High Voltage Electrical Discharges as an Alternative Extraction Process of Phenolic and Volatile Compounds from Wild Thyme (Thymus serpyllum L.): In Silico and Experimental Approaches for Solubility Assessment

The objective of this study was to evaluate the potential of green solvents for extractions of bioactive compounds (BACs) and essential oils from wild thyme (Thymus serpyllum L.) using theoretical and experimental procedures. Theoretical prediction was assessed by Hansen solubility parameters (HSPs) and conductor-like screening model for realistic solvents (COSMO-RS), to predict the most suitable solvents for extraction of BACs. An experimental procedure was performed by nonthermal technology high voltage electrical discharge (HVED) and it was compared with modified conventional extraction (CE). Obtained extracts were analyzed for chemical and physical changes during the treatment. Theoretical results for solution of BACs in ethanol and water, as green solvents, were confirmed by experimental results, while more accurate data was given by COSMO-RS assessment than HSPs. Results confirmed high potential of HVED for extraction of BACs and volatile compounds from wild thyme, in average, 2.03 times higher yield of extraction in terms of total phenolic content was found compared to CE. The main phenolic compound found in wild thyme extracts was rosmarinic acid, while the predominant volatile compound was carvacrol. Obtained extracts are considered safe and high-quality source reach in BACs that could be further used in functional food production.


Introduction
Nowadays, there is growing interest among consumers in functional food production. Functional foods can be defined as natural dietary items that, besides providing nutrients and energy, have health-promoting, disease-preventing, or medicinal properties [1]. Spices and aromatic herbs have  Relative energy difference (RED): very good solubility 0-1 (green color); medium solubility 1-3 (yellow color); poor solubility > 3 (red color). CPME-cyclopentyl methyl ether, DMC-dimethylcarbonate, MeTHF-2-methyltetrahydrofuran.  The qualitative modelling using principal component analysis (PCA) to identify potential grouping and correlations between HSPs and COSMO-RS results was performed. PCA graph including both components and HSPs and COSMO-RS variables is given in Figure 1.

Experimental Procedure
Extracts obtained by CE and HVED were analyzed for changes during the extraction, therefore, physical and chemical parameters of extracts were measured. Physical parameter results, including pH, conductivity, temperature and power used for HVED treatment, are given in Table 3. Also, cell disintegration index (Zp) was calculated.

Experimental Procedure
Extracts obtained by CE and HVED were analyzed for changes during the extraction, therefore, physical and chemical parameters of extracts were measured. Physical parameter results, including pH, conductivity, temperature and power used for HVED treatment, are given in Table 3. Also, cell disintegration index (Z p ) was calculated.
Furthermore, chemical analyses of wild thyme extracts were performed. In order to compare extraction efficiency of BACs by means of HVED, compared to CE, obtained extracts were analyzed by spectrophotometric methods-total phenolic content (TPC) and antioxidant activity (2,2-diphenyl-2-picrylhydrazyl (DPPH) free radical assay and ferric reducing antioxidant power (FRAP) assay). Results are given in Figure 2. Extraction yields are presented as g GAE/g of sample × 100 and expressed as percentages (%).
In order to quantify individual phenolic compounds, an ultra performance liquid chromatographytandem mass spectrometry characterization (UPLC-MS/MS) analysis was performed. Wild thyme extracts were analyzed for apigenin, carnosol, diosmetin, hydroxytyrosol, luteolin, oleanolic acid, quercetin, rosmarinic acid, p-cymene, thymol, carvacrol and camphor, results are given in Table 4. In order to quantify individual phenolic compounds, an ultra performance liquid chromatography-tandem mass spectrometry characterization (UPLC-MS/MS) analysis was performed. Wild thyme extracts were analyzed for apigenin, carnosol, diosmetin, hydroxytyrosol, luteolin, oleanolic acid, quercetin, rosmarinic acid, p-cymene, thymol, carvacrol and camphor, results are given in        The analysis of volatile compounds from wild thyme extracts was performed by HS-SPME/GC-MS method. Analyzed compounds included 1,8-cineol, linalool, thymol and carvacrol, and results are shown in Table 5. Statistical analysis was performed in in XLStat (MS Excel 2010). An analysis of covariance (ANCOVA) was used to analyze the impact of independent variables: treatment time, voltage, ethanol content and treatment type (CE, HVED-nitrogen or HVED-argon) to measured physical and chemical parameters. For UPLC-MS/MS and GC-MS results, all individual (data not shown) and total of all compounds was assessed. The p-values present the statistical significance of each of the factor and it was significant at p ≤ 0.05, and are shown as bolded in Table 6. Furthermore, in order to assure safe and high-quality final product, residue levels of pesticides and metals has been analyzed. At the moment, trace content of metals and pesticides in foods are regulated via European Commission Regulations No 1881/2006 and No 396/2005, respectively. Regulations for dietary supplements were chosen as the nearest category to evaluate health safety requirements for wild thyme and provide maximum residue levels (MRLs). Results of analyzed pesticides levels in dried wild thyme is given in Table 7.  Results of residue levels of metals found in dried wild thyme and selected extracts are given in Table 8. Due to possible abrasion of electrodes during the treatment with HVED, content of metals was analyzed also in wild thyme extract obtained by HVED. For this purpose, extract TN8 was chosen that was treated with highest voltage (25 kV), nitrogen, for longer time (9 min) and highest ethanol content (50%), since highest abrasion is expected in such extract.

