Oil samples tested in this work were prepared by the cold pressing of seeds obtained from grapes during the process of wine production. The fresh wine pomace from the red (Merlot and Hamburg) and white grape (Italian Riesling and Sila-Serbian autochthonous) varieties was gathered within a few hours after leaving the largest winery on the slopes of the Fruška Gora vineyard in the Republic of Serbia. Then, for a very short time, the pomace was purified using a machine with a rotating, perforated drum, separating the seeds from the grapes’ skins. Immediately after that, the seeds were dried in a stream of warm air at about 30 ^oC for 24 hours. The content in impurities in cleaned and dried seeds was about 0.2-0.3% (mostly parts of skins and stunted seeds). Two samples of a blend of different varieties of red, as well as white grape seeds were purchased from small wineries. In this case, wine pomace was collected over a longer period of time; the samples were dried naturally at ambient temperature (at about 20 ^oC for several days) and the seeds were separated. All grape cultivars were grown in the same geographical region with similar climatic conditions and soil characteristics, and were harvested at their best maturity. In grape seed samples prepared for pressing, by standard laboratory methods, the contents of moisture (ISO 665:2000) and oil (ISO 659:2009-Soxhlet method) were determined. The obtained results, as well as label information, are presented in Table 1.

The oil samples were obtained by pressing the cleaned and dried grape seeds with a screw press (Koprulu Machine, Type KYP20D, Turkey), run by a 1.8 kW electric motor, capacity of 5-7 kg/h in a small-size facility for cold-pressed oils. Initially, the press head was heated up to 90 °C. Subsequently, the heaters were switched off and the pressing material was released. The speed of the screw was 25 rpm. When the oil started coming out, the speed slowed down to 20 rpm. The temperature of the outgoing oil, due to pressure and friction, was about 50-55 °C. After pressing the oil samples were stored for three days at room temperature for natural sedimentation of insoluble impurities to occur, after which they were decanted. Finally, clear oil samples for further analysis were used as average samples from several series of pressing. The samples (each of 1000 mL) were kept in a refrigerator at 8 ± 1 °C until analysis to avoid oxidation processes (Bjelica et al., 2019).

Free acidity, given as acid value (AV) (ISO 660:2009), expressed in mg KOH/g, was determined by volumetric titration of a solution of oil dissolved in ethanol:diethyl ether (1:1) with an ethanolic solution of 0.1 M KOH.

The peroxide value (PV) (ISO 3960:2017), expressed in meqO₂/kg, was determined by the reaction of oil (dissolved in acetic acid and isooctane) with a solution of potassium iodide. The liberated iodine was then titrated with a standard volumetric sodium thiosulfate solution.

The p-anisidine value (p-AnV) was determined following the ISO standard spectrophotometric method (ISO 6885:2016), measuring the absorbance at 350 nm.

The Totox value was calculated as follows: Totox = 2 PV + p-AnV

The content of total tocopherols and tocotrienols was determined by using the spectrophotometric method according to Paqout et al., (1967). The method is based on the reduction properties of tocophenols/tocotrienols, which reduce Fe³⁺ from FeCl₃ in the presence of 2-2’dipyridyl and create a red-colored complex with Fe²⁺ ions. The determination was made from unsaponified matter, so the oil saponification and the extraction of unsaponifiable matter was performed, and then a benzene solution of the unsaponified matter developed a colored reaction, where absorbance was measured at 520 nm wavelength using a spectrophotometer with a blank test.

The determination of tocopherols was carried out using detector RF-535 (Shimadzu, Fluorescence HPLC) coupled to HPLC (Waters M600E, USA) on a reversed-phase Nucleosil 50-5 C18 column (Macherey-Nagel, Germany), using a method based on the procedure described by Carpenter (1979) with some modifications. Operating conditions of the HPLC system were the following: mobile phase: 95% CH₃OH; flow rate: 1.0 mL/min.; column temperature: 20 °C; wavelength: λex = 295 nm, λem = 330 nm. The relative retention time and maximum values of absorption at the given relative retention time were used for the identification of tocopherols in the oil samples. Commercial tocopherol standards were suitably diluted and were used for method validation (solution series from 0.01 to 0.5 μg/mL), as well as for quantification purposes (solution series from 1.0 to 20.0 μg/mL).

