For experimental assays, 20 kg of fresh alperujo were obtained from a continuous two-phase centrifugation system (Oliomio 200 Eco, Toscana Enologica Mori, Italy) placed in the olive oil extraction plant of INTA EEA San Juan, Argentina. Milled olive fruits were from the Arbequina cultivar at a maturity index of 4. The raw material was homogenized and stored in individual pots of 1 kg at -20 °C. Before starting the assays, 3 individual samples of alperujo were subjected to the following chemical determinations: pH, water content, organic matter content, total soluble phenols, DHPG, HT and Ty.

Raw alperujo was characterized according to the following determinations: Water content was measured gravimetrically at 70 ºC (Martinez et al., 2021); pH was determined by potentiometric determination in 1:5 (w/v) water extract (Martinez et al., 2021); total organic matter and ashes were determined by loss on ignition at 550 ºC for 24 h (Martinez et al., 2021). Total organic carbon was calculated from organic matter according to (Navarro et al., 1993). The phenolic content in alperujo was determined in methanolic extract (methanol:water 80:20 (v/v) using 2 mL per gram of fresh alperujo) by a modified version of the method described by (Singleton and Rossi, 1965) with Folin-Ciocalteu reagent and measured in a Shimadzu 1240 UV-Visible spectrophotometer at 725 nm. The results were expressed as mg of caffeic acid per kg of initial sample. The individual phenols of interest (DHPG, HT and Ty) were measured in methanolic extract (explained above) by high-performance liquid chromatography (HPLC Hewlett-Packard series 1100), using a Kinetex® EVO 5 μm C18 (250 × 4.6 mm) Phenomenex® column. The analysis was performed at room temperature, the elution was at a flow rate of 1.0 mL/min, with a mobile phase A of acetonitrile and B ultrapure water, using the following gradient over a total run time of 55 min: 95% A initially, 75% A at 30 min, 50% A at 45 min, 0% A at 47 min, 75% A at 50 min, 95% A at 52 min until the run was completed. The chromatograms were analyzed at 280 nm by integration peaks at different wavelengths, according to calibrations performed with external standards. The results were expressed in mg/Kg or mg/L.

For determinations in liquid fractions, total soluble solids were determined by a portable refractometer (Atago, JPN), and expressed as Brix degrees. Total soluble phenols were determined in 1/10 water dilutions by the colorimetric method described above. The results were expressed as mg of caffeic acid/L of liquid fraction. The individual phenols of interest (DHPG, HT and Ty) were measured following the protocol described above.

In order to evaluate the relevant conditions required to increase the total phenolic concentration in the alperujo liquid fraction a factorial assay was performed. Closed plastic tubes containing 40 g of alperujo were positioned in thermostatic baths at two temperature levels (30 and 70 ºC) and two exposure times (60 and 180 min). In addition, a control treatment of untreated alperujo was included. Three replicates of each treatment were performed in parallel. Once treatments were applied, samples were centrifuged at 3500 g for 20 min to separate the solid and liquid phases. After centrifugation, weight, total soluble solids and total soluble phenols were determined in liquid phases. The percent of phenols recovered was calculated by the following equation:

Where: “T liquid phase PC” represents the phenolic content in the liquid phases obtained after treatment application. “T liquid phase weight” represents the weight of the liquid phase obtained after treatment application. “Control liquid phase PC” represents the phenolic content in liquid phases obtained without treatment application. “Control liquid phase weight” represents the weight of the liquid phase obtained without treatment application.