Discussion
The extraction of BACs and volatile compounds from wild thyme was assessed by theoretical and experimental procedures. Theoretical prediction was calculated by computational simulation methods using HSPs and COSMO-RS software, while experimental analysis was done by HVED as a nonthermal extraction technology. For better comparison, both methods were performed at room temperature.

Wild Thyme Compounds Solubility in Green Solvents by HSPs
The solubility of wild thyme compounds in selected green solvents assessed by HSPs is given in Table 1 and presented as a comparison with conventionally used n-hexane (first column). Results are shown as a RED value that estimates the potential of a solvent to dissolve solutes. RED values from 0-1 present very good solubility, medium solubility is 1-3, and poor solubility > 3. It is clear that some alternative (green) solvents have better solubility then n-hexane and for that reason, it is not the best solvent for extraction of wild thyme BACs from theoretical perspective. According to solvent with highest number of very good soluble compounds, followed by medium and poor solubility, the order of solvents that could be used for better solubility of wild thyme compounds was as following: CPME It can be concluded from the results that wild thyme compounds have the lowest solubility in water, followed by primary alcohols (ethanol, 1-butanol and methanol) and secondary alcohol isopropanol. On the other side, these solvents are the cheapest green solvents that is the great advantage for their usage. CPME showed the highest theoretical prediction for solution of wild thyme compounds.
A classical rule "like dissolves like" can be applied to theoretical prediction for solvent-solute solubility. According to this rule, more polar solvents will have higher probability of solubility for polar compounds, while non-polar solvents will have a tendency to dissolve less polar or non-polar solutes. This was shown with HSPs results, where non-polar (or weakly polar) solvents like n-hexane, limonene, CPME, methylacetate, ethyloleate, cymene, β-myrcene and α-pinene had higher probability of solubility of less polar compounds such as hydrophobic thymol, α-thujene, piperitone and other.

Wild Thyme Compounds Solubility in Green Solvents Assessed by COSMO-RS Software
The COSMO-RS software was also used for evaluation of solubility of wild thyme compounds in selected green solvents. Results are presented in probability of solubility (%) where low probability is marked with red color, medium probability with yellow color, and high probability of solubility with green color. Similar trend was shown as with HSPs, but higher potential for solution of solutes was predicted with COSMO-RS software. Results showed that MeTHF has the highest probability of solubility of wild thyme compounds, and that all solvents, except water, have higher potential for solution during extractions with wild thyme, compared to n-hexane. Solvents are ranked as follows: These results show the solubility of each individual compound in the solvent, but the extraction does not depend only on solubility, but also on the quantity of each compound in the plant. Therefore, experimental analysis was performed. Taking into consideration results obtained by two theoretical prediction methods, it was decided to perform extraction using green solvents with potential to replace n-hexane: water, ethanol, limonene, α-pinene, DMC and ethyl acetate. Since no electrical discharge was achieved with limonene, α-pinene, DMC and ethylacetate, for further analysis, only water and aqueous ethanol 25% and 50% (v/v) were taken.

