3.1. Experimental Model Fitting
The present study was conducted to evaluate the optimal conditions for the extraction of phenolic antioxidants from Merlot’s grape stems (
Vitis vinifera L.). This variety is one of the most representative grape varieties as it is one of the most widespread cultivars. For this purpose, a central composite rotatable design was applied as displayed in
Table 2. Three independent variables or factors, namely, ethanol concentration (X
1, 0–100%), temperature (X
2, 40 °C–120 °C) and time (X
3, 1–11 min), were studied to assess their influence on antioxidant activity, flavan-3-ol monomers and oligomers, polymer procyanidins and total phenolic compounds. These factors are considered as the main extraction-independent variables, at the expense of others less significant, such as solid–solvent ratio or solvent pH [
13,
32]. Aqueous methanol has been proposed as the most suitable solvent for the extraction of phenolic compounds from grape products [
11,
22]. Nevertheless, this toxic solvent is currently being replace by aqueous ethanol [
6,
12]. Its high efficiency and GRAS status makes it suitable for food or pharmaceutical applications [
12]. Heating up to 200 °C has been commonly used in PLE extractions, since higher temperatures lead to a greater extraction yield [
23]. Nevertheless, values above 120 °C should be avoided as they may cause degradation of phenolic compounds [
16]. Accordingly, the maximal temperature value was set at 120 °C in the present study. Moreover, the extraction time variable was limited in low values, since higher extraction time may promote phenolic degradation without an enhancement of extraction yields [
17,
32].
Table 2 shows response variable data corresponding to each experimental condition.
The regression coefficients of linear, quadratic and interaction terms of the experimental factors were calculated by fixing the experimental values of response variables to a quadratic linear regression model. The effect of each term in the model and its statistical significance on the response variables were analysed from the standardized Pareto chart (data not shown). The quadratic and interaction terms not significantly different from zero (
p ≤ 0.05) were excluded from the model, and the mathematical model was refitted by multiple linear regression (MLR), resulting in the polynomial equations shown in
Table 3.
These equations suggest that RSM was successfully applied for the optimisation of the considered variables. The models did not show significant lack of fit (
p > 0.05), indicating well-fitting models for yield, TPC, ABTS and DPPH, opposite to flavan-3-ol monomers and oligomers, and polymer procyanidin behaviour. However, determination coefficients (R
2) for all the studied variables were over 0.90. Therefore, the proposed models could be used as an approach to the real behaviour of these compounds regarding these extraction parameters. Concerning extraction yields, the obtained data exhibited a very good fit for a quadratic model. Moreover, all the experimental factors were significant, mainly ethanol proportion in a quadratic manner (
Table 3). In addition, some significant interaction factors were found. Response surface plot showed that an increase of ethanol proportion caused an extraction yield rise of up to almost 25% (
Figure 1). Therefore, higher and lower ethanol:water ratios led to a decrease in the extraction yield, more markedly at high ethanol proportions (
Figure 1A,B). Moreover, temperature caused a linear effect, that is, extraction yield was enhanced linearly when temperature increased (
Figure 1A). On the other hand, a weak quadratic influence of extraction time was noticed.
