Development and Validation of Benzophenone Derivatives in Packaged Cereal-Based Foods by Solid–Liquid Extraction and Ultrahigh-Performance Liquid Chromatography–Tandem Mass Spectrometry

We established and validated a sensitive multi-residue analytical method for identifying benzophenone (BP) and nine BP derivatives (2,4-dihydroxybenzophenone [BP-1], 2,2′,4,4′-tetrahydroxydroxybenzophenone, 2-hydroxy-4-methoxy benzophenone, 2,2′-dihydroxy 4-methoxy benzophenone, 2-hydroxybenzophenone [2-OHBP], 4-hydroxybenzophenone, 4-methylbenzophenone [4-MBP], methyl-2-benzoylbenzoate, and 4-benzoylbiphenyl). Solid–liquid extraction pretreatment and ultra-high-performance liquid chromatography–tandem mass spectrometry (UHPLC–MS/MS) were employed in an analysis of 85 packaged cereal-based food samples (25 pastry, 50 rice, and 10 noodle samples). The method had satisfactory linearity (R2 ≥ 0.995), low limits of detection (pastry: 0.02–4.2 ng/g; rice and noodle: 0.02–2 ng/g), and favorable precision, with within-run and between-run coefficient of variation ranges of 1–29% and 1–28%, respectively. BP and 4-MBP were detected in 100% of the pastry samples, and BP-1 and 2-OHBP were found in 76% and 56% of the pastry samples, respectively. BP and 2-OHBP were found in 92% and 38% of the rice samples, respectively. BP was found in 50% of the noodle samples. BP contributed the most to the total level of BPs in pastries, with significantly higher mean ± standard deviation (range) levels for pastries (26.8 ± 32.6 [1.8–115.4] ng/g) than rice (1.2 ± 2.0 [0.4–13.4] ng/g) and noodles (0.7 ± 0.7 [0.4–1.9] ng/g); p < 0.0001). The trace levels of 4-MBP identified in the samples demonstrate the need for the development of analytical methods with high sensitivity and specificity; the proposed method satisfies this need.


Introduction
Cereal-based products, such as rice and noodles, are staple foods in many countries; they comprise approximately 75% of the average individual's carbohydrate intake (mainly in the form of starch) and 6-15% of their protein, vitamin, and mineral intake [1]. Studies have investigated the importance of cereals-based products, particularly wholegrain products, in providing these carbohydrates due to the presence of dietary fiber and bioactive compounds in such products. Diets rich in dietary fiber and whole grains are associated with a decreased risk of atherosclerotic cardiovascular diseases [2,3], type 2 diabetes [3], and obesity [2]. In the packaging of these cereal-based products, benzophenone (BP) and its derivatives (BPs) are often added to the plastic packaging as ultraviolet (UV) blockers and are used as photoinitiators (PIs) in UV-curable inks for printed food-packaging Sigma-Aldrich (St. Louis, MO, USA), and BP-2 and BP-8 were supplied by Tokyo Chemical Industry (Tokyo, Japan). SIL-IS, BP-d 5 , and BP-3-d 5 were purchased from Sigma-Aldrich (Burlington, MA, USA); BP-1-d 5 , 4-MBP-d 3 , BP-8-d 3 , diOHBP-13 C 6 , and 4-OHBP-d 4 were obtained from Toronto Research (North York, Toronto, Canada). All standards and SIL-IS had purities of >97%. Anhydrous magnesium sulfate (≥99%, MgSO 4 ), LC-grade ACN, formic acid (≥88%), acetic acid (≥99.7%), and LC-MS-grade methanol (MeOH) were purchased from J.T. Baker (Phillipsburg, NJ, USA). Sodium chloride (>99%, NaCl) was obtained from PanReac (Castellar del Vallès, Barcelona, Spain). The bulk sorbent, Sepra C18-E and PSA, and Strata C18-T SPE cartridges (1 mL, 100 mg) were obtained from Phenomenex (Torrance, CA, USA) and Sigma-Aldrich. FaPEx-cer was obtained from Silicycle (Quebec, QC, Canada). Oasis PRiME HLB cartridges (1 mL, 30 mg) were obtained from Waters (Milford, MA, USA). The Milli-Q water of the study was produced by a Sartorius Ultrapure water system to reach a resistivity of 18.2 MΩ cm (Savska, Zaprešić, Croatia).