Principal Component Analysis (PCA) of Theoretical Prediction Results
The PCA analysis was conducted in order to correlate the results from HSPs and COSMO-RS software. The biplot of PCA assessment is given in Figure 1. A total of 52.37% of variance was clarified in the observed data set. The biplot showed that analysed compounds have spread over all quadrants, while HSPs and COSMO-RS parameters were placed in first, second and third quadrants. A grouping of selected compounds from the same group (i.e., oxygenated monoterpenes, sesquiterpenes, flavonons, etc.) is notable, where all compounds from the same group were grouped in the same quadrant or at least, at the same side of x or y axis, except monoterpenes that were placed in first, second and fourth quadrants. Additionally, grouping of solvents was also notable. For example, short chain alcohols (ethanol, methanol, 1-butanol and isopropanol) were placed all in the third quadrant according to HSPs and in second and third quadrants according to COSMO-RS results. All monoterpenes, regarding the analysed method, were placed in the first quadrant of the PCA biplot.
It is clear that neither of solvents showed completely different results with two theoretical prediction models, i.e., no solvents were placed in inversely proportional quadrants (I-III or II-IV). Furthermore, most solvents were placed in the same quadrant with HSPs and COSMO-RS prediction including n-hexane (1), methyl acetate (3), ethyl oleate (4), ethanol (5), methanol (8), limonene (9), α-pinene (10), cymene (11), β-myrcene (12), and CPME (13). For that reason, it can be concluded that similar data are given with HSPs and COSMO-RS results, although some differences could be noted, especially with higher probability of solution given with COSMO-RS data ( Table 2).

Physical Properties of Wild Thyme Etracts Obtained by HVED Compared to Conventional Extraction
The aim of the experimental analysis was to extract BACs from wild thyme by HVED extraction. During the HVED treatment, a gas is flowing through the needle (electrode) and it is being ionized forming a cold plasma. The type of gas used in the treatment since different gas ionize at different voltages and different radical species are being formed during discharge. Therefore, different results could be obtained with different gases [31]. For that reason, it was difficult to obtain electrical discharges with nitrogen under 20 kV so 20 and 25 kV were chosen for nitrogen treatments, while lower voltages of 15 and 20 kV were chosen for argon treatments.
Results of physical parameters, including pH, conductivity, temperature before and after the treatment and power used during HVED treatment, are given in Table 3. pH of wild thyme extracts ranged from 5.12 ± 0.18 to 6.31 ± 0.15 meaning that all extracts were slightly acidic, and no significant changes in pH were noted between HVED extracts and extracts obtained by CE. Electrical conductivity was in range from 89.0 ± 6.3 µS/cm (extract TA11) to 646.0 ± 30.1 µS/cm (extract TA2). Statistical analysis of influence of main effect (treatment time, voltage, ethanol content and treatment type) to results of physical parameters is shown in Table 6. Statistically significant influence (p ≤ 0.05) to pH and conductivity had only ethanol content. With higher ethanol content, pH increased, while conductivity decreased. Regarding the temperature of extracts, it was shown that the maximum measured temperature after the HVED treatment was 37.8 ± 1.3 • C. For statistical analysis, a temperature difference calculated as a difference after and before HVED treatment, was taken in account for statistics. Maximum temperature difference was 12.2 • C for sample TA2 treated for 9 min with argon at 15 kV and water as a solvent, where also the highest conductivity was noted. Temperature difference was higher with longer treatment time, higher voltage, argon and lower ethanol content, while statistically significant influence was noted for all parameters except voltage. Regarding the power used for HVED treatment, statistically significant influence of treatment time and voltage was observed, where longer treatment time and higher voltage increased power. This trend was statistically confirmed for treatment time and voltage.
Additionally, Z p was used to select the optimal HVED treatment conditions where the highest cell membrane permeabilization degree was achieved. According to results (Table 3), the highest permeabilization happened in extract TA2 (9 min, 15 kV, 0% of ethanol), sample where highest conductivity and highest temperature difference was noted, indicating the highest yield of extraction, although it was not proven with results of phenolic compounds and antioxidants ( Figure 2). Z p lowered with higher ethanol content and increased with higher voltage applied. These results indicate potential use of Z p for assessment of electroporation during HVED treatment. However, statistical data showed that only ethanol content had a significant influence to the electrical conductivity, and Z p accordingly. Therefore, no clear conclusions could be provided regarding Z p index.