Vitamin E activity, i.e. vitamin E equivalent of tocopherols (Mag et al., 2002) was calculated following the formula:

The content in total phenolic compounds was estimated according to the Folin-Ciocalteu colorimetric method (based on the reaction of the reagent with the functional hydroxyl groups of the phenols) as described by Haiyan et al. (2007). The oil was dissolved in hexane and extracted with methanol (three times). The sample was left to stand overnight. After that, the methanolic extract was washed with hexane and an aliquot (1 mL) was transferred to a volumetric flask (10 mL), to which the Folin-Ciocalteu reagent (0.5 mL) was added. The solution was shaken and left to stand for 3 min. prior to the addition of a saturated solution of sodium carbonate, and dilution with water was made. After 1hour, the absorbance at 725 nm was measured against a reagent blank using a spectrophotometer. Calibration was performed using gallic acid. The total phenol content was expressed as the equivalent of gallic acid in mg per kilogram of oil.

Determination of single polyphenols. Phenols from the grape seed oil samples were extracted using a modified procedure by Gouvinhas et al., (2014) and determined on Agilent Technologies 1200 HPLC-DAD, coupled with Agilent Technologies 6410A ESI-QqQ-MS/MS, described previously (Bjelica et al., 2019). Approximately 800 μL of the sample were transferred to the 4 mL vial; the mass was accurately measured, and the sample was diluted with 400 μL of hexane. The mixture was extracted with 80% methanol (600 μL) with vigorous shaking using the vortex device. After centrifugation (10 minutes at 2500 rpm), the aqueous methanol layer was transferred to a 2 mL vessel, and the oil layer was extracted two more times with 80% methanol. The combined extracts were supplemented with 80% methanol to 2 mL, filtered through a 0.45 μm regenerated cellulose membrane filter and analyzed using LC-DAD-MS/MS. The content of the selected secondary biomolecules was determined by the LC-DAD-MS/MS method. A 5 μL sample was injected. The separation was performed by the Zorbax Eclipse XDB-C18 column (50 mm x 4.6 mm, 1.8 μm) at 50 °C. The components were eluted with a mobile phase based on 0.05% aqueous solution of formic acid (A) and methanol (B) at a flow rate of 1.0 mL/min, in gradient mode: 0 min., 30% B, 6 min. 70% B, 9 min. 100% B, 12 min. 100% B; re-equilibrium time 3 min. For identification purposes and identity validation, the UV/VIS signal was observed in the range of 190-700 nm. Effluents passed the MS/MS detector without flow sharing. The ion source parameters were: the pressure of the nebulizer 50 psi, temperature and flow of drying gas (N₂) 350 °C and 10 L/min.; the voltage on the capillary 4000V; negative polarity. The compounds were monitored in a dynamic SRM (selected reaction monitoring) mode, with optimized parameters. By sequential dilution of 1:1, a series of commercial standard solutions were prepared in a concentration range of 1.53 ng/mL to 25.0 μg/mL. The concentration of the analyte was determined by the external standard method, using calibration curves in a narrower band-adjusted concentration in the sample. For the quantification, points that directly surrounded the readings in the samples were taken. In most cases, the correlation was not linear but square. For all the compounds, peak areas were determined using Agilent MassHunter Workstation Software - Qualitative Analysis (ver. B.03.01). Calibration curves were plotted and sample concentrations calculated by using the OriginLabs Origin Pro (ver. 8.0) software.

The antiradical capacity of oils in the current study was assessed by the evaluation of the free radical scavenging effect on 2, 2-diphenyl-1-picrylhydrazyl radicals (DPPH•), according to the method proposed by Martínez and Maestri (2008). The absorbance was measured on spectrophotometer at 515 nm using toluene as the blank. Antiradical action toward DPPH radicals was estimated from the difference in absorbance with or without oil sample (control) and the percent of inhibition was calculated from the following formula:

where: A_C = absorbance of the control; A_S = absorbance of the sample.