Similar results were found regarding the total phenolic content (TPC) and antioxidant activity (ABTS, DPPH), where good fitting models for these response variables were established (
Table 3). Furthermore, extraction solvent was the most important factor, showing a quadratic effect, together with a linear effect of temperature (
Figure 2A,
Figure 3A and
Figure 4A). Although time and some interaction factors resulted in meaningful effects, a lower contribution to response variables was determined for these parameters (
Figure 2B,
Figure 3B and
Figure 4B), where extraction time significance is generally linked to temperature [
32]. The optimum ethanol concentration effect was determined to be close to 30%; meanwhile, 120 °C allowed reaching the highest response variable values. Similar results were found in other studies, where ethanol:water mixtures’ behaviour was analysed [
12,
13]. These authors indicated that extraction solvent was the main factor for both antioxidant activity and TPC of the extracts, showing a quadratic main effect. The optimum ethanol concentration in ethanol:water mixtures to achieve the maximum phenolic extraction, and therefore the greatest antioxidant capacity, is generally observed between 30% and 80% [
4,
12,
13]. In accordance with the present study, the optimization of conventional solid–liquid extraction for two grape stem samples through RSM determined that ethanol and temperature were the main factors during phenolic extractions. The optimum ethanol:water concentration was determined as 57.9% and 63.8%. Although a strong negative and quadratic effect was also observed in these studies for ethanol concentration effect, noticeably higher amounts of ethanol were found [
12]. The lower optimum ethanol content in the PLE extraction solvent of the present study (30%) is probably due to the improvement extraction capacities of the solvents because of the pressure and higher temperatures applied during the extraction process. Moreover, PLE modified the extraction capacities of solvents by reducing their polarity [
17] and enhancing the extraction of low polar compounds, such as phenolic compounds. Therefore, PLE reduces the required ethanol:water proportion compared to conventional solid–liquid extraction [
12,
13] or ultrasound extraction [
4]. Besides, temperature increase is generally associated with enhancements of phenolic compound extraction [
12,
13]. Since PLE allows increasing the temperature over the solvent boiling point, the extraction capacity of the used solvents is generally enhanced [
17]. Because of that, optimum extraction time is reduced when PLE is used in comparison with conventional extraction [
12,
13].
Furthermore, temperature was found to be an important but controversial factor, regarding phenolic compound extraction and antioxidant activity. In general, a high extraction temperature is correlated with an increase in the solubility of phenolic compounds from the matrix [
33], as it reduces solvent viscosity and enhances solvent penetration [
34]. On the other hand, high temperatures may lead to breakdown of thermolabile compounds [
13]. Nevertheless, this shortcoming could be avoided by using high extraction temperatures (100–120 °C) at a short extraction time [
35]. Accordingly, in the present study, a short extraction time (up to 10 min) at 120 °C avoided thermal degradation of some phenolic compounds.
Regarding total flavan-3-ol monomers and oligomers or total polymers of procyanidin content, experimental data did not fit with the quadratic proposed model, suggesting more complex behaviour. Nevertheless, according to response surface plots, extraction solvent resulted as the main factor in the total mono-oligomer demeanour, whereas temperature and time were remarkably less meaningful. Ethanol seems to enhance flavanol-3-ol monomer and oligomer extraction (
Figure 5A), while temperature or time influences are less clear (
Figure 5A,B). Opposite to that, higher ethanol proportions seem to reduce the extraction of total polymer procyanidins, while temperature increased their extraction yield (
Figure 6A). In this case, although extraction time turned out to show a positive tendency, an important interaction between time and the rest of studied factors distorted the effect of time into an overall negative trend (
Figure 6A,B). Moreover, temperature increase resulted in a higher extraction of polymers, while ethanol had a suppressive effect, but enhanced the flavan-3-ol monomer and oligomer yield. Sun and Spranger [
36] indicated that temperature allowed higher proanthocyanidin extraction rates if the temperature did not reach a degradation point. Likewise, solvent mixtures with higher polarities improve the extraction of these compounds, disrupting the bonds between phenolic compounds and the matrix. In this sense, Karvela et al. [
13] found that greater flavanol monomers contents were reached at 60% of ethanol:water, whereas oligomers and polymer proanthocyanidins were achieved at 44% and 55%, respectively.
Nevertheless, it is important to highlight that, in all cases, individual optimisation of extraction conditions is required as grape variety, agro- and weather conditions or plant material are other factors that are not taken into consideration and can have significant influence on phenol recovery [
12].