Effect of HVED-Assisted Extraction on Recovery of Bioactive Compounds from Wild Thyme
BACs are generally used in various industries such as food, cosmetic and pharmaceutical industries, especially due to the their antioxidative, antiseptic and antitumor capacity [32]. Extraction of BACs from wild thyme was performed by HVED as a green extraction technology in order to obtain high-quality extracts reach in BACs that could be used for functional food production. The obtained extracts were analyzed for total phenolic content and antioxidant activity (DPPH and FRAP) and compared with modified CE (Figure 2). HVED extraction showed higher results for all measured methods including TPC, DPPH and FRAP. The highest content of TPC was found in extract TN7 that was extracted using HVED, nitrogen at 20 kV, and with 50% ethanol (42.86 ± 2.38 mg GAE/g, Figure 2c). The highest antioxidant activity was found in extracts obtained by HVED using argon: TA7 and TA9 for DPPH and FRAP methods, respectively. Better extraction of polyphenols and antioxidants was enhanced generally with longer treatment time, higher ethanol content and higher voltage. TPC and DPPH were higher for treatment with nitrogen than argon, while FRAP values were higher for argon treatment. In order to define the best extraction parameters, it is necessary to establish the amount of BACs extracted according to the defined experimental design. Statistically significant influence of ethanol content was noted for TPC and FRAP, while longer treatment time significantly influenced to DPPH results (Table 6). Based on the model, the extraction yield was expressed as percentage of extracted polyphenols per g of sample (g GAE/g of sample). Therefore, yield of extraction was in line with results of TPC and the same trend was noted. The highest yield of extraction was seen for extract TN7 extracted with HVED using nitrogen for 9 min, at 20 kV with 50% of ethanol. In average, HVED extraction showed 2.03 higher yield of extraction compared to CE under same extraction conditions (ethanol content and time of extraction), and it was 2.07 times higher for nitrogen and 1.99 times higher for extraction with argon. A correlation analysis between TPC, DPPH and FRAP showed that the highest correlation was found between TPC and FRAP (0.259). Although these correlations were not high, all methods showed positive correlations. This is not an unusual occurrence since each of this method is based on different mechanism and is performed in different conditions. Jovanović et al. measured TPC in wild thyme extracts after extraction with 1:30 solid-to-solvent ratio and 50% ethanol. With these conditions, results were 26.6, 29.8 and 32.7 mg GAE/L for extraction by maceration, heat and ultrasound-assisted techniques, respectively [5]. These results are much lower compared to HVED, since extract TN7 that had the highest TPC (42.86 ± 2.38 mg GAE/g) had the equivalent of 214.31 mg GAE/L. Ðukić et al. evaluated various conventional (Soxhlet and macerate extraction) and non-conventional extraction (ultrasound, microwave and subcritical water extractions) methods from wild thyme. The results showed that the highest TPC and antioxidant activity was found using subcritical water extraction (141.12 ± 0.23 mg GAE/g and 170.32 ± 0.87 mg AA/G) [33]. Although these results seems higher compared to our study, no conclusions could be made regarding the best extraction method since different analytical methods were used and different solvents with different solvent to solid ratio were used.

