Recovery of Anthocyanins from Passion Fruit Epicarp for Food Colorants: Extraction Process Optimization and Evaluation of Bioactive Properties

The potential of passion fruit (Passiflora edulis Sims) epicarp to produce anthocyanin-based colorants with bioactive properties was evaluated. First, a five-level three-factor factorial design coupled with response surface methodology was implemented to optimize the extraction of anthocyanins from dark purple epicarps. The extraction yield and cyanidin-3-O-glucoside content were used as response criteria. The constructed models were fitted to the experimental data and used to calculate the optimal processing conditions (t = 38 min, T = 20 °C, S = 0% ethanol/water (v/v) acidified with citric acid to pH 3, and RS/L = 50 g/L) that lead to maximum responses (3.4 mg/g dried epicarp and 9 mg/g extract). Then, the antioxidant, antimicrobial, and cytotoxic activities of anthocyanin extracts obtained using the optimized method and a conventional extraction method were evaluated in vitro. The extract obtained by the optimized method revealed a higher bioactivity, in agreement with the higher cyanidin-3-O-glucoside content. This study highlighted the coloring and bioactive potential of a bio-based ingredient recycled from a bio-waste, which promotes a sustainable bioeconomy in the agri-food sector.


Introduction
The food industry is a highly regulated sector in Europe, notably in terms of safety and hygiene, with the priority of reducing the environmental impact, especially the production of bio-waste [1]. In this regard, the European Waste Framework Directive has required Member States to take measures to encourage the selective collection of bio-waste for composting, digestion, and treatment, and the use of environmentally safe materials produced from this bio-waste [2].
In the fruit processing industry, the amount of bio-waste obtained during transformation is generally between 30-50%. Seeds, pulp, bagasse, and peel are the most common bio-waste resulting

Colour Parameters of Fresh and Dried Epicarp
The color parameters of the plant matrices may reflect the type of molecules present in their composition. Thus, in this study, the color of the passion fruit peel was measured in order to preliminarily assess the typology of the coloring compounds, as well as the color tone and intensity. It was also important to understand the interference of the dehydration process in relation to the peel color, that is, if there was a significant degradation of the coloring pigments. The results of the color analysis of the purple passion fruit epicarp are presented in Table 1. The parameter L* denotes lightness, a* corresponds to chromaticity on a green (−) to red (+) axis, and b* to chromaticity on a blue (−) to yellow (+) axis. As observed, the freeze drying process affected the epicarp color. The L* parameter did not suffer a significant change, with values close to 34 in the CIE L*a*b* color space, while the other two parameters were significantly (p < 0.05) affected by drying. A higher a* value was recorded on the dried epicarp, thus translating a more intense red color. In turn, the same sample showed a lower b* value. These changes in color coordinates indicate that the passion fruit epicarp acquired a shade closer to purple after lyophilization. To better visualize these color differences, data were converted into RGB values and the obtained colors are shown in Table 1.

Anthocyanin Profile of the Epicarp Extract
Data of the anthocyanin analysis in the passion fruit epicarp hydroethanolic extract obtained by the conventional solid-liquid extraction are shown in Supplementary Materials, Table S1. The compound identification was made based on the retention time, UV-Vis spectrum, and MS fragmentation patterns. The analysis revealed the presence of one anthocyanin ( Figure S1), identified as cyanidin-3-O-glucoside by chromatographic data comparison with those of the commercial standard. It has a [M+H] + ion at m/z 449 and a MS 2 fragment at m/z 287 (Table S1). In this study, the conventional extraction method, previously described by Jabeur et al. [12], where solid-liquid extraction is performed twice (1 h + 1 h) with 80% ethanol acidified with 0.05% citric acid as the solvent, yielded 34% extract that contained 8.3 ± 0.1 mg of cyanidin-3-O-glucoside/g extract. Therefore, it was possible to recover 2.82 mg of cyanidin-3-O-glucoside per gram of dried epicarp.