Packaged Sample Collection and Preparation
A total of 85 packaged cereal-based foodstuffs were randomly selected from bakeries and supermarkets in Taiwan. All the samples were domestic and were classified into three groups: pastries (n = 25), rice (n = 50), and noodles (n = 10). The collected foods were popular brands in Taiwan. The types of packaging materials used on the internal side of the products' packaging were examined, and the food contact materials were recorded to be polypropylene (PP) and polyethylene terephthalate (PET) plastics. PP was the food contact material in 19 pastry, 16 rice, and 1 noodle samples, and PET was the food contact material in 6 pastry, 0 rice, and 8 noodle samples. The types of packaging materials were unknown in 34 rice and 1 noodle samples. Because no blank matrices with an absence of BPs were available, unwrapped samples of pastry (n = 6), rice (n = 6), and noodles (n = 6) were analyzed through UHPLC-MS/MS; among them, three samples of each type were mixed and selected as three types of blank matrices because the BPs were below the limit of detection (LOD).

Preparation of Standards
A stock standard solution of each analyte (500 mg/L) was individually prepared in ACN. The standard working solutions of 50 mg/L, which were prepared by dilution with MeOH in an appropriate volume of the solution, were used to spike the calibration curves of solvents and cereal-based matrices. BP-d 5 , BP-1-d 5 , 4-MBP-d 3 , BP-8-d 3 , diOHBP-13 C 6 , 4-OHBP-d 4 , and BP-3-d 5 were used as SIL-IS for BP, BP-1, 4-MBP, BP-8, 2-OHBP, 4-OHBP, and BP-3, respectively. Because no SIL-IS was available for BP-2, M2BB, and PBZ, 4-OHBPd 4 was used as an SIL-IS for BP-2 and M2BB, and BP-3-d 5 was used as an SIL-IS for PBZ. The SIL-IS mixture was diluted to 20 µg/L.

Analysis of UHPLC-MS/MS Method
With the exception of BP-2, BPs were detected using a UHPLC-MS/MS system (Shimadzu 8045, Kyoto, Japan) connected to a triple-quadrupole MS system with an electrospray ionization positive mode. The analytical column was a UPLC Waters BEH C18 column (1.7 µm, 2.1 mm × 100 mm) and had flow rate of 0.3 mL/min. Mobile phase A was MeOH containing 0.1% formic acid, and mobile phase B was deionized water. The elution gradient was 80-20% B at 3.5 min, 20% B at 1 min, 20-10% B at 1 min, 10% B at 4 min, and 10-80% B at 0.1 min, and re-equilibrated at 80% for 3.9 min. The total run time was 13.5 min, and the sample injection volume was 10 µL. BPs were monitored under multiple reaction monitoring (MRM) modes. The interface, desolvation line, and heat block temperatures were 300, 240, and 400 • C, respectively. The flow rates of the nebulizing, heating, and drying gases were 3, 10, and 10 L/min, respectively. The tandem MS parameters, ion transitions for quantification and qualifications, retention time, and collision energy are presented in Table 1. Data acquisition and processing were conducted using LabSolution (version 5.93, Shimadzu, Kyoto, Japan).

