HPLC with Fluorescence Detection for Determination of Bisphenol A in Canned Vegetables: Optimization, Validation and Application to Samples from Portuguese and Spanish Markets

: Bisphenol A (BPA) is one of the chemicals used to produce both polycarbonate plastics and epoxy resin coatings. Research has shown that small amounts of BPA can migrate into the foods and beverages enclosed in these types of containers. In this research, an analytical method based on high-performance liquid chromatography with ﬂuorescence detection (HPLC-FLD) was developed and validated for the determination of BPA in canned vegetables. The results were conﬁrmed by liquid chromatography coupled to tandem mass spectrometry (LC-MS / MS). Attenuated total reﬂectance Fourier transform infrared spectroscopy (ATR-FTIR) was performed, to identify the coating material of each tin can. Nineteen cans of vegetables were taken as study samples (eleven samples from the Spanish market, and eight samples from the Portuguese market). Excellent linear correlation ( r 2 = 0.9999) was observed over the range of 0.01 to 0.25 mg / L. Limit of detection (LOD) and limit of quantiﬁcation (LOQ) values were calculated to be 0.005 mg / kg and 0.01 mg / kg, respectively. Good recoveries, between 72% and 90% were obtained at three di ﬀ erent levels of concentration (RSD% = 4.6). BPA was not detected in the samples. The proposed HPLC-FLD was found to be suitable for the determination of BPA in canned vegetables.


Introduction
People are exposed to contaminants through several sources. Packaging can be a potential source of contamination through the migration of substances from the packaging material into the food [1]. This fact is especially relevant, as most of the food found at the supermarkets is packaged in order to prevent its deterioration and to extend its shelf life [1][2][3] In order to ensure food safety, the monitoring of migration from food contact materials has become a priority issue. For researchers, the study of the contaminant bisphenol A (BPA) and its impact on the endocrine system are of major concern, because BPA exhibits estrogenic properties arising from its structural resemblance to the human 17β-estradiol [4] and disturbs the endocrine and reproductive systems [5][6][7]. Following the preparation of foods described by Nerín et al. (2002) [17] and because the objective of the paper was also to first screen BPA in canned vegetables, the liquid phase in contact with canned foods was analyzed. To determine the amount of BPA present in each sample, the net content of each can was leaked into a container. Then 900 µL were removed from this container and 100 µL of can was added. This mixture was filtered and then injected into HPLC-FLD and LC-MS/MS.

HPLC-FLD
The high-performance liquid chromatography (HPLC) system used in this study was an Agilent Technologies 1200 Series (Waldbronn, Germany) constituted by a quaternary pump, a degassing device, an autosampler, a column thermostat system, a diode array detector, a fluorescence detector and an Agilent ChemStation for LC and ChemStation for LC 3D Systems software (version B.04.0x).

LC-MS/MS
The liquid chromatography-tandem mass spectrometry (LC-MS/MS) system used in this study consisted of an Accela autosampler, a column thermostat system and Accela 1250 pump fitted with a degasser, coupled to a TSQ Quantum Access Max Triple Quadrupole controlled by Xcalibur 2.1.0 (Thermo Fisher Scientific, San José, CA, USA). MS data were acquired in the negative Atmospheric Pressure Chemical Ionization (APCI) mode. The operating conditions were: spray voltage 2500 V; nebulizer gas (N 2 ), 35 psi; vaporizer temperature 400 • C; capillary temperature 350 • C. As the collision gas was used Argon (1.5 mTorr). The selected precursor ion for BPA was m /z 227.2 and the two SRM transitions monitored were 227.2-212.1 and 227.2-133.2 m/z with a collision energy of 21 and 29 V respectively (Table 3).

Chromatographic Conditions
The chromatographic conditions for the HPLC-FLD and the LC-MS/MS analyses are similar. Table 2 shows the chromatographic conditions for BPA separation/quantification and Table 3 indicates the gradient elution conditions [18]. For LC-MS/MS analysis the organic solvent used was methanol instead of acetonitrile. Table 2. Chromatographic conditions for the determination of bisphenol A (BPA) as results of its migration to canned vegetables.

