30 L of integral CPO (a mixture of liquid and solid phases) industrialized at Nazaré city (state of Bahia, Brazil) and conditioned in two tinplate cans with a capacity of 15 L each were purchased at São Joaquim Fair, in Salvador (Bahia). This oil was placed in a stainless steel recipient and heated up to 45 °C, allowing oil fusion and homogenization to be used in frying. At first, 20 mL (time 0) of it were filtered in a glass wool filter, inerted with nitrogen gas and stored in an amber bottle at –20 °C (Jorge and Gonçalves, 1998) and thawed for analysis (initial sample, CPO).

The preparation of akara was performed by a baiana de acarajé according to her traditional practice. 5 kg of crude mass of akara were bought daily from her traditional source at the same fair. Frying was always performed outdoors. First the raw mass was placed in an aluminum pan and seasoned with grated onions and salt, then whipped and shaped into balls (60–80 g each) with a wooden spoon. Then, 5 L of CPO in an enameled pan were heated for 12 minutes in the presence of an onion (baiana's traditional practice – an intention of making the oil quality lasts longer the oil quality longer during frying). Subsequently, 5 of those balls were successively added to the oil and deep-fried during approximately 6 minutes, with some intervals in which the oil remained with only the onion, while the baiana whipped the mass again, and a total of 110 akaras/day was obtained – simulating akara's average quantity at the baiana's point of sale (Rogerio, 2010).

The total frying time per day was 5 intermittent hours and the frying process was performed on 5 consecutive days – totaling 25 hours. The temperature ranged from 143 to 188 °C in the beginning (day 1) and 159–178 °C in the end (day 5).

Every frying day began with 5 L in the enameled pan. On day 1, only fresh oil was used, and on subsequent days a mixture of the previous day's oil (used oil) was made with fresh oil. The oil turnover was performed daily with 2 L at the most. On days 1 and 2 only fresh CPO was added, and on the other days, the replacement was made with a mixture of fresh plus used oil, according to the baiana's techniques.

Then, at the end of each frying day, the oil was left undisturbed in a pan with a lid, at room temperature, until the food waste settled, and was filtered afterwards, stored until the next frying day and the same procedure was repeated over four subsequent days, for a total time of 25 hours. At the end of each day and the frying process (25 hours), a new aliquot of 20 mL was filtered in a glass wool filter, obtaining the last sample (FPO of frying) and the others for TPC (%) content analysis.

Total polar compound (TPC) content was determined gravimetrically according to a mini column method described previously, with slight modifications (Dobarganes et al., 2000b). In brief, approximately 0.5 g of crude palm oil were dissolved in the elution solvent and introduced into a glass column filled with a silica gel slurry and elution solvent. This elution solvent was a mixture of light petroleum (b.p. 40–60 °C) and diethyl ether 94:6 (v:v). A chromatographic glass column with an internal diameter of 1 mm and a length of 15 mm, containing 5 g of silica gel (with a particle size 0.063–0.200 mm and 70–230 mesh) adjusted to a water content of 5% was used. Non-polar compounds were eluted with 60 mL of elution solvent and the polar compound fraction was eluted with 50 mL diethyl ether. A dropping funnel was used and the flow rate was adjusted to approximately 1.5 mL·min⁻¹. Solvent was removed by rotary evaporation and the flask was flushed with a stream of nitrogen to ensure dryness. The completeness of fractionation was evaluated by analytical thin-layer chromatography (TLC) with an elution system of petroleum ether:diethyl ether:acetic acid (70:40:1; v:v:v).

The reported data of TPC were obtained from duplicate measurements of each sample and expressed in terms of the mean and standard error (SE). A regression analysis was also applied to these data.

The mutagenic evaluation of CPO and FPO was performed by the pre-incubation method of Salmonella/microssome assay as described by Maron and Ames (1983), with slight modifications (Mortelmans and Zeiger, 2000). Briefly, 100 µL of different concentrations of CPO or FPO (all dilutions were in DMSO) and 100 µL of stationary growth cultures of Salmonella Typhimurium strains TA97, TA98, TA100, and TA102 were added to 500 µL of 0.2 M of a sodium phosphate buffer (pH 7.4) or to 500 µL of an S9 mix. After pre-incubation of the mixture at 37 °C for 30 min, 2 mL of top agar were added to it and the contents poured onto a minimal glucose agar plate. The plates were incubated at 37 °C and the numbers of His⁺ were scored after 72 h. Each assay (in triplicate) was repeated twice to confirm the results.