EC₅₀ (mgoil/mgDPPH•) values were expressed as the inhibitory concentration of the oil necessary to decrease the initial DPPH• absorbance by 50%. A lower EC₅₀ value indicates a higher antiradical capacity.

All above-mentioned spectrophotometric measurements were conducted using a UV/VIS spectrophotometer model T80+ (PG Instruments Limited, London).

The solvents and chemicals used in these investigations were of analytical grade, except chemicals for the extraction procedure and chromatographic analysis, which were of chromatographic purity. Methanol, ethanol, KOH, pyrogallol and n-hexane were purchased from Merck (Darmstadt, Germany); Folin-Ciocalteu reagent and gallic acid were from Institute Mol Ltd.; formic acid was from J.T. Baker (Deventer, Netherlands), as well as other solvents and reagents. Standards were purchased from Sigma-Aldrich Co. (St. Louis, USA).

All results are presented as a mean value ± standard deviation (n = 3). One-way analysis of variance (ANOVA) with a Duncan test was used to determine significant differences among the data at the level of p < 0.05. Also, two-way ANOVA with the Tukey test was performed at a statistically significant level (p < 0.05). The degree of linear relationship between two variables was measured using the Pearson’s correlation coefficient (r). Statistical analysis was performed using the software SPSS Statistica 20 (IBM Corporation, Armonk, New York, U.S.) and OriginPro 8 (OriginLab Corporation, Northampton, MA 01060 USA).

Some quality parameters of grape seed oil samples, like AV, PV, p-AnV and Totox, are presented in Table 2.

The acid value measures the content in free fatty acids formed by the hydrolytic degradation of oil molecules. Free fatty acids can contribute to the reduction in the shelf-life of oil, so acidity is a very important quality parameter (Prescha et al., 2014). Table 2 shows that all oil samples had low acidity values, below the maximum level (4 mgKOH/g) allowed by Codex Alimentarius (1999). Similar values have been reported by other authors (Pardo et al., 2009; Boso et al., 2018). The lowest acid value was found in the oil of the red grape Merlot variety (0.62 ± 0.03 mgKOH/g), as well as in the oil of the Serbian autochthonous white grape Sila variety (0.66 ± 0.03 mgKOH/g). The acidity of the oils was slightly higher in the two other grape varieties (Hamburg - 0.94 ± 0.04 and Ital. Riesling - 0.73 ± 0.03 mgKOH/g); however, the oils extracted from a blend of different red (R) and white (W) grape varieties had increased acidity, 1.61 ± 0.05 and 2.42 ± 0.28 mgKOH/g, respectively. An explanation for the low acid values can be the fact that grape cultivars were harvested at their highest level of maturity. Namely, Rubio et al., (2009) revealed that the oil from the grape seeds collected at an early stage of growth had a considerably higher free acidity value compared to the oil extracted from grapes collected at later stages of maturity. Furthermore, the pomace from a large winery company was dried quickly, in contrast to collecting and drying pomace from small wineries. Careful and fast drying of the pomace after production of wine is required to achieve high valuable grape seed oils. Otherwise, the quality of the resulting oil is affected by various microorganisms which enhance acidity (Matthäus, 2008).

PV defines the contents in lipid hydroperoxides in oils formed during auto- and photo-oxidation processes. All investigated oil samples were characterized by such values for PV that did not exceed the limit for cold-pressed oils of 15 meqO₂/kg, recommended by Codex standards for vegetable oils (Codex, 1999). The PV for the oil from red grape seeds were lower and ranged from 1.90 ± 0.04 (Hamburg) to 8.14 ± 0.07 meqO₂/kg (R-blend), in contrast to the oil from white grape seeds with a level of 5.08 ± 0.21(Ital. Riesling) to 10.08 ± 0.18 meqO₂/kg (W-blend). Regarding the PV, it can be concluded that there were statistically significant differences among all the oil samples at p < 0.05.