ATR-FTIR
Infrared spectra were acquired using an ATR-FTIR spectrometer (ATR-PRO ONE FT-IR4700, Jasco, Japan) equipped with a Diamond optical element (in the range from 4000-650 cm −1 ). ATR-FTIR spectrometer was controlled by the software Spectra Manager™ Suite (version 2). The identification was carried out by using the IR Spectral libraries of Polymers & Related Compounds (Bio-Rad Laboratories, Inc., Hercules, CA, USA) using the software KnowItAll ® (version 17.4.135.B).

Recovery
To determine recovery of BPA, a spiking experiment using a canned vegetable BPA free (blank) was performed. The spiking levels chosen were 0.01, 0.05 and 0.25 mg kg −1 of BPA, respectively. HPLC results were plotted into the BPA calibration curve. Spiking assays were performed by different operators, in three different days and the recovery was calculated by comparing the theoretical spiking and practical BPA value obtained from HPLC.

Precision
Repeatability and precision were determined by running spiking experiments on a representative sample at 3 different spiking levels (0.01, 0.05 and 0.25 mg kg −1 ). Repeatability was carried out in a single day and precision (internal reproducibility) was performed during three different days and with different operators. Data are expressed in terms of % of relative standard deviation (%RSD).
The use of a series of standards solutions of known concentration allowed to test the linearity of the method. The calibrations curves were constructed using six concentration levels (0.005, 0.010, 0.025, 0.050, 0.100 and 0.250 mg L −1 ) and they were fitted to a linear equation. The linearity was tested in acetonitrile/water (10:90, v/v) as solvent. Three calibration curves were performed in three different days. All of them have shown good linearity in the studied concentration range with coefficients of determination (r 2 ) equal to 0.9999, indicating suitability for BPA quantification in the selected calibration range. Table 4 shows the parameters of linearity. The limits of detection and quantification, calculated according to ACS guidelines [21] (defined as three and ten times the height signal of the noise level, respectively) were presented in Table 5. The method presents enough sensitivity to detect the BPA at the regulatory level, once the experimental LOD (0.005 mg kg −1 ) is considerably lower than the specific migration limit (SML) of 0.05 mg kg −1 established by the Regulation (EU) no. 10/2011 and its amendments, which are in force [4,10]. Table 5. Linearity, limit of detection and quantification of BPA according to the International Conference on Harmonisation (ICH) guidelines Q2 (R1), and using high-performance liquid chromatography (HPLC) with fluorescence detection (FLD).

