Methanol, n hexane, acetonitrile, acetone, dichloromethane, toluene, water and tetrahydrofuran were purchased from Merck (Darmstadt, Germany), all of which were of HPLC grade.

The standard mixture of the 16 EPA (United States Environmental Protection Agency) PAHs (PAH-mix 4S8743) was purchased from Sigma Aldrich (Bellefonte, PA). The mixture consisted of (acenaphthene, acenaphthylene, anthracene, fluoranthene, fluorine, naphthalene, phenanthrene, pyrene, benz [a] anthracene, benzo [b] fluoranthene, benzo [k] fluoranthene, benzo [ghi] perylene, benzo [a] pyrene, chrysene, dibenz [a,h] anthracene and indeno [1,2,3-cd] pyrene).

The neutralized soybean oil was bought from the Behshahr Company of Iran. The samples were stored at room temperature until analysis.

A bleaching earth (Tonsil 210 optimum) was purchased from the Kanisaz-Jam Company, Iran with the following characteristics: a chemical mixture of bentonite activated with acid; white in color with a powdery appearance; insoluble, with a moisture percentage of 9; free acidity in sulfuric acid of 0.66; density of 500; and pH of 3.

Activated carbon was purchased from Merck (Darmstadt, Germany) with the following characteristics: a chemical mixture of bituminous carbon; 0.5–2 millimeters in size with a diameter of 1.25 millimeters; moisture percentage of 2; and actual density of the particle in kilograms per liter of 0.81; a special surface area per square meter per kilogram of 1.2; the volume of the pores in milliliters per gram of 0.73, with a pore diameter of 2 nanometers.

An HPLC system (YL 9100 HPLC, South Korea) with binary gradient elution, a solvent reservoir of 1L capacity, a gradient washing program, a solvent storage tank with 1L capacity and rotary filter, a YL 9110 quaternary pump, a YL 9101 vacuum degasser, a sampler and YL 9130 temperature regulator at 25 °C, a fluorescence detector with time schedule for types of propagation waves (FP-2020), computer software for information processing (YL Clarity software program), a Waters C18 reversed-phase column, and a WAT045905 symmetry column reversed-phase 5 μm spherical silica, 4.6 mm × 150 mm (Water, Ireland) was used for chromatographic analysis. Florisil-bonded phase cartridges were Chromabond, 3 mL, 500 mg LN 0113/1 (Macherey-Nagel, Germany). The additional equipment included a rotary evaporator (Heidol, Germany), with centrifugation (Dynamc, United Kingdom) at the minimum speed of 4000 rpm, a vortex mixture (Velp Scientifica, Italy), an automatic evaporator (Heraeus, Germany) with nitrogen gas pressure at 34.5 kPa, a scale (Mettler Toledo, Switzerland), an ultrasonic bath (Elma, Germany) and an oven (Heraeus, Germany).

The bleaching process was not performed on sample a (blank), so it was the same neutralized soybean oil that was injected into the HPLC-FLD device after the process of separating the PAH compounds.

Bleaching of the other neutralized soybean oil samples was done according to the method described by Azadmard and Dutta with slight modifications (Azadmard and Dutta, 2007). Approximately 10 g of the neutralized soybean oil with 1% w/w of bleaching earth and different levels of activated carbon (0.1%, 0.2%, 0.27%, 0.3%, 0.4% and 0.5% w/w) and a standard solution of PAHs at the concentration of 200 μgkg⁻¹ were added to 250 ml round bottom flasks. Then the flasks were attached to the rotary evaporator and mixed at 85 °C. The bleaching process was carried out for 1 h at this temperature under vacuum (1000 mbar) with vigorous stirring of the mixture by the device (100 rpm). After the elapsed time, the mixture was cooled down and immediately filtered on a Buchner funnel under vacuum to remove the bleaching earth and activated carbon. The obtained samples included b¹, c², d³, e⁴, f⁵, g⁶ and h⁷.

A standard solution with the concentration of 200 μgkg⁻¹ was prepared for spiking the soybean oil samples. After the extraction and purification of PAH compounds, these compounds were analyzed by YL 9100 HPLC system using a fluorescence detector (HPLC-FLD). Also, five standard solutions of PAHs in acetonitrile with the concentrations of 0, 12.5, 25, 50 and 100 μg·L⁻¹ for the calibration curve were injected into the HPLC-FLD device.

The column temperature of HPLC-FLD was isothermal at 25 °C, injection volume was 20 μL with a flow rate of 0.5 ml·min⁻¹ and a mobile phase acetonitrile (A) and acetonitrile/water, 50:50 (B), were applied. Separation was performed using the gradient elution program on the reverse phase column C18, as described in Table1.

For the PAH determination by the fluorescence detector (FLD), the following programmed excitation and emission wavelengths (Ex/Em): were used: 270/385 nm (BαA: CHR) to 18.05 min, 256/446 nm (BβF) to 28 min, 292/410 nm (BαP) to 31.1 min (ISO 15753: 2006).