Pretreatment Approaches
Six procedures, namely SLE, FaPEx, QuEChERS with and without cleanup, and the SPE-based method with HLB and C18 cartridges, were compared to select a suitable sample pretreatment method and were tested with five replicates.
Method A, SLE: A homogenized cereal-based food sample (0.5 g) was loaded into a 12 mL glassware tube, and a standard (STD) and SIL-IS (20 and 8 ng/g) and ACN (5 mL) with 1% acetic acid were added. The mixtures were shaken for 1 min, allowed to stand for 60 min, and centrifuged at 2000× g for 10 min. Subsequently, the supernatant solution was dried by nitrogen gas. Finally, the residue was dissolved in 200 µL of MeOH and filtered using 0.22 µm polytetrafluoroethylene filters.
Method B, FaPEx: A homogenized sample (0.5 g) was loaded into a 12 mL glassware tube. Subsequently, pure water (1 mL) was added, and the sample was spiked with an STD and SIL-IS and vortexed for 1 min, and we then waited for 30 min. ACN (5 mL) with 1% acetic acid was added; the mixtures were vortexed for 30 s and centrifuged at 2000× g for 10 min. Thereafter, the supernatant was transferred to a FaPEx-cer kit with MgSO 4 , PSA, C18, and GCB, at a liquid flow rate of 1 drop/s and dried by nitrogen gas. Finally, the residue was treated in the same manner as that used in the SLE method.
Methods C-D, QuEChERS with and without a cleanup: The original QuEChERS method comprises several steps: (1) addition of an IS; (2) extraction using ACN; (3) partitioning with MgSO 4 and NaCl; (4) dispersive SPE aliquot purified with MgSO 4 and SPE sorbents; and (5) adjustment of the pH [30]. The QuEChERS method employed in this study was described in a previous study [27]. A homogenized sample (0.5 g) was placed in a 12 mL glassware tube, and an STD and SIL-IS and deionized water (1 mL) were added. The mixtures were shaken for 1 min and allowed to stand for 60 min, and ACN (5 mL) with 1% acetic acid was added. The mixtures were shaken for 1 min, extraction salt packages (4 g of anhydrous MgSO 4 and 1 g of NaCl) were added, and the mixtures were vigorously shaken again for 1 min. For samples without a cleanup procedure, the extract was centrifuged for 10 min at 2000× g. For those with a cleanup procedure, a supernatant was added to the cleanup sorbents, which contained 0.24 g of MgSO 4 , 0.24 g of C18-E, and 0.08 g of PSA. Finally, the residue was treated in the same manner as that in the SLE procedure.
Methods E-F, SPE-C18, and SPE-HLB: A homogenized sample (0.5 g) was loaded into a 12 mL glassware tube, and an STD, SIL-IS, and ACN (5 mL) with 1% acetic acid were added. The mixtures were shaken for 1 min and centrifuged at 2000× g for 10 min. Thereafter, the supernatant was prepared for purification through SPE by using Strata C18-T and Oasis PRiME HLB. First, the SPE cartridges were conditioned using 2 mL of each MeOH and deionized water in sequence. Subsequently, the supernatant was loaded onto the cartridges. The prepared solutions (no washing solvents, water, and 10% MeOH) were examined with 2 mL for washing efficiency, and MeOH and ACN were tested with 2 mL for elution efficiency. Finally, the eluent was collected and concentrated to dryness by nitrogen gas per the aforementioned SLE procedure.

Method Validation
The method was validated according to the guidelines established in the United States [31] for linearity, the matrix effect (ME), the LOD, the limit of quantification (LOQ), recovery, and precision. Blank matrix samples were unwrapped samples of pastries, rice, and noodles, which were analyzed through UHPLC-MS/MS and had BP contents below the LOD. Linearity was evaluated using solvent-matched and matrix-matched calibration standards with 8 levels (0.04, 2, 10, 30, 60, 72, 92, and 100 ng/g, with an SIL-IS of 8 ng/g) for pastries and 10 levels (0.04, 0.4, 0.6, 1.2, 2.4, 4, 6, 12, 24, and 40 ng/g, with an SIL-IS of 8 ng/g) for rice and noodles. The calibration curves of solvents and matrices were obtained by plotting the quotients of the peak areas of BPs and their corresponding SIL-IS against the levels of the standards. The MEs were evaluated through the comparison of the slopes of standards in solvents with matrix-matched standards. The LOD and LOQ were defined as the levels with signal-to-noise ratios of ≥3 and ≥10, respectively. Blank pastry samples with 3 spiking levels (4, 40, and 80 ng/g) and rice and noodle samples with 3 levels (0.4, 4, and 40 ng/g) were used to evaluate the recovery and precision of the method. The within-run and between-run recoveries were evaluated on the basis of the mean recovery for these spiked samples on the same day (n = 5) and over 3 consecutive days (n = 15), respectively. The precision was used to determine the relative standard deviation (RSD); the within-run and between-run precisions were assessed on the basis of the standard deviation (SD) of the recovery percentage of the spiked samples on a given day (n = 5) and 3 consecutive days (n =15). The validation criteria of the Codex Alimentarius were 70-120% for ME and recovery and RSD < 20% for precision.
To evaluate the potential procedure-associated contamination and verify the instruments' performance, a solvent blank, procedural blank, and medium-spiked sample were applied at intervals of every 10 samples during sample analysis.