The data published for the PV of grape seed oil commercially available in the UK was 1.0 ± 0.0 meqO₂/kg (Madawala et al., 2012). In pressed oil obtained from different red grape varieties grown in Spain, Pardo et al., (2009) found PV values from 5.99 meqO₂/kg (hot air drying of seed, at 50 ^oC for 6h) to 13.50 meqO₂/kg (ambient temperature dried seeds, at 20 ^oC for 7 days). In oils obtained from cold extraction of bagasse-derived seeds from the white grape varieties, PV values were in the range of 11 ± 1.69 to 26 ± 2.89 meqO₂/kg (Boso et al., 2018). Kiralan et al., (2019) concluded that the PV values for fresh cold-pressed oils were 1.82, 12.2 and 26.5 meqO₂/kg for flaxseed, grape seed and black cumin seed oils, respectively. The PV value of commercial cold-pressed grape seed oil from the European market by Radočaj and Dimić (2013) was 5.07 ± 0.05 meqO₂/kg.

The p-AnV reflects the contents in secondary lipid oxidation products (saturated and unsaturated carbonyl compounds), resulting from the decomposition of hydroperoxides, so p-AnV along with PV, indicate the progression of rancidity in the oils. The p-AnV in all the analyzed oil samples (Table 2) were in the range of 0.56 ± 0.03 to 2.14 ± 0.06; whereby the lowest p-AnV was found in the oil of the red grape Hamburg variety. Generally, all three red grape seed oils had somewhat lower p-AnV in relation to oils from white grape seeds. The p-AnV of the oils had the following values: Merlot (0.64 ± 0.03) and Hamburg (0.56 ± 0.03) (without significant differences between these two samples, p < 0.05), < R-blend (1.78 ± 0.05) and Sila (1.83 ± 0.05) (without significant differences, p < 0.05), < W-blend (1.94 ± 0.04) and < Ital. Riesling (2.14 ± 0.06).

The food safety of these oils, with respect to the p-AnV value, is difficult to evaluate because of the lack of an established limit value for this parameter in cold-pressed oils (recommended p-AnV value for fresh fully refined oil is <10). However, cold-pressed seed oils, as a rule, have very low p-AnV, indicating low levels of carbonyl compounds because of the absence of heat during the oil extraction process. The mean values of p-AnV in a number of fresh cold-pressed oils from the seeds and fruits of different plants (other than grape seed oil) found in the Polish market (Prescha et al., 2014) were in the range of 0.25 ± 0.05 to 8.60 ± 3.20. According to their results, the lowest p-AnV was found for sesame oil (0.2-0.3), but the largest variability occured in rose hip oil (4.22-11.88) and pumpkin seed oil (1.48-8.17).

The total secondary oxidation products, i.e. Totox, in our samples, measured as a sum of PV and p-AnV, ranged from 2.45 ± 0.09 to 12.03 ± 0.38. The Totox value for oils from the red grape Merlot and Hamburg varieties were the smallest; whereas they were 2 to 3 times larger in the oils of white grape Ital. Riesling and Sila varieties. The largest Totox values of 10.24 ± 0.53 (R) and 12.03 ± 0.38 (W) were found in the oils from blended seeds, which may be a reflection of pomace collection and the drying process. In the reviewed literature, data on the Totox values for grape seed oil were very limited. In grape seed oil, as a specialty oil from a local retail store in Reading (UK), the Totox value was 17.5 ± 0.1 (Madawala et al., 2012), but there was a lack of accurate background information on the sample.

As cold pressing does not involve chemicals or heat treatment during the oil production process, it implies that cold-pressed seed oils may contain phytochemicals, as well as natural antioxidants, more than refined oils. Some results from our previous work about total tocopherol and phenol contents and their fractionation (Bjelica et al., 2019) have been included to support this new investigation.