The data obtained from the results of determining the levels of PAH compounds in the soybean oil samples were analyzed. All the results were the average of three separate experiments. Linear least-square regression equations were used for calibration curves. A factorial experiment was conducted in a completely randomized design. Independent variables included the bleaching earth (1% w/w) and different levels of activated carbon (0.1%, 0.2%, 0.27%, 0.3%, 0.4% and 0.5% w/w) and the studied traits (BαA, CHR, BβF and BαP). The significance of differences among the main effects in the investigated methods was considered by the Duncan test at the significance level of α=1%. For the data analysis, MSTAT-C software (version 2.18, Michigan State University) was employed.

After passing an extraction and two clean-up steps, the analysis was performed by HPLC with fluorescence detector. Chromatograms of the samples obtained are given in Figure 1 (a, b, c, d, e, f, g and h).

Validation of the method was conducted based on the linearity, repeatability and recovery, limits of detection (LOD) and quantification (LOQ). The linearity of the method was tested separately using a standard curve design for each PAH composition. Standard solutions with the concentrations of 0, 12.5, 25, 50 and 100 μg·L⁻¹prepared for each PAH compound, BαP and Ʃ4PAHs (BαA, CHR, BβF, BαP) were injected with three replications. The mean peak area obtained for each concentration was calculated based on five-point calibration curves. The results indicated that there was a linear response with high linear regression coefficients of determination for all the 4 PAHs analyzed. (R² > 0.9950), (Table 2). In addition, linear response with high linear regression coefficients of determination for target compounds determined in previous studies was found, with correlation coefficients between 0.9929 and 0.9999. (Rojo Camargo et al., 2012; Tfouni et al., 2014; Jiang et al., 2015; Taghvaee et al., 2015; Akdogan et al., 2016). Accordingly, the developed extraction method provided reasonably good accuracy for the analysis of PAHs in the oil samples and it can be assured that linear data or regression were reliable with a probability of higher than 90% (less than 10% error).

In order to evaluate the repeatability and the recovery, blank samples of soybean oil were spiked with two levels of all PAHs (5 and 10 μgkg⁻¹). Reproducibility was evaluated by performing three analyses on the same day under the same conditions. The repeatability of the method was expressed as the relative standard deviation (RSD) in %, and recoveries ranged from 83.8% to 106.2% (RSD=3–5%) (Table 2). According to the Commission Regulation no 835/2011 on the maximum levels for polycyclic aromatic hydrocarbons in foodstuffs, the recovery range is between 50–120% and the results from this study and previous studies were within the limit set for BαP and Ʃ4PAHs. (Teixeira et al., 2007; Camargo et al., 2012; Yu et al., 2014; Taghvaee et al., 2015a,b).

The LOD and LOQ were defined as the lowest analyte concentrations and measured as the detection limit and the quantification threshold, which could generate a signal-to-noise ratio of 3.33 and 10 in the resulting chromatogram. The detection of limit and the quantification of limit were calculated by drawing a five-point standard curve using five different concentrations of PAH (0, 12.5, 25, 50, 100 μg·L⁻¹) and calculating the standard deviations of the standard samples and the calibration curve slope. The results of the detection of limit were in the range of 0.06–0.020 μgkg⁻¹ and the quantification of limit was in the range of 0.20–0.61 μgkg⁻¹ (Table 2). This result complied with the range of LODs and LOQs obtained in previous research. (EC 2011a,b; Camargo et al., 2012; Yu et al., 2014). The results showed adequate sensitivity of the method for the target compounds.

More than 90% of oil is imported in Iran and contamination levels can vary according to different sources of vegetable oils. Studies have shown that almost none of the oil refineries in Iran use activated carbon for bleaching oils due to its high cost. Therefore, this research is essential for food safety. In this research we tried to find the appropriate level of bleaching earth, especially activated carbon, to reduce or eliminate this contamination and in addition to the determination of the BαP content, Ʃ4PAHs (BαA, CHR, BβF, BαP) were evaluated in bleached soybean oil.

Commission Regulation no. 835/2011a has specified the maximum level of 2.0 μgkg⁻¹ for BαP and 10 μgkg⁻¹ for Ʃ4PAHs in oils and fats as the most suitable indicators of PAHs in food (EC, 2011a,b). In this research the average results obtained from the peak area of each chromatogram, the analyses were done at least in triplicate. Levels of PAH compounds detectable by HPLC_FLD were analyzed in the neutralized and bleached soybean oil samples ranging from 2.2 to 15.1 μgkg⁻¹ (Table 3). Based on the results, the amount of PAH compounds decreased as the result of the bleaching process with bleaching earth and activated carbon. By reducing these compounds by 100% compared to the blank sample and not detecting them by the HPLC-FLD in samples e, f, g and h, none of the target compounds (BαP and Ʃ4PAHs) were observed, indicating the positive effect of bleaching earth and activated carbon on reducing the levels of these compounds. The PAH concentrations in the neutralized soybean oil (blank) and bleached soybean oil samples were significantly different: (P < 1% according to Duncan’s test).

The results of previous studies (Teixeira et al., 2007; Camargo et al., 2012; Yu et al., 2014), and Codex Alimentarius (CAC/RCP 68, 2009) also show that the use of activated carbon in the bleaching of vegetable oils at the right dosage during the refining process can eliminate PAH contamination and the application of an HACCP system in accordance with the principles and steps as recommended by Codex is one option for reducing PAH compounds.