Aroclor 1254-induced rat liver S9 fraction was purchased from Molecular Toxicology Inc. (Moltox™, USA). The metabolic activation mixture (S9 mix) was prepared according to Maron and Ames (1983) and used as described earlier (Aiub et al., 2003). The criterion for a positive mutagenic response was the number of reverting colonies in the test assay amounting to at least twice the number of spontaneous reverting (Mutagenic Index, MI ≥ 2) (Maron and Ames, 1983).

The mutagenic response was considered positive when the number of revertant colonies in the test was at least twice the number of spontaneous revertants and no cytotoxicity was detected (survival rates <70%). Significant statistical differences between negative and tested concentrations under the same experimental conditions were verified by statistical analysis (Student t-test, p ≤ 0.05) (Stankevicins et al., 2008).

To determine the citotoxic effect of CPO and FPO, samples from the pre-incubation assay mixture in the Salmonella/microssome assay were diluted in 0.9% NaCl (w/v) to obtain a suspension containing 2×10² cells·mL⁻¹. An aliquot of 100 µL of this suspension was poured onto a nutrient agar plate. The plates were incubated at 37 °C for 24 h and colonies were counted. Citotoxicity was considered positive when the cell survival rate was <70% of the rate observed for the negative control (Aiub et al., 2003).

A murine macrophage cell line was used from a culture at confluence and was incubated in an atmosphere of 5% CO₂, at a temperature of 37 °C. Cell detachment was carried out mechanically by scraping. After that the suspension was centrifuged at 5000 rpm for 5 min and cells were re-suspended in Modified Eagle Medium (MEM) Ca²⁺, 1.8 mM (Gibco^®) supplemented with 10% fetal bovine serum, 1.76 g·L⁻¹ of NaHCO₃, 0.88 g.L⁻¹ of pyruvate, 22 mg·L⁻¹ of aspartic acid and 17 mg·L⁻¹ of _L-serine. Cell viability was determined after trypan blue staining and the cells were plated on 24-well plates at a density of 2×10⁵ cells/well and then incubated for 24 h.

The micronucleus assay was performed as previously described (Aiub et al., 2011), with modifications. Cells were treated with 100 µL of various dilutions of CPO or FPO in DMSO, corresponding to 10% of the total volume of each well and then the plates were incubated for 24 h. After this period, the medium was removed from the wells, cells were washed with 1 mL of MEM Ca²⁺, 1.8 mM and re-incubated, and after that with 1 mL of MEM Ca²⁺, 1.8 mM, supplemented with fetal bovine serum for 24 h in an atmosphere of 5% CO₂. N-methyl-N-nitro-N-nitrosoguanidine (MNNG) 0.5 µM was used as positive control. The medium was then replaced with cold Carnoy's fixative (3 parts methanol to 1 of part glacial acetic acid) for 15 min. Fixed cells were washed with McIlvaine buffer (21 g·L⁻¹ of citric acid and 36 g·L⁻¹ of Na₂HPO₄, pH 7.0) for 2 minutes and left to dry at room temperature. Afterwards, the cells were stained with 4’-6-diamidino-2-phenylindole (DAPI) (0.2 µg·mL⁻¹), dissolved in McIlvaine buffer for 40 min. Then, the cells were washed twice with McIlvaine buffer and left to dry again at room temperature.

To determine the mitotic index and number of micronucleated cells, as well as percentages of necrotic and apoptotic cells, 1000 cells per concentration were analyzed in a fluorescence microscope (Reichert Univar) at an excitation wavelength of 350nm. This experiment was conducted in quintuplicate. The data were analyzed using a one-way ANOVA and the Tukey-Kramer multiple comparison test. Results were considered to be statistically significant at p < 0.05 (Aiub et al., 2011, Carpes et al., 2013). GraphPad Instat^® software, version 3.01 (GraphPad Software, Inc., USA) was used for all analyses.