For the development of optimization studies, Box-Behnken experimental design followed by response surface methodology were applied. Box-Behnken is a second-order design based on three-level incomplete factorial designs. It is used to measure the combined effect of the factors under study, reducing the number of experiments, improving the performance of time and resources (Box and Behnken, 1960). The response surface methodology (RSM) is a technique to model and analyze statistical problems. The main objective is to optimize the response surface which is influenced by various parameters, quantifying the relationship among them and plotting the response (Montgomery and Wiley, 1984). A total of fifteen experimental trials combining three factors at three levels were analyzed: temperature (50 , 70 and 90 ºC), exposure time (30 , 75 and 120 min) and alperujo water content (70 , 80 and 90%). Trials were performed in closed plastic tubes containing 100 g of alperujo in thermostatic baths. As performed in a previous stage, once trials were conducted, phase separation was accomplished by centrifugation at 3500 g for 20 min. Liquid phases were weighed, and total phenols, DHPG, HT and Ty concentrations were determined. Then the content of phenols recovered was calculated considering both concentration in liquid phase and amount of liquid phase recovered by the following equation:

Relevant variable selection was performed under factorial design. Analysis of variance (ANOVA) was achieved using Infostat Professional Software 2.0 version (UNC). For comparison of means LSD Fisher (p ≤ 0.05) was employed.

Variable optimization was performed in the Box-Behnken design according to Response Surface methodology, using Design Expert software 7.0.0 (Stat-Ease Inc., Minneapolis, USA). Statistical differences were established at p ≤ 5%. Model significance, polynomic equation, data normality and surface response graphs were determined for each variable. Model efficiency was determined by R2 and adjusted R2. The response surface statistical model responded to the equation adjusted by three factors (temperature, water content and time) and four response variables (Total phenols, DHPG, HT and Ty). The optimal levels of each response variable were obtained and the polynomic equation was determined.

As shown in Table 1, the chemical composition of the alperujo employed in the assays was similar to others employed in different reports (Alburquerque et al., 2004; Morillo et al., 2009). It was characterized by high water content and slightly acidic pH. Regardless of the water content, it was mainly composed of organic matter and a minor proportion of mineral ashes. Among organic constituents, the total phenolic content (7233.8 mg/Kg) was in the lower limit compared to other reports, presenting values ranging from 6000 to 26000 mg/Kg of raw dry alperujo (Alburquerque et al., 2004; Rubio-Senent et al., 2017). As stated, the amount and profile of phenolic compounds present in olives is highly variable and dependent mainly on cultivar and maturity stage as well as agronomic and climatic conditions (Obied et al., 2008). In the present work, the alperujo employed belonged to the Arbequina cultivar, grown under intensive drip irrigated system, harvested at a maturity index of 4 and immediately milled. Arbequina represents one of the most abundant olive cultivars in Argentina (Gómez del Campo et al., 2010) and generally the olive oils obtained with this cultivar are characterized by low phenolic content (Monasterio et al., 2017). Regarding the individual phenols of interest, Hydroxytyrosol was the main hydrophilic phenolic compound observed representing almost 20% of the total phenol content, while tyrosol and DHPG together did not reach 4%. These results agree with previous data (Rubio-Senent et al., 2013) and support the significance of applying pre-treatments in order to induce phenolic compound hydrolysis and increase the total amount of hydrophilic forms.

Table 2 shows the effect of thermal treatments performed on alperujo samples (40 g) on the weight of liquid fraction recovered as well as the total soluble solids and total phenols. As shown, the weight of the liquid fraction reached its highest value when thermal treatments were performed at 70 ºC, without presenting statistical differences between 60 and 180 minutes. On the other hand, the weight of the liquid fraction obtained by treatments achieved at 30 ºC did not present differences compared to untreated control samples. In addition, the level of total soluble solids and total phenols presented a similar pattern, showing an increase when thermal treatments were performed at 70 ºC and remaining similar to untreated alperujo in treatments carried out at 30 ºC (Table 2). These results agree with previous data which suggest that thermal treatments provoke the relaxation of organic matter constituents allowing the exit of intracellular water and favoring the hydrolysis of complex molecules and the solubilization of lower compounds such as phenols, sugars, organic acids and proteins, among others (Rodríguez et al., 2007).