Day
Equation Figure 1 shows the chromatogram obtained by HPLC-FLD with 0.005 mg kg −1 standard solution, that is the limit of detection of the method.
As an important step in the validation process, is to demonstrate that the proposed methodology is fit to its purpose. This was evaluated through recovery assays, spiking a BPA free-sample. A working session with six replicates for each spiking level was performed in three different days. According to the results report in Table 4, it is confirmed that the method internal reproducibility is appropiate even at low concentrations, with average recoveries (n = 18) ranging from 72 % to 90%. Repeatability was always lower than 3.48%, except at 0.01 mg kg −1 , which was 9.94%, and intermediate precision was always lower than 4.70%, indicating good precision. Poustka et al. (2007) [22] in their research, reported the results of bisphenols monitoring in various categories of canned food obtained from the Czech retail market through the validation of analytical procedure HPLC coupled with FLD. They obtained recoveries, not below 75%, and LOD and LOQ of 0.002 and 0.01 mg kg −1 , respectively. In its turn, Errico et al. (2014) [23] in its work performed the determination of bisphenol A (BPA) in canned tomatoes by a method based on solid-phase extraction followed by liquid chromatography coupled to fluorescence spectrophotometry obtaining a LOQ of 0.26 µg BPA kg −1 tomato. As an important step in the validation process, is to demonstrate that the proposed methodology is fit to its purpose. This was evaluated through recovery assays, spiking a BPA free-sample. A working session with six replicates for each spiking level was performed in three different days. According to the results report in Table 4, it is confirmed that the method internal reproducibility is appropiate even at low concentrations, with average recoveries (n = 18) ranging from 72 % to 90%. Repeatability was always lower than 3.48%, except at 0.01 mg kg −1 , which was 9.94%, and intermediate precision was always lower than 4.70%, indicating good precision. Poustka et al. (2007) [22] in their research, reported the results of bisphenols monitoring in various categories of canned food obtained from the Czech retail market through the validation of analytical procedure HPLC coupled with FLD. They obtained recoveries, not below 75%, and LOD and LOQ of 0.002 and 0.01 mg kg −1 , respectively. In its turn, Errico et al. (2014) [23] in its work performed the determination of bisphenol A (BPA) in canned tomatoes by a method based on solid-phase extraction followed by liquid chromatography coupled to fluorescence spectrophotometry obtaining a LOQ of 0.26 μg BPA kg −1 tomato. Figures 2 and 3 show the chromatogram obtained by HPLC-FLD with a fortification level of 0.01 mg kg −1 and a chromatogram obtained by HPLC-FLD of a sample of "sweet corn" where BPA was not detected, respectively.  As an important step in the validation process, is to demonstrate that the proposed methodology is fit to its purpose. This was evaluated through recovery assays, spiking a BPA free-sample. A working session with six replicates for each spiking level was performed in three different days. According to the results report in Table 4, it is confirmed that the method internal reproducibility is appropiate even at low concentrations, with average recoveries (n = 18) ranging from 72 % to 90%. Repeatability was always lower than 3.48%, except at 0.01 mg kg −1 , which was 9.94%, and intermediate precision was always lower than 4.70%, indicating good precision. Poustka et al. (2007) [22] in their research, reported the results of bisphenols monitoring in various categories of canned food obtained from the Czech retail market through the validation of analytical procedure HPLC coupled with FLD. They obtained recoveries, not below 75%, and LOD and LOQ of 0.002 and 0.01 mg kg −1 , respectively. In its turn, Errico et al. (2014) [23] in its work performed the determination of bisphenol A (BPA) in canned tomatoes by a method based on solid-phase extraction followed by liquid chromatography coupled to fluorescence spectrophotometry obtaining a LOQ of 0.26 μg BPA kg −1 tomato. Figures 2 and 3 show the chromatogram obtained by HPLC-FLD with a fortification level of 0.01 mg kg −1 and a chromatogram obtained by HPLC-FLD of a sample of "sweet corn" where BPA was not detected, respectively.

LC-MS/MS
In order to optimize and obtain the appropriate signal response in LC-MS/MS, acetonitrile and methanol were introduced as mobile phases. The best signal was achieved with methanol-water as a mobile phase.

LC-MS/MS
In order to optimize and obtain the appropriate signal response in LC-MS/MS, acetonitrile and methanol were introduced as mobile phases. The best signal was achieved with methanol-water as a mobile phase.
MS data were acquired in selected reaction monitoring (SRM) mode once the optimization of the MS/MS parameters was performed using the built-in perfusion system. The ion m/z 227.

LC-MS/MS
In order to optimize and obtain the appropriate signal response in LC-MS/MS, acetonitrile and methanol were introduced as mobile phases. The best signal was achieved with methanol-water as a mobile phase.
MS data were acquired in selected reaction monitoring (SRM) mode once the optimization of the MS/MS parameters was performed using the built-in perfusion system. The ion m/z 227.   In our study, any of the samples presented detectable BPA according to HPLC-FLD, therefore it was not necessary to analyse them by LC-MS/MS. However, we decided to analyse the samples by LC-MS/MS and also none of them showed detectable amounts of BPA by this technique.