Effect of HVED on Phenolic Composition of the Extracts
Results of the performed UPLC-MS/MS analysis from wild thyme extracts have shown ( Table 4) that the main compounds in wild thyme extracts were in the following order, according to their average content in all extracts: rosmarinic acid, oleanolic acid, luteolin, apigenin, diosmetin, hydroxytyrosol, quercetin, carnosol, camphor, thymol, p-cymene, and carvacrol. Rosmarinic acid was found to be the main phenolic compound found in wild thyme extracts, which was already confirmed by Jovanović et al. [5]. Rosmarinic acid was found in higher concentrations in HVED extracts compared to CE and it was higher with higher concentrations of ethanol, up to 50% in extracts. Chromatograms of representative extracts obtained by CE ( Figure S1) and HVED ( Figure S2) present the main components found by UPLC-MS/MS and their differences. The most pharmacological importance of rosmarinic acid is its antioxidant and anti-diabetic properties [34]. For that reason, wild thyme extracts are a valuable source for functional food production. Apigenin, carnosol, luteolin, oleanolic acid, quercetin, rosmarinic acid and camphor were found to be statistically significant higher with higher ethanol content. Also, luteolin and cymene significantly depended on treatment type. The sum of all BACs measured by UPLC-MS/MS was compared with results of TPC and correlation of 0.384 (Table S1) was found between UPLC-MS/MS results and TPC, which is the highest correlation compared to DPPH and FRAP.
Experimental results have confirmed theoretical results obtained by HSPs and COSMO-RS software. Apigenin, luteolin and rosmarinic acid were confirmed to be better soluble with higher ethanol content of the solvent, which was also indicated by the theoretical results (Tables 1 and 2). Although theoretical results showed better solubility of these compounds in ethanol, compared to water, still poor solubility was expected in both solvents calculated by HSPs (Table 1). Low probability of solubility (red color, Table 2) was also expected by COSMO-RS prediction in water. However, for apigenin, luteolin and rosmarinic acid, a very high probability (100%) was predicted for solution in ethanol. Therefore, it can be concluded that experimental results confirmed theoretical results regarding higher solubility of wild thyme compounds in ethanol compared to water. On the other hand, better results are given with COSMO-RS software than HSPs since high concentrations of apigenin, luteolin and especially rosmarinic acid were found after extraction with HVED.

Effect of HVED on the Content of Volatile Compounds in Wild Thyme Extracts
Results of HS-SPME/GC-MS analysis of volatile compounds from wild thyme extracts are given in Table 5. Carvacrol and thymol were already measured by UPLC-MS/MS, but very low concentrations were found by this methods (< 3 ng/mL), so these volatile compounds were measured also by HS-SPME/GC-MS. Results showed that monoterpenoid phenol carvacrol (0.47-21.86%) was the predominant volatile compounds found in wild thyme extracts, followed by linalool, thymol, and finally, 1,8-cineole that was found in small amounts (0.96-3.46%). A similar trend was found in a study where wild thyme essential oil was extracted by supercritical carbon dioxide as another green extraction method, and carvacrol and thymol were found as main compounds in the extract. That study demonstrated that oil rich fraction of the CO 2 extract yields were 0.3-0.5% [35] All measured compounds were found in higher concentrations in HVED extracts, compared to CE. 1,8-cineol and linalool were higher in HVED extracts obtained with nitrogen, while thymol and carvacrol were higher in extracts where argon was used. The differences between CE and HVED treated extracts in the same extraction conditions are presented in Figure S3a-c. The influence of ethanol content was difficult to assess, since overlapping of peaks profile happened with ethanol peak in chromatograms. For that reason, most extracts containing ethanol do not have available data for compound concentration. However, a traceability of results is visible in HVED extracts with nitrogen (RN2-RN6 and RN9-RN10) and argon (RA2-RA6 and RA9-RA10), except for linalool that was found in extracts MA5 and MA6, while it was not found in extracts MN5 and MN6. Similar results were presented in study by Sonmezdag et al., where thymol and carvacrol were found to be the main components of essential oil of wild thyme, assessed by GC-MS [36].
Results of thymol, carvacrol and linalool were compared with theoretical prediction models where low solubility was expected in water measured with both methods (HSPs and COSMO-RS) and in ethanol when calculated by HSPs. However, very good solubility of carvacrol and thymol (100%) was expected in ethanol, while medium solubility was expected for linalool (52.48%) evaluated by COSMO-RS software. Since it was difficult to evaluate experimentally extracts containing ethanol, no clear correlations can be drawn with theoretical results. On the other hand, COSMO-RS results showed lower probability of solubility of linalool, compared to thymol and carvacrol, which was proven with experimental results.