Statistical Analysis
The experimental results are presented in terms of the means (SDs, range). Mann-Whitney U test and Kruskal-Wallis tests were used to evaluate the differences in BPs in three types of cereal-based foods and their packaging materials. Statistical analysis was performed using SPSS (version 19.0, SPSS, Chicago, IL, USA), and the significance level was p < 0.05.

Selection of Sample Pretreatment Method
Many studies have investigated the migration of UV ink to foodstuffs [32][33][34]; however, few have simultaneously investigated BPs in cereals and cereal-based foods. Bugey et al. (2013) [21] reported a pretreatment method involving PLE and analysis performed through GC-MS, which yielded an LOQ of 60 ng/g for BP in cereal-based foods. Anderson et al. [19] and Ibarra et al. [10] reported a technique featuring SLE followed by GC-MS for quantifying BP in packaged foods (e.g., burgers, cakes, and rice). The LOD and LOQ of BP were 10 ng/mL and 50-250 ng/mL, respectively. Furthermore, Jung et al. [23] and Van Den Houwe et al. [24] reported an SLE method, with HPLC-MS/MS applied for the quantification of BP, 4-MBP, and 4-OHBP in cereal-based foods with LODs of 31-38, 3-13, and 3 ng/g, respectively. Van Den Houwe et al. [25] performed SLE followed by SPE with an HLB cartridge and then applied the UPLC-MS/MS method to analyze BP, 4-MBP, and 4-OHBP in dry foods, with LODs ranging from 0.1 to 4 ng/g. Chang et al. [26] applied the QuEChERS technique without a cleanup procedure and then used UPLC-MS/MS to detect PIs in breakfast cereals, with LOQs of 20, 10, and 40 ng/g for BP, 4-MBP, and 2-OHBP, respectively. Gallart-Ayala et al. [22] used the QuEChERS technique followed by HPLC-MS/MS for the analysis of 11 PIs in packaged food, which yielded the LOQs of 2.3 and 0.7 ng/g for BP and PBZ, respectively. However, the aforementioned methods were limited to the few BPs that can be simultaneously analyzed and that have a high LOD. Huang et al. [27] developed a FaPEx method, with UHPLC-MS/MS used to analyze BPs in cereals, with LODs ranging from 0.001 to 0.3 ng/g. UHPLC achieves a rapid chromatographic technique with a higher resolution, and lower mobile phase volume compared to HPLC.
In this study, we applied the SLE, FaPEx, QuEChERS with or without a cleanup, SPE-HLB, and SPE-C18 methods in spiked blank samples (pastries, rice, and noodles) at standards and SIL-IS level of 20 and 8 ng/g, respectively. A comparison of the six methods for the extraction and cleanup of the BPs revealed that the recoveries were in the order of SLE > FaPEx > QuEChERS without cleanup ≈ QuEChERS with cleanup > SPE-C18 > SPE-HLB among three matrices (Figure 1). To optimize the SPE procedures, two SPE cartridges, Oasis PRiME HLB and Strata C18, were used to investigate washing and elution efficiencies. In our experiment, 2 mL of water and 10% MeOH were added in the washing step, and 2 mL of MeOH and ACN were added in the elution step. The results are displayed in Figure S1. The SPE-C18 had a higher recovery, although all wash solvents minimized the recovery. The MeOH elution solvent achieved the best recovery. Therefore, the optimization of SPE-C18 was proposed to eliminate the need for a wash procedure and elution with MeOH. In addition, as presented in Figure S2, the blank matrix among the six methods indicated BPs below the LOD, with the exception of 2-OHBP and BP. 2-OHBP was detected using the QuEChERS, and BP was mainly identified using the FaPEx and QuEChERS. Because of its high recovery and lack of background contamination, the SLE approach was proposed for BPs analysis in cereal-based foods. A flowchart of sample pretreatment is listed in Figure 2. Figure 3 presents the chromatograms of a solvent sample spiked with 80 ng/g of BP and BP analog standards.