3.2. Optimal Conditions and Validation of the Developed Model
Optimal experimental conditions were achieved for extraction yield, TPC and antioxidant activity since these response variables were fitted to the proposed model. As can be seen in
Table 4, only slight differences were found regarding the optimal conditions of these variables. A tight range of ethanol concentrations (22%–30%) was found as the optimum, while the highest temperature (120 °C) was optimal in every response variable. Moreover, high extraction time showed better results, except for extraction yield and TPC, where a slight decrease was shown at 11 min. Therefore, 30% ethanol, 120 °C and 10 min were selected as the most suitable extraction conditions in order to obtain an extract with the highest contents of phenolic antioxidants (optimum extract) by PLE from Merlot grape stem.
Under such optimal conditions, the statistical model predicted an extraction yield of 29.2%, 192.4 mg GAE/g extract (TPC), and TEAC values of 3.81 and 1.31 mmol Trolox/g extrac regarding the ABTS and DPPH methods, respectively. To corroborate these values, additional extractions were made at the optimal extraction conditions (
Table 4). The results display that optimum extract showed values very close to the predicted ones, validating the proposed model. In addition, ORAC assay was carried out in order to perform a deeper antioxidant characterization of this optimum extract, with an ORAC value of 1.48 ± 0.17 mmol Trolox/g extract. Moreover, 26.8 ± 0.4 mg of mono-oligomers and 79.17 ± 1.36 mg of proanthocyanidin polymers were quantified in this extract.
These optimal experimental conditions allowed obtaining an extraction yield, TPC and antioxidant activity in agreement with the wide range of values determined by González-Centeno et al. [
22] for several grape stem varieties of
Vitis vinifera, or even slightly higher results regarding PLE Merlot extract. However, it should be noted that these authors proposed substantially different extraction conditions.
Experimental designs have been proposed previously to find optimal extraction conditions for grape stems [
12,
13]. In these studies, ethanol:water mixtures, time, temperature or pH were studied as experimental factor using SLE. Slightly higher TPC, along with greater TEAC values, was found in the present study. It is worth mentioning that a lower ethanol:water proportion was required at optimal extraction conditions compared to SLE [
12,
13]. As has been mentioned before, even if the grape variety or environmental conditions should affect this behaviour, this result could be ascribed to an electric constant decrease in the extraction solvent at high pressure [
37]. Furthermore, shorter extraction time and less consumption of solvents were used for PLE than for SLE.
Moreover, similar or lower values of TPC and antioxidant activity were observed when data presented in this study were compared with other SLE or PLE extracts from grape stems [
8,
18,
22]. However, regarding these parameters, the few data in the literature concerning grape stem extracts show a wide range of values. In this regard, differences caused by ripening stage, grape variety, geographic factors, climatological factors or oenological practices, as well as different extraction procedures applied, should be considered [
38,
39].
3.3. Correlation between Response Variables
Correlations between TPC and TEAC values (ABTS and DPPH) were established in order to confirm the influence of phenolic compounds on the antioxidant activity of the extracts. Antioxidant activity, measured by two methods in the present study (ABTS and DPPH) resulted in a strong correlation with TPC, achieving r = 0.993 and r = 0.987, respectively (
p ≤ 0.001). Besides, antioxidant activities obtained by both methods showed a high correlation between them (r = 0.993;
p ≤ 0.001)). This confirms that phenolic compounds are the main factor responsible for the extract antioxidant activity, according to the similar optimal conditions predicted by RSM models. In addition, Pearson’s test indicated a good correlation between TPC and the two fractions of phenolic compounds, flavan-3-ol monomers and oligomers (r = 0.557;
p ≤ 0.05) and polymers (r = 0.586;
p ≤ 0.01). Furthermore, statistical correlations between antioxidant activity and flavan-3-ol monomers and oligomers were found (r = 0.571;
p ≤ 0.05 for ABTS method, and r = 0.617;
p ≤ 0.01 for DPPH method). However, correlations turned out to be stronger with polymer procyanidins, being r = 0.621 (
p ≤ 0.01) for ABTS and r = 0.637 (
p ≤ 0.05) for the DPPH method. These results reveal that the antioxidant activity of the grape stem extracts could be ascribed to their general phenolic content [
38], although particular contributions of phenolic groups, such as polymer proanthocyanidins, could be higher [
13].