Determination of Pesticides and Heavy Metals in Wild Thyme Samples
High levels of pesticides and metals in food could cause serious toxicological effects. For that reason, it is important to monitor level of pesticides and metals even in raw material, in order to assure high-quality final product. The results of pesticides from dried wild thyme are given in Table 7. Residue levels of pesticides were determined according to EC Regulation No 396/2005. This Regulation include EU Pesticides database with all active substance of pesticide (EC 1107/2009) and their MRLs (EC 396/2005). Results showed that residue levels of all pesticides measure in wild thyme were lower than limit of quantitation of the method. For that reason, no exact data are provided. For most results, limit of quantification was below MRL and for other data it was not possible to quantify exact level and analyze if it was below MRL.
Residue level of heavy metals was determined according to EC Regulation No 1881/2006. Results showed that levels of heavy metals lead (Pb), cadmium (Cd) and mercury (Hg) in dried wild thyme sample were below MRLs for each substance (Table 8). Further analysis of safety and quality included analysis of other metals (chromium (Cr), nickel (Ni), manganese (Mn), iron (Fe), copper (Cu) and zinc (Zn)). These metals were measured in dried wild thyme material and in HVED extract (TN8). For these metals, no MRL data is provided because these are not included in EC Regulations. Although results are given for dried wild thyme per kilogram of plant and for extract per kilogram of final extract, it is clear that levels of Ni, Mn, Fe, Cu and Zn decreased, while level of Cr significantly increased. It could have happened because of abrasion of electrodes during HVED treatment and release of metals, such as Cr, in a solvent. However, it is possible to obtain safe and high-quality final extract of wild thyme since dried plant is considered as safe for human use. Detailed analyzes should be performed regarding levels of metals released in extracts during HVED treatment.

Hansen Solubility Parameters (HSPs)
HSPs is a method for characterization of solute-solvent interactions according to the classical "like dissolves like" rule. A detailed concept of HSPs is described in paper by Aissou et al. [30]. For solvent optimization, a simple composite affinity parameter, the RED number, has been calculated to determine the solubility between solvents and solutes: where R o is the radius of a Hansen solubility sphere and R a is the distance of a solvent from the center of the Hansen solubility sphere. The Hansen solubility sphere is determined by three Hansen parameters: dispersion (δ d ), polar (δ p ) and hydrogen bonding (δ h ) where R 0 determines the radius of the sphere in Hansen space and its center is the three Hansen parameters. A potentially good solvent has RED number smaller than 1 (the compound has similar properties and will dissolve), while medium and poor solvents have RED values of from one to three and more than 3, respectively. The chemical structures of the solvents and solutes discussed in this article could be mutually transformed by JChemPaint version 3.3 (GitHub Pages, San Francisco, CA, USA) to their simplified molecular input line entry syntax (SMILES) notations, which were subsequently used to calculate the solubility parameters of the solvents and compounds (HSPiP Version 4.0, Hansen Solubility, Hørsholm, Denmark).

Conductor-Like Screening Model for Real Solvents (COSMO-RS) Software
The COSMO-RS software is a statistical thermodynamic method for molecular description and solvent screening based on a quantum-chemical approach [37]. The prediction is based on a two-step procedure-microscopic and macroscopic. The procedure was explained in details by Aissou et al. [30]. The COSMOthermX program (version C30 release 13.01) was used to calculate the relative solubility between the solid compound and the liquid solvent in terms of the logarithm of the solubility in mole fractions (log 10 (x solub )). The logarithm of the best solubility was set to 0 and all other solvents were given relative to the best solvent. Also, the logarithm was transformed into probability of solubility (%). The calculation was performed at room temperature (20 • C). As an example, Figure 3 depicts the molecular structure of thymol and its sigma surface. [30]. The COSMOthermX program (version C30 release 13.01) was used to calculate the relative solubility between the solid compound and the liquid solvent in terms of the logarithm of the solubility in mole fractions (log10(xsolub)). The logarithm of the best solubility was set to 0 and all other solvents were given relative to the best solvent. Also, the logarithm was transformed into probability of solubility (%). The calculation was performed at room temperature (20 °C). As an example, Figure  3 depicts the molecular structure of thymol and its sigma surface.

Plant Materials
Dried wild thyme (Thymus serpyllum L.) material was used for extractions. It was provided by local specialized herb store (Suban d.o.o., Samobor, Croatia). Plant material was collected during the flowering season in 2017, in the northwestern part of Croatia, dried naturally, and stored in polyethylene bags in a dark and dry place, at ambient temperature until extractions. Measured plant particle size distribution was measured by the laser particle size analyzer Mastersizer 2000 (Malvern Instruments GmbH, Herrenberg, Germany) and results were as following: d(0.1) ≤ 158.4 µm; d(0.5) ≤ 289.0 µm; d(0.9) ≤ 457.4 µm. All extractions were performed using 1 ± 0.01 g of herb material added to 50 mL of extracting solvent (distilled water, 25% and 50% aqueous ethanol (v/v)) based on preliminar study results were different solvent to herb ratios were used and higher range of ethanol content.