Method Validation
The efficacy of the developed method was validated in accordance with the of the Codex Alimentarius of the United States [31]. Because the analyte contents in the pastry samples were higher than those in the rice and noodles samples, for the pastries, we used a wider working range in validating the recovery and precision of the method, with two 5-point standard calibration curves covering the range of 0.04-60 ng/g constructed for 4 and 40 ng/g and a curve covering the range of 60-100 ng/g constructed for 80 ng/g. For the rice and noodle samples, two 5-point standard calibration curves covering the range of 0.04-4 ng/g were constructed for 0.4 ng/g and a curve covering the range of 4-40 ng/g was constructed for 4 and 40 ng/g.  , and quick, easy, cheap, effective, rugged, and safe (QuEChERS) using an extraction kit (4 g MgSO4 + 1 g NaCl), QuEChERS using an extraction kit (4 g MgSO4 + 1 g NaCl), and a cleanup procedure (0.24 g C18 + 0.24 g MgSO4 + 0.08 g PSA), SPE with HLB, and SPE with C18 in samples (n = 5 replicates) of (A) pastry, (B) rice, and (C) noodle spiked with standards and IS (20 and 8 ng/g). , and quick, easy, cheap, effective, rugged, and safe (QuEChERS) using an extraction kit (4 g MgSO 4 + 1 g NaCl), QuEChERS using an extraction kit (4 g MgSO 4 + 1 g NaCl), and a cleanup procedure (0.24 g C18 + 0.24 g MgSO 4 + 0.08 g PSA), SPE with HLB, and SPE with C18 in samples (n = 5 replicates) of (A) pastry, (B) rice, and (C) noodle spiked with standards and IS (20 and 8 ng/g). , and quick, easy, cheap, effective, rugged, and safe (QuEChERS) using an extraction kit (4 g MgSO4 + 1 g NaCl), QuEChERS using an extraction kit (4 g MgSO4 + 1 g NaCl), and a cleanup procedure (0.24 g C18 + 0.24 g MgSO4 + 0.08 g PSA), SPE with HLB, and SPE with C18 in samples (n = 5 replicates) of (A) pastry, (B) rice, and (C) noodle spiked with standards and IS (20 and 8 ng/g).    With regard to the recoveries and precision in within-run and between-run tests at 3 spiking levels, the mean within-run and between-run recoveries of the BPs among the pastry, rice, and noodle samples were 45-148% and 44-150%, 75-125% and 70-128%, and 75-145% and 73-152% (Table 4). The mean within-run and between-run precisions (%RSD) of BPs among the 3 matrices were 1-11% and 3-18%, 1-29% and 1-22%, and 1-26% and 1-28%, respectively ( Table 5). Most of the BPs satisfied the validation criteria, with the exception of BP-1 and PBZ [31].   Table 4. Within-run and between-run recoveries in spiked pastry, rice, and noodle samples.
Thus, we developed a more efficient and simpler approach that achieved satisfactory results with a lower LOD and higher precision for detection of BPs in the samples of pastries, rice, and noodles. Moreover, the application of SIL-IS in this approach achieved highly accurate quantification by compensating for recoveries and MEs and by reducing measurement uncertainty.