Additionally, a PCA was carried out to understand correlations between procyanidins and antioxidant activity. The principal component of the analysis explains a 90.79% of the samples. The antioxidant activity, determined using the ABTS method, was mainly explained by PC1 (76.48%) whereas PC2 contributed to a lesser extent (14.32%) (
Figure 7A). Therefore, PCA analyses showed a strong correlation between the polymer content and the total antioxidant activity of the samples. Therefore, those samples characterised by higher amounts of polymers, being samples 7, 12 and the optimum extract, showed greater antioxidant activity. These samples were characterised by the use of high temperatures during PLE extraction (104 °C, 120 °C and 120 °C, respectively), being in concordance with the polynomial equations of the fitted models (
Table 3). Additionally, the proximity of samples 3 and 14 in the PCA graph, as well as sample 7, evidences the influence of the extraction time. Similar results were observed when PCA analysis was conducted for DPPH values (
Figure 7B), explaining 89.65% of the samples.
3.4. Phenolic Composition of the Optimum Extract
Forty-two phenolic compounds were identified by HPLC-PAD-MS. Optimal PLE grape stem extract showed a complex composition of phenolic compounds, including phenolic acids, stilbenes, flavonols and, especially, flavanols (
Table 5). Regarding phenolic acids, gallic and caftaric acids were the main hydroxybenzoic and hydroxycinnamic acids of the extract, followed by vanillic and syringic acid [
19]. In addition, stilbenes were identified, including
trans-resveratrol,
ε-viniferin,
trans-resveratrol-glucoside (piceid), along with different dimmers and trimers of
trans- and
cis-resveratrol [
9,
10,
40].
Numerous flavan-3-ols were also determined, including monomers, dimers and oligomers [
40]. Catechin was the main monomeric compound, followed by epicatechin, whereas dimer B
1 turned out to be the highest dimer compound [
11].
Moreover, different flavonols were quantified, mainly as quercetin derivatives. The most remarkable compound corresponded to quercetin-3-
O-glucuronide, followed by quercetin-3-
O-glucoside [
40]. These forms of quercetin are the main flavonols of grape stems, along with others such as quercetin-3-
O-rutinoside and quercetin-3-
O-galactoside [
8,
19,
40]. In addition, low quantities of different anthocyanins where detected in the extract, malvidin-3-
O-glucoside being the most abundant of this group [
8].
It is remarkable that, as far as we know, it is the first time that compounds such as ethyl gallate, ellagic acid, delphinidin-7-
O-glucoside or cyanidin-3-
O-glucoside have been identified in grape stem extracts, although ethyl gallate has been previously reported in grape seed extracts [
25].
Just a few articles depict the composition of proanthocyanidin fraction in grape stem extracts. In the present study, 80 mg of CE/g extract was quantified as procyanidins according to the procedure of Sun et al. [
27]. Procyanidin characterization revealed a mDP of 12 units. Structural composition showed catechin as a predominant terminal unit and epicatechin as a principal extension unit, where approximately 12% of the total constituent units of procyanidins were galloylated (
Table 6). Focusing on general composition, epicatechin was the main monomer unit (72.3%), followed by catechin (14.6%), epicatechin gallate (12.4%) and, to a lesser extent, epigallocatechin (0.7%). All these results are consistent with the stem proanthocyanidin fraction described in literature [
14,
22], and particularly with Merlot’s proanthocyanidins [
7]. Nevertheless, slightly higher mDP was found in the present study. This result might be attributable to the effect of pressure extraction conditions used in this study, allowing a higher penetration of the solvent during extraction process.