Optimization of the Extraction Process for Obtaining Anthocyanin-Rich Extracts
Although there are some studies on extraction of anthocyanins from passion fruit peel, there are few that detail the processing conditions that maximize their recovery. Moreover, the different qualitative and quantitative anthocyanin profiles of the different natural sources does not allow the direct extrapolation of the extraction conditions delineated for other matrices [16]. Therefore, the approach adopted to reduce costs (energy and solvent consumption) and optimize the extraction of cyanidin-3-O-glucoside from purple passion fruit epicarp was to use an RSM-coupled central composite design (CCD).
The CCD design presented in Table 2 was used to optimize the extraction conditions for anthocyanins considering the following independent variables and ranges: time, t (5-85 min); temperature, T (20-90 • C); and solvent proportion (% ethanol in water), S (0-100%). A comprehensive summary of the different steps carried out in the optimization study is illustrated in Supplementary Materials, Figure S2. The response criteria used to optimize the anthocyanin extraction are shown in Table 3, namely the anthocyanin (cyanidin-3-O-glucoside) content in both dried epicarp (Y 1 ) and extract (Y 2 ) and the extraction yield (Y 1 /Y 2 ). These response criteria are of interest to industrial sectors dealing with the recovering of high value-added molecules from plant materials or the production of bio-based ingredients, and provide information on the amount of plant material required to obtain a certain quantity of the desired compounds and the concentration of these natural compounds in the produced extracts.
According to the CCD design matrix in Table 3, it was possible to recover from 0.74 to 2.58 mg of cyanidin-3-O-glucoside from 1 g of dried epicarp (with runs 5 and 1, respectively), and to obtain from 2.84 to 9.34 mg of the same compound per gram of extract (with runs 5 and 2, respectively). The extraction yield was also affected and ranged from 12.1 to 44.6% (with runs 10 and 13, respectively). Table 3. Experimental results obtained under the conditions of the CCD design matrix for the responses Y 1 (mg anthocyanin/g dried epicarp), Y 2 (mg anthocyanin/g extract) and Y 3 (Y 1 /Y 2 , g extract/g dried epicarp) used in the anthocyanin extraction optimization process by RSM.

Run
Coded Values Natural Values Experimental Responses

Model and Response Surface Analysis
The response values shown in Table 3 were fitted to a second-order polynomial equation to obtain the parametric values shown in Table 4, where the results of the statistical verification of the obtained models are also presented. The fitting procedure was made using nonlinear least-squares estimations, in which the non-significant terms were excluded; the significant ones were assessed at a 95% confidence level (α = 0.05). The parametric values reveal the complexity associated with the extraction process and reflect the expected variation in the response per unit change in factor value when all remaining variables are kept constant. The intercept is the overall average response of the RSM design matrix runs (Table 3). Table 4. Parametric results obtained for the responses Y 1 (mg anthocyanin/g dried epicarp), Y 2 (mg anthocyanin/g extract), and Y 3 (Y 1 /Y 2 , g extract/g dried epicarp) used to optimize the extraction of anthocyanins from passion fruit epicarp by RSM (part A). Part B shows the statistical information of the fitting procedure. Part C presents the variable conditions in natural values that lead to optimal response values.

(C) Optimal variable conditions for response maximization
Individual conditions Global conditions The developed models had a non-significant lack-of-fit and were statistically validated and used to navigate the design space in the optimization steps.
Although parametric values show responses patterns, the best way to visually describe the effects of any independent variable in the recovery process of bioactive molecules is to generate 3D response surface plots, representing the combined effects of two variables and keeping the other variable constant. Figure 1 shows the 3D surface graphs for the three studied response criteria, and the corresponding 2D contour graphs focusing the optimal extraction points are represented in Figure 2. Both surface and contour graphs are in accordance with the parametric values derived from the multiple regression analysis presented in Table 4.
The 3D surface graphs translate the parametric results of Table 4 and visually represent the influence of the three independent variables involved in the process (Figure 1). Part A shows 3D response surfaces. In each graph, the combined effects of two variables were represented with the excluded variable positioned at its individual optimal value, which is shown in Table 4. In turn, part B illustrates the capability of the models to predict the obtained results and the residual distribution as a function of each variable. A good agreement between experimental and model-predicted values was observed, with R 2 values higher than 0.95 in all cases (Table 4), which translates the percentage of variability explained by the models. In addition, the errors had constant variance, with the residuals scattered randomly around zero.  Graphical results for the response criteria Y1 (mg anthocyanin/g dried epicarp), Y2 (mg anthocyanin/g extract), and Y1/Y2 (g extract/g dried epicarp). The 3D response surface graphs are shown in Part A. In each graph, the excluded variable is fixed at the individual optimum presented in Table 4. The goodness of fit is illustrated in Part B through the ability to simulate response changes between predicted and observed data and the residual distribution as a function of each variable. . The 3D response surface graphs are shown in Part A. In each graph, the excluded variable is fixed at the individual optimum presented in Table 4. The goodness of fit is illustrated in Part B through the ability to simulate response changes between predicted and observed data and the residual distribution as a function of each variable.