There was no significant (p > 0.05) increase in the number of reverting colonies at any of the tested concentrations of CPO and FPO over the respective negative control plates in all four tested strains, both with or without the S9 mix (Tables 1 and 2).

The results from the assay with CPO and FPO are presented in Table 3. Both samples did not induce micronuclei formation, apoptosis or necrosis on macrophage cells. Moreover, they did not induce an increase in mitotic division. There was no significant difference between the results obtained with the oils sampled before and after frying (p < 0.05).

TPC contents increased linearly with frying time with high correlation coefficients (R²=0.96; TPC = 15.24 + 0.6 time) (Table 4). After 25 hours of heating, the oil showed around 29% polar compounds, above the established limit in most of the current regulation regarding the discharging of frying oils and fats for human consumption (Berger, 2005; Saguy and Dana, 2003). It was observed that the values were higher from 15 h of frying (Table 4), due to the greater heating time and replenishment with reused oil.

Crude palm oil is largely used in Brazilian cuisine and it is essential in akara deep frying. The latter is now regarded as one of Brazil's immaterial national treasures (IPHAN, 2005) and a cultural and touristic icon in the city of Salvador (Bahia, Brazil). During deep-frying, the high temperatures used in the presence of oxygen and water induced important chemical changes in the palm oil, inevitably reducing its shelf life and directly affecting its nutritional characteristics (Matthäus, 2007). Therefore, it is necessary to carry out toxicological studies to determine the security of the CPO used in this process.

No mutagenic or citotoxic activity were detected in the CPO used in the traditional akara frying process in all tested strains (Tables 1 and 2). Furthermore, it was observed that CPO did not induce genotoxic potential in the micronucleus assay in macrophages (Table 3). These results were contrary to the expected ones since the TPC content in the analyzed frying oil sample (25 h) exceeded the considered safe limit for health (Table 4) (Berger, 2005). Nevertheless, some researches had found no mutagenic activity in deep-frying oils (Scheutwinkel-Reich et al., 1981; Taylor et al., 1982; Van Gestel et al., 1984), and only Fong et al. (1980) had observed mutagenicity in peanut oil, however it had been correlated to an aflatoxin B1 contamination in the oil.

As already mentioned, TPC include TGP, TGD, oxTGM, DG, and FFA (Dobarganes et al., 2000b). As DG is measured as a fraction of TPC, it may be misleading to use TPC in judging oil quality. The higher amounts of total polar compounds in palm oils are mainly due to a higher level of DG contents (4.0–7.5%) compared to other vegetable oils (2–3%) (Berger, 2005; DeMarco et al., 2007). Otherwise, the CPO processing in Bahia displays high FFA %, a consequence of the poor condition of the oil extraction (Almeida et al., 2013). Thus TPC elevation in samples (CPO and FPO) may be due to the predominance of such fractions which are not associated with the toxicity of heated oils (González-Munõz et al., 1998; Morita et al., 2008).

One unique characteristic of crude palm oil is its high content of carotenoids and tocopherols. Carotenoids, which impart to CPO its distinctive orange-red color, together with tocopherols and tocotrienols contribute to stability. A research made by our group demonstrated that the mean of carotenes in the industrial CPO processed in Bahia were 526.77 µg·g⁻¹ (Almeida et al., 2013) and this carotenoid value is extremely high when compared to other oils: sunflower (8.5±0.7 mg·L⁻¹) and extra virgin olive oil (11.0±0.3 mg·L⁻¹) (Andreu-Sevilla et al., 2008). On the other hand, CPO contains unusually high levels of tocotrienols (Norhaizan et al., 2013; Berger, 2005). Most other common vegetable oils have negligible contents of tocotrienols (Lin, 2011). In CPO, the presence of these antioxidants seems to offer some oxidative protection to the oil through a mechanism in which they are oxidized prior to triglycerides (Lin, 2011). Together with CPO constitution factors, the baiana de acarajé practice of heating oil in the presence of a whole onion considerably preserves the oil quality with a significant reduction in the levels of lipid peroxidation indicators as well as a reduction in tocoferol loss (Oyewole and Olayinka, 2007).