 g/L
Acrylamide in sample
DI water (blank) dry potato French fries potato chips
spiking 0 0 58 93
spiked 143 144 201 243
spiked 278 281 336 374
spiked 66 66 66 69
spiked 64 65 64 65
(g/kg) < LOD < LOD 441 723
A common way of evaluating the recovery rate is by the spiking of internal standards. The isotope-labeled internal standards, such as deuterium-labeled (2 H3 -Acrylamide or d3 -Acrylamide) or carbon-labeled (13 C3 -Acrylamide), are the most ideal internal standards, but they can only be used in MS-based analysis and are inappropriate for nitrogen chemiluminescence detector. Non-isotope-labeled internal standards such as methylacrylamide, N,N-dimethylacrylamide, and acetamide are widely used also; however, satisfactory repeatability might not be achieved due to the differing stability and structure of the compounds [6]. A standard addition calibration is appropriate when there is no blank matrix to carry out the calibration [40–42]. Known amounts of analyte are introduced into the sample and are treated along with the sample. A one-point calibration was used for the quantification process [40], and the sample amount was obtained using the equation: Csample = Cadded × Ssample / Ssample+added − Ssample
(1)
where Csample was the initial concentration in the sample, Cadded was the added concentration to the sample, Ssample was the peak area of the sample, and Ssample+added was the peak area that corresponds to the sample with the added concentration. The recovery rate was calculated as follows: Recovery% = Ssample+added −Ssample /Sadded × 100 where, S’ added was the calculated peak area of standard added in MeCN from the external standard calibration curve.
3.2. Method assessment 3.2.1. Linearity The linearity of the method was tested by using standard solutions (eight levels and six replicates per level). A linear regression over the range of nominal concentrations of 10– 1280 g/L in MeCN had an R2 value ࣙ0.999 (Table 1).
C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
3.2.2 Recovery The method was validated for recovery rates of the acrylamide in DI water (as blank), dry potato, French fries and potato chips at 0, 2.17 and 4.34 g of acrylamide for every 2 g of sample, and 0, 70 and 140 L of the 31 ng/L of acrylamide in MeCN solution was added to samples, respectively, before extraction. The samples were treated as described before. The acrylamide recovery rate in MeCN from the spiked solutions varied between 64 and 69% (Table 2). 3.2.3 LOD and LOQ The LOD was established using an LOD = 3.3×(s/S) and an LOQ was established using LOQ = 10×(s/S), where s was the SD of the 5 g/L standard with a S/N of approximately 3, and S was the slope of the calibration curve. The LOD and the LOQ obtained by this method were 3.5 and 10.6 g/L in MeCN (equivalent to 27 and 81 g/kg in the food samples, with an average 66% recovery rate), respectively. The lower LOQ could be determined by increasing the injection volume. The EFSA monitoring data from 2007–2012 (2013/647/EU) showed that the indicative acrylamide values were 50 g/kg to 4000 g/kg in fried food; therefore, GC–NCD could be used to screen those samples. 3.2.4 Comparison of GC–NCD with other techniques It was found that the typical LOD of LC–MS was 2 –10 g/kg [7–14]; for direct analysis, it was 40 g/kg by GC–MS [19] and 150 g/kg by GC with nitrogen phosphorous detection [41]; it was 2–10 g/kg with derivatization GC–MS [16–27]; and it was 0.1–0.6 g/kg by GC with electrochemical detection (ECD) [42–44]. The analysis time for LC–MS was 8–20 min [16] and for GC–MS was 24.6–30 min [16, 17]. Although the LOD of GC– nitrogen chemiluminescence detector NCD was higher than that for LC–MS or GC–MS, the system’s low cost, the method’s simplicity, and nitrogen chemiluminescence detector’s high selectivity were still attractive advantages. The GC total cycle time was approximately 12 min (6.5 min of run time, 3 min of post-run time and approximately 2 min of oven cooling time); therefore, this method could be used as a screening step before a final LC–MS or GC–MS analysis.
www.jss-journal.com
J. Sep. Sci. 2015, 38, 2272–2277
4 Conclusions Through proper sample preparation and configuration of the gas chromatograph, GC–NCD analysis can be used for the determination of acrylamide in starchy foods. The key aspects of the approach are the following: a rapid and simple sample preparation through modified QuEChERS, a largevolume injection to improve the detection limits to below 100 g/kg, heart-cutting to eliminate interferences and the acetonitrile solvent that is incompatible with the detector, and backflushing to reduce the cycle time and carryover. Combining these advantages enabled this first reported approach of GC with nitrogen chemiluminescence detector to determine acrylamide in food matrices. The authors have declared no conflict of interest.