High Voltage Electrical Discharge (HVED) and Conventional Extraction
For high voltage electric discharge generation, a IMP-SSPG-1200 generator (Impel group d.o.o., Zagreb, Croatia) was used previously described in more detail by Nutrizio et al. [38]. Previously optimized extraction parameters were set to frequency of 100 Hz, pulse width of 400 ns, voltage of 15 and 20 kV for argon gas and 20 and 25 kV for nitrogen gas, and treatment time of 3 and 9 min. The extraction was performed in a 100 mL beaker shaped reactor where the herb-mixture was transferred. The gap between electrodes was 15 mm. Argon and nitrogen gases were flowed in through the needle with the flow 0.75 L min −1 . Power used during the HVED treatment was measured directly from the HVED instrument. A modified conventional extraction (CE) method was performed for comparison purposes, with same extraction conditions as HVED: at (room) temperature by dissolving the dried thyme material (1 g) in the solvent (50 mL) with light magnetic stirring during 3 or 9 min. Both extractions, HVED and conventional, were performed in duplicates.

Physical Properties of Wild Thyme Extracts
The pH and electrical conductivity of extracts were measured immediately after HVED treatment using a pH and conductivity meter HI-2030-edge (Hanna Instruments, Bedfordshire, UK). Temperature was measured using an infrared thermometer PCE-777 (PCE Instruments Ltd., Southampton Hampshire, UK).
Electrical conductivity measurements of the extracts were used to quantify the degree of cell permeabilization induced by HVED treatment of given intensity. The results are presented as cell disintegration index (Z p ) and are calculated according to literature data for pulsed electric fields [39] with some modifications, expressed as follows: where σ t is the actual measured conductivity value of the extract, σ i is the conductivity of the extract obtained from untreated samples (intact cell tissue), while σ d is the highest value of conductivity, related to the maximally damaged cell tissue. For each treatment condition investigated the Z p value ranged between 0 (for intact tissue) and 1 (for fully permeabilized tissue).

Determination of Total Phenolic Content (TPC)
TPC of thyme extracts was determined using Folin-Ciocalteu method as previously described [38]. The calibration curve was prepared using 50 to 500 mg/L of gallic acid in ethanol. The concentration of TPC was expressed in mg of gallic acid equivalents per gram of sample (mg GAE/g of sample). All measurements were performed in duplicates.

Determination of Antioxidant Capacity
2,2-Diphenyl-2-Picrylhydrazyl (DPPH) Free Radical Assay DPPH assay of thyme extracts was determined according to previously reported procedure [38]. The results were calculated using calibration curve for Trolox and expressed as µmol of Trolox equivalents per gram of samples (µmol TE/g of sample).

Ferric Reducing Antioxidant Power (FRAP) Assay
The FRAP assay was conducted as previously reported [40]. FRAP values were calculated according to the calibration curve for FeSO 4 ·7H 2 O and expressed as µmol of Fe 2+ equivalents (FE) per gram of sample (µmol FE/g of sample).

Headspace Solid-Phase Microextraction (HS-SPME) Followed by Gas Chromatography and Mass Spectrometry Analysis (GC-MS)
HS-SPME was performed with manual SPME holder using three fiber covered with DVB/CAR/PDMS (Supelco Co., Bellefonte, PA, USA). For HS-SPME, the finely samples 2 mL were placed separately in 10 mL glass vials and hermetically sealed. The vials were maintained at 60 • C during equilibration (15 min) and extraction (45 min). Thereafter, the SPME fiber was withdrawn and inserted into GC-MS injector (250 • C) for 6 min for thermal desorption. The procedure was similar as reported [42]. GC-MS analyses were done on an Agilent 7890A Gas Chromatograph (Agilent Technologies, Palo Alto, CA, USA) equipped with a mass spectrometer (MSD) model 5977E (Agilent Technologies) and HP-5MS capillary column (5% phenyl-methylpolysiloxane, Agilent J & W). The GC conditions were same as reported previously by Jerković et al. (2016). The oven temperature was set at 70 • C for 2 min, then increased from 70 to 200 • C (3 • C/min) and held at 200 • C for 18 min; the carrier gas was helium (1.0 mL/min). The compounds identification was based on the comparison of their retention indices (RI), determined relatively to the retention times of n-alkanes (C9-C25), with those reported in the literature [43] and those from Wiley 9 (Wiley, New York, NY, USA) and NIST 14 (National Institute of Standards and Technology; Gaithersburg, MD, USA) mass spectral database. The percentage composition of the samples was computed from the GC peak areas using the normalization method (without correction factors).