ATR-FTIR
For the analysis of the cans by ATR-FTIR, small pieces of each sample (can) were cut with an area suitable to be used in the spectrophotometer. Pieces of the cover, bottom and side of each tin can were analyzed. Subsequently, each piece of the sample was analyzed by ATR-FTIR regarding their inner part. The identification was carried out by comparing obtained spectra with those of the IR Spectral libraries of Polymers and Related Compounds (Bio-Rad Laboratories, Inc.) database.
Analyzing the results listed in Table 6, it can be seen that polyester, phenoxy and epoxy coatings are those that are used in the samples. However, the polyester is the most used, as 78% of the samples have a polyester coating. Figure 5 shows, as an example, two spectra obtained for the samples 2P (phenoxy coating) and 3S (polyester coating), respectively.  Phenoxy resin has high molecular weight, a thermoplastic polymer having pendant hydroxyl groups and no oxirane ring [24]. Therefore, a broad peak of the hydroxyl group was observed at 3450.0 cm −1 in the FTIR spectrum, shown in Figure 5a [24]. In Figure 5b the band at 3454 cm −1 is related to stretching vibrations of OH groups. The bands in the range between 2800-3000 cm −1 correspond to stretching vibrations of CH groups such as CH 2 and CH 3 [25]. Weak bands at 1410 and 1462 cm −1 observed in the spectrum of polyester can be assigned to aromatic ring. The strong band at 1269 cm -1 that appears in the spectrum of polyester is due to the twisting vibration of CH 2 groups [25]. The similarity between the two spectra is because Phenoxy coating is a thermoplastic polymer derived from BPA ((CH 3 ) 2 C(C 6 H 4 OH) 2 ) and epoxy-epichlorohydrin (Cl-CH 2 -(C 2 H 3 O)) [26].  Phenoxy resin has high molecular weight, a thermoplastic polymer having pendant hydroxyl groups and no oxirane ring [24]. Therefore, a broad peak of the hydroxyl group was observed at 3450.0 cm −1 in the FTIR spectrum, shown in Figure 5a

Conclusions
An analytical method was validated, with success, for the accurate quantitation of Bisphenol A in canned vegetables by high-performance liquid chromatography with fluorescence detection (HPLC-FLD). Good recoveries were obtained showing that method is valid to determine the migration of BPA in these foods. The optimized method was also shown to be precise, and it presents a very low limit of detection (0.005 mg kg −1 ) and limit of quantification (0.01 mg kg −1 ). All the samples were analyzed by LC-MS/MS to confirm that BPA was not detected in any of them. The proposed HPLC-FLD method is an excellent tool to monitor the levels of BPA in canned vegetables due to being suitable to ensure compliance with the limit established by actual European Union legislation (SML 0.05 mg·kg −1 ). Its application was demonstrated in real samples. We are aware of the fact that the main limitation of this study is the reduced number of samples. Therefore, in the near future, we are planning to design a study with a wider number of samples. If possible, we will try also to include samples from other countries besides Portugal and Spain, in order to compare the results and be able to reach conclusions on the influence of different the legislation systems in the number of positive samples. Funding: MF Vaz acknowledges FCT, through IDMEC, under LAETA, project UID/EMS/50022/2019. The work was supported by UIDB/00211/2020 with funding from FCT/MCTES through national funds. Fernanda Vilarinho acknowledges National Institute of Health Dr. Ricardo Jorge through the project "Development of methodologies for the evaluation of polymeric food packaging components and determination of their structural and mechanical properties" (2016 DAN 1289). Antía Lestido Cardama is grateful for her grant "Programa de axudas á etapa predoutoral" da Xunta de Galicia (Conselleriía de Cultura, Educación e Ordenación Universitaria). This research was cofounded by the "Consellería de Cultura, Educación e Ordenación Universitaria, Xunta de Galicia," reference number GRC 2019/030.

Conflicts of Interest:
The authors declare no conflict of interest.