References [1] Swedish National Food Administration, Information About Acrylamide in Food, April 24, 2002. [2] Taeymans, D., Wood, J., Ashby, P., Blank, I., Studer, A., Stadler, R. H., Gonde, P., VanEijck, P., Lalljie, S., Lingnert, H., Lindblom, M., Matissek, R., Muller, D., Tallmadge, D., O’Brien, J., Thompson, S., Silvani, D., Whitmore, T., Crit. Rev. Food Sci. Nutr. 2004, 44, 323–347. [3] Zhang, Y., Zhang, G., Zhang, Y., J. Chromatogr. A 2005, 1075, 1–21. [4] Keramat, J., LeBail, A., Prost, C., Soltanizadeh, N., Food Bioprocess Technol. 2011, 4, 340–363. [5] Tekkeli, S. E. K., Onal, C., Onal, A., Food Anal. Methods 2012, 5, 29–39. [6] Elbashir, A. A., Omar, M. M. A., Ibrahim, W. A. W., Schmitz, O. J., Aboul-Enein, H. Y., Crit. Rev. Anal. Chem. 2014, 44(2), 107–141. [7] Jezussek, M., Schieberle, P., J. Agric. Food Chem. 2003, 51, 7866–7871. [8] Delatour, T., Perisset, A., Goldmann, T., Riediker, S., Stadler, R. H., J. Agric. Food Chem. 2004, 52, 4625–4631. [9] Aguas, P. C., Fitzhenry, M. J., Giannikopoulos, G., Varelis, P., Anal. Bioanal. Chem. 2006, 385, 1526–1531. [10] Wenzl, T., Karasek, L., Rosen, J., Hellenaes, K.-E., Crews, C., Castle, L., Anklama, E., J. Chromatogr. A 2006, 1132, 211–218. ´ K.-E., J. Chromatogr. A. [11] Rosen, J., Nyman, A., Hellenas, 2007, 1172, 19–24. [12] Karasek, L., Wenzl, T., Anklam, E., Food Chem. 2009, 114, 1555–1558. [13] Bortolomeazzi, R., Munari, M., Anese, M., Verardo, G., Food Chem. 2012, 135, 2687–2693. [14] Ozer, M. S., Kola, O., Altan, A., Duran, H., Zorlugenc, B., J. Food Agric. Environ. 2012, 10, 74–77. [15] Yang, F., Li, Z., Bian, Z., Tang, G., Fan, Z., Wang, Y., Liu, S., Zhang, H., J. Sep. Sci. 2014, 37, 3625–3631. [16] DIN 10785:2013 Analysis of coffee and coffee products – Determination of acrylamide – Methods using HPLC– MSMS and GC–MS after derivatization
C 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Gas Chromatography
2277
[17] Rothweiler, B., Kuhn, E. Prest, H., GC–MS approaches to the analysis of acrylamide. PittCon 2003 Poster 2003. [18] Pittet, A., Perisset, A., Oberson, J. M., J. Chromatogr. A. 2004, 1035, 123–130. [19] Becalski, A., Lau, B. P.-Y., Lewis, D., Seaman, S. W., Sun, W. F., Friedman, M., Mottram, D., Eds., Chemistry and Safety of Acrylamide in Food, Springer: New York, 2005, pp 271–284. [20] Tateo, F., Bononi, M., Andreoli, G., J. Food Compos. Anal. 2007, 20, 232–235. [21] Fernandes, J. O., Soares, C., J. Chromatogr. A. 2007, 1175, 1–6. [22] Palazoglu, T. K., Gokmen, V., J. Agric. Food Chem. 2008, 56, 6162–6166. [23] Soares, C. M. D., Alves, R. C., Casal, S., Oliveira, M. B. P., Fernandes, J. O., J. Food Sci. 2010, 75, T57–T63. [24] Bent, G.-A., Maragh, P., Dasgupta, T., Food Chem. 2012, 133, 451–457. [25] Komthong, P., Suriyaphan, O., Charoenpanich, J., Food Addit Contam Part B Surveill 2012, 5, 20–25. [26] Lim, H. H., Shin, H. S., J. Sep. Sci. 2013, 36, 3059– 3066. [27] Blanch, G. P., Morales, F. J., Moreno, Fde L., del Castillo, M. L., J. Sep. Sci. 2013, 36, 320–324. [28] Castle, L., Eriksson, S., J. AOAC Int. 2005, 88, 274–284. [29] Mastovska, K., Lehotay, S. J., J. Agric. Food Chem. 2006, 54(19): 7001–7008. [30] Dunovska, L., Cajka, T., Hajslova, J., Holadova, K., Anal. Chim. Acta. 2006, 578, 234–240. [31] Weisshaar, R., Routes. Eur. J. Lipid Sci. Technol. 2004, 106, 786–792. [32] Yan, X., J. Chromatogr. A 2002, 976, 3–10. [33] Deans, D. R., Chromatographia 1968, I, 187–194. [34] Tranchida, P. Q., Sciarrone, D., Dugo, P., Mondello L., Anal. Chim. Acta 2012, 716, 66–75. [35] Boeker, P., Leppert, J., Mysliwietz, B., Lammers, P. S., Anal Chem. 2013, 85, 9021–30. [36] Grob K., On-column Injection in Capillary Gas Chromatography, Huthig, Heidelberg, 1987, pp. 512, 110. [37] Klee, M. S., Wylie, P., Wilson B., Agilent Application Note 5965–7770E, 1997. [38] Hoh, E., Mastovska, K., J. Chromatogr. A 2008, 1186, 2– 15. [39] Klee, M. S., J. Sep. Sci. 2009, 32, 88–98. [40] Garrido Frenich, A.; Martinez Vidal, J. L.; Fernandez Moreno, J. L.; Romero-Gonzalez, R. J., Chromatogr. A 2009, 1216, 4798–4808. [41] Kim, S. H., Hwang, J.-H., Lee, K.-G., Food Sci. Biotechnol. 2011, 20, 835–839. [42] Zhang, Y., Dong, Y., Ren, Y., Zhang, Y., J. Chromatogr. A 2006, 1116, 209–216. [43] Zhu, Y., Li, G., Duan, Y., Chen, S., Zhang, C., Li, Y., Food Chem. 2008, 109, 899–908. [44] Qu, Y., Liu, C., Luo, F., Qiu, B., Chen, X., J. Sep. Sci. 2013, 36, 3889–3895.
www.jss-journal.com