Experimental Design and Statistical Analysis
The experiment was designed in STATGRAPHICS Centurion (StatPoint Technologies, Inc, Warrenton, VA, USA) software and it is presented in Table 9. All experimental results are shown for extracts obtained by HVED and modified conventional extraction (CE) for comparison. Each sample denotes different process parameters, where T stands for thyme, N for nitrogen gas used during HVED treatment and A for argon. Multi-factor categorical design consisted of 12 experimental trials per gas (argon and nitrogen). The three chosen independent variables for HVED assisted extraction were: treatment time (3 and 9 min), voltage applied during HVED (15 kV or 20 kV for argon, and 20 kV or 25 kV for nitrogen) and concentration of ethanol (0%, 25% or 50%). For HVED extraction, numbers 1-12 are the order of conducted treatment, and in CE extraction, 3 and 9 are referred to treatment time while 0, 25, and 50 stands for concentration of an ethanol solvent (%). A total of 30 extracts were prepared in duplicates and all results are given as average ± standard deviation (SD). Table 9. Denotation of samples, experimental design and process parameters.

Sample
High Voltage Treatment Time Statistical analysis was performed in XLStat (MS Excel 2010). The PCA analysis of theoretical prediction results (using HSPs and COSMO-RS) was performed in XLStat (MS Excel 2010). The PCA was used as a multivariate statistical analysis tool in the processing of the theoretical results to detect qualitative similarities or differences between two different prediction models. Coding of results was performed before modelling with the purpose of uniformed results: 0 (red color in Tables 1 and 2), 1 (yellow color) and 2 (green color). A PCA model after Varimax rotation was chosen.
ANCOVA was used to analyze the impact of independent variables: treatment time, voltage, ethanol content and treatment type (CE, HVED-nitrogen or HVED-argon) to measured physical and chemical parameters. For UPLC-MS/MS and GC-MS results, all individual (data not shown) and sum of concentrations of all compounds was assessed. The p-values present the statistical significance of each of the factor, and it was significant at p ≤ 0.05.

Conclusions
In this work the potential of green extraction of BACs from wild thyme by theoretical and experimental approach was demonstrated. The results of experimental study confirmed the potential of HVED for extractions of wild thyme bioactive compounds using water and ethanol. Extraction yield of polyphenols and antioxidants was enhanced with longer treatment time, higher ethanol content and higher voltage, and it was higher for treatment with nitrogen than argon. In average, HVED extraction showed 2.03 higher yield of extraction compared to CE under same extraction conditions (ethanol content and time of extraction), and it was 2.07 times higher for nitrogen and 1.99 times higher for extraction with argon. The main phenolic compound found in wild thyme extracts was rosmarinic acid, while the predominant volatile compound was monoterpenoid phenol carvacrol. Furthermore, experimental results confirmed theoretical results since higher solubility of wild thyme bioactive compounds was found in ethanol than water. However, more accurate results were obtained using COSMO-RS software, compared to HSPs. Wild thyme extracts obtained by HVED are considered as safe, in terms of pesticides and metals levels, and present a high-quality source of valuable BACs for further use in functional food production.
HVED was presented as fast and effective nonthermal technology where thermolabile BACs are being preserved and recovered in high concentrations. Furthermore, HVED has a potential to be scaled-up to industrial level to replace less environmentally acceptable conventional extraction methods. Finally, the results from this study encourage further investigation of HVED as a green extraction method and its comparison with other green extraction methods.