Applications of Samples of Popular Food Products in the Taiwanese Market
To verify the suitability of the validated method, 85 samples were subjected to analysis. The detection rates and mean (SD, range) levels of BPs in the samples of the pastry (n = 25), rice (n = 50), and noodle (n = 10) samples are presented in Table 7. Of the 10 analytes, 6 were detected within the range of 2-100% among the 3 cereal-based foods, with BP-2, BP-8, M2BB, and PBZ being below the LOD.
In the pastry samples, six analytes were identified within the range of 4-100%, with BP and 4-MBP detected in 100% of the samples and BP-1, 2-OHBP, BP-3, and 4-OHBP detected in 76%, 56%, 32%, and 4% of the samples, respectively. In the rice samples, four analytes were identified within the range of 2-92%, with BP, 2-OHBP, 4-MBP, and BP-1 detected in 92%, 38%, 16%, and 2% of the samples, respectively. In the noodle samples, BP and 4-OHBP were detected in 50% and 10% of the samples, respectively. These results suggest that BP, BP-1, 2-OHBP, and 4-MBP were prevalent in pastries, and BP was ubiquitous in rice and noodles, which is in line with the findings reported in Belgium [25], Germany [23], Italy [34], Spain [22], Switzerland [21], the United Kingdom [19,20], and Taiwan [27]. BP contributed the most to the total level of BPs in the pastry samples; the second highest contributor was BP-3, with mean ± SD (range) levels of 21.6 ± 37.6 [0.4-85.3] ng/g. BP levels were significantly higher in the pastry samples (26.8 ± 32.6 [1.8-115.4] ng/g) than they were in the rice (1.2 ± 2.0 [0. 4-13.4] ng/g) and noodle (0.7 ± 0.7 [0.4-1.9] ng/g); p < 0.0001) samples. 2-OHBP levels were significantly higher in the pastry (6.9 ± 6.4 [0.7-23.0] ng/g) than in rice (0.8 ± 3.6 [0.5-25.3] ng/g; p < 0.0001) samples. 4-MBP levels were similarly significantly higher in the pastry (5.1 ± 4.2 [0. 5-14.4] ng/g) than in rice (0.1 ± 0.2 [0.4-0.9] ng/g; p = 0.0001) samples. The higher levels of BP, 2-OHBP, and 4-MBP in pastry samples than in rice and noodle samples may be attributed to the fat content of food products having a strong influence on the migration process [33]. The BP and 4-MBP levels found in this study agree with those reported in Belgium [24] and Germany [23], which were 4 and 13 ng/g. The range of BP levels found in this study was slightly higher than those in samples of cakes, pastries, rice, and noodles reported in Belgium [25] and Spain [10], which were <4-20 and <10-54 ng/g, respectively, but lower than those reported in Germany and the United Kingdom, which were 15-1559 [23], <LOD-439 [20], and 180-2000 ng/g [19]. The high levels of BP in cereal-based foods may be attributed to food contamination from packaging, printed cardboard packaging material, or additional plastic wrapping [10,20,23,32]. In this study, 19 and 6 pastry samples were packaged with PP and PET plastics, respectively, and no significant difference in the mean BPs (BP, BP-1, and 4-MBP) levels was observed among the packaging materials (all p > 0.05). However, studies with larger sample sizes are warranted to investigate the association between packaging materials and BPs. This study

Conclusions
We developed and validated the FaPEx method for the oatmeal and corn flakes [27] and rice cereal [35] samples. This is the first study in which a simple and efficient SLE UHPLC-MS/MS method was simultaneously applied for the identification of BP and nine BPs in samples of pastry, rice, and noodle. The method exhibited excellent results with the advantages of low background contamination. More specifically, LODs were below the ng/g level, and analyses of BP and BPs had high recovery and precision in the analyses of cereal-based foodstuffs. The results indicate that BP, BP-1, 2-OHBP, and 4-MBP re the most abundant analytes in pastries. The observed trace levels of 4-MBP in the samples indicate a need for the development of analytical methods with high sensitivity and specificity; the proposed method fulfills this need.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/foods11091362/s1, Figure S1: Comparison of the peak areas for BPs standards and IS (20 and 8 ng/g) across the cartridges with (A) HLB-SPE and (B) C18-SPE in three wash conditions (no wash, water, 10% MeOH) and elution solvents in MeOH and ACN (n = 5). Figure S2: Comparison of the peak areas for (A) BP and (B) 2-OHBP across six pretreatment methods in a blank pastry sample.