Taken together, both the increase in liquid fraction weight and the total phenol content in these fractions, the amount of recoverable phenols increased to 136-138% with respect to the control when thermal treatments at 70 ºC where applied (Figure 1). The observed increase in the total amount of recoverable phenols is lower than that observed in other reports where thermal treatments were performed at higher temperatures (Lama-Muñoz et al., 2011; Rubio-Senent et al., 2013). As these authors explain, the importance of the application of a heat treatment is to achieve a solubilization of high value-added components, such as phenols, and to enable phase separation. Obviously, the higher the treatment temperature, the higher the solubilization of phenols and the better the phase separation, but also the higher the formation of degradation metabolites, such as furfurans from sugars, and the higher the investment and operating costs. For this reason, the application of thermo-malaxation is sought to improve the use of alperujo, a first step being the use of temperatures of up to 70 ºC which solubilize phenols and allow an adequate separation of phases, even if a quantity of solids in suspension is reached, which implies an intermediate sedimentation step prior to the extraction of phenols from the liquid phase (Fernández-Prior et al., 2020).

Table 3 summarizes the experimental design conditions applied as well as the values of response variables (Total phenols, DHPG, HT and Ty) expressed as mg in liquid fractions obtained from 100 g of initial sample. Response surface methodology was used to optimize each response variable considering the linear, quadratic, and cross-product interactions of the independent variables at the 95% confidence level. The analysis of the variance (ANOVA) for each response variable is shown in Table 4. As shown, Total phenols, DHPG and Hydroxytyrosol described quadratic response surface models with high R2 and R2 adjusted values. Temperature, water content and the quadratic effect of water content were highly desirable (p-value ≤ 0.05) for Total phenols and HT, while Temperature and its quadratic effect were significant for DHPG. On the other hand, the interaction between independent variables and Tyrosol was explained by a linear model, with temperature and water content as the significant variables.

The polynomial that describes the relationships between the process variables and each response variable are represented by the following equations:

Where: “T” represents temperature; “t” represents time; “WC” represents water content.

As can be seen in Figure 2, all the response variables increased with temperature. Regarding time and water content, these factors presented particular patterns for each response variable. Total phenols and HT showed maximum levels when time and water content were set at central points (Figure 2 A, B, E, F). Tyrosol was greater when the water content was set at the lowest level without being highly affected by time (Figure 2 G, H), while DHPG was not highly affected by time or water content (Figure 2 C, D). Based on the different models and equations, the optimal conditions and the higher values for the different response variables are summarized in Table 5.

Table 6 shows the theoretical levels of the different compounds evaluated (total phenols, DHPG, HT and Ty) expressed as concentrations in liquid fractions obtained and in raw dry alperujo. As can be observed, the concentration of total phenols solubilized reached 3804 mg/kg liquid fraction, and 17434 mg/kg raw dry alperujo. The values obtained are interesting and similar to those obtained experimentally by other authors (Lama-Muñoz et al., 2011). The comparison of theoretical values obtained for each response variable (Table 6) with the initial alperujo content (Table 1) indicates increases of 2.4, 957.8, 3.4 and 6.4 folds in initial total phenols, DHPG, HT and Ty, respectively, from raw dry alperujo. These results confirm that applying low thermal treatments to alperujo is possible to obtain phenol-enriched fractions. The observed increase in the content of hydrosoluble phenols supports the hypothesis that thermal treatments provoke the hydrolysis of complex hydrophobic phenolic compounds and favor the solubilization of hydrophilic molecules. The results show that the extraction of phenols is temperature and time dependent, and that for the tested conditions the solubilized amount of the main phenols in free form can justify their extraction as mentioned by other authors (Fernández-Prior et al., 2020). Likewise, at these temperatures it is necessary to continue studying the solubilization of polymerized or more complex phenols that could increase with subsequent chemical or enzymatic hydrolysis processes by the addition of acids or enzymes or by the action of those already present during a certain storage time, which will also help to separate suspended solids that hinder the application of chromatographic processes to obtain final phenolic extracts.