Y1 Y2 Y1/Y2
Projections in XY plane of the 3D response surfaces for the response criteria Y1 (mg anthocyanin/g dried epicarp), Y2 (mg anthocyanin/g extract), and Y1/Y2 (g extract/g dried epicarp) that describe the most favorable extraction conditions. In each projection, the excluded variable is fixed at the individual optimum presented in Table 4.
The 3D surface graphs translate the parametric results of Table 4 and visually represent the influence of the three independent variables involved in the process (Figure 1). Part A shows 3D response surfaces. In each graph, the combined effects of two variables were represented with the excluded variable positioned at its individual optimal value, which is shown in Table 4. In turn, part B illustrates the capability of the models to predict the obtained results and the residual distribution as a function of each variable. A good agreement between experimental and model-predicted values was observed, with R 2 values higher than 0.95 in all cases (Table 4), which translates the percentage of variability explained by the models. In addition, the errors had constant variance, with the residuals scattered randomly around zero.

Optimal Extraction Conditions and Verification of the Predictive Models
Once the models were statistically validated, they were used determine the optimal extraction conditions that maximize the individual and global responses. Table 4, part C, presents the optimal processing conditions that maximize the recovery of anthocyanins from passion fruit pericarp. It was possible to recover 3.4 ± 0.6 mg of cyanidin-3-O-glucoside from 1 g of plant material when processing Figure 2. Projections in XY plane of the 3D response surfaces for the response criteria Y 1 (mg anthocyanin/g dried epicarp), Y 2 (mg anthocyanin/g extract), and Y 1 /Y 2 (g extract/g dried epicarp) that describe the most favorable extraction conditions. In each projection, the excluded variable is fixed at the individual optimum presented in Table 4.

Optimal Extraction Conditions and Verification of the Predictive Models
Once the models were statistically validated, they were used determine the optimal extraction conditions that maximize the individual and global responses. Table 4, part C, presents the optimal processing conditions that maximize the recovery of anthocyanins from passion fruit pericarp. It was possible to recover 3.4 ± 0.6 mg of cyanidin-3-O-glucoside from 1 g of plant material when processing it at 20 • C for 35 min and using 0% ethanol, while 78 ± 8 mg of the same anthocyanin were obtained per each gram of extract when processing the sample at 20 • C for 78 min and using 29% ethanol (v/v). In order to maximize the amount of anthocyanin in the extract and the recovery rate of the target compound from the dried epicarp, global processing conditions were also calculated ( Table 4). These conditions (t = 38 min, T = 20 • C; S = 0% ethanol, v/v) allowed 9 ± 1 mg of cyanodin-3-O-glucoside/g of extract to be obtained and 3.4 ± 0.5 mg of the same anthocyanin/g of dried epicarp to be recovered, representing a higher extraction yield (37 ± 5%) than the conventional method. Although the improvements obtained with the optimized process are not extremely pronounced in terms of yield and anthocyanin contents compared to the conventional method, the optimized protocol stands out for requiring a much shorter extraction time (37 min vs. 2 h) and just water (vs. 80% ethanol) as extraction solvent. At the industrial level, these processing conditions are thus time-saving, economically attractive, and "greener" and therefore more desired and advantageous. These global conditions were then applied to obtain a new extract and thus evaluate the predictive model's ability to predict the experimental results. The model was experimentally validated with a good correlation between model-predicted and experimental results.

Effect of the Solid-to-Liquid Ratio
The individual R S/L analysis was performed under the optimal conditions predicted by the models developed for each response variable (Table 4), and its effects were assessed at ratios between 5 and 100 g/L. The obtained R S/L responses were consistent with the results obtained in the RSM analysis, and were described by a linear relationship (data not shown). All experimental points were distributed around the equation with only one independent variable and, consequently, the dose-response effect was explained by the slope (m) and intercept (b) of the linear equation.
Although the lower R S/L yielded similar results, the responses decreased as the R S/L increased. The negative values of m show that increasing this ratio causes a decrease in extraction efficiency, resulting in a maximum extraction value at 5 g/L and a minimum at 100 g/L. However, the observed decrease was marked, meaning that the increase of 1 g/L implies the loss of a significant amount of anthocyanin per gram of extract. Such values produce losses of~45% in the maximum experimental value tested, compared to the value extracted at 5 g/L. However, the economic advantages of working at 100 g/L may outweigh the possible benefits of extracting at the optimal R S/L . From the optimization point of view, it was decided that the value used would be around 50 g/L.

Antioxidant Activity
The antioxidant activity of the anthocyanin extracts obtained from passion fruit epicarp using the optimized (optimal extract) and conventional (normal extract) methods was evaluated through its capacity to prevent the formation of thiobarbituric acid reactive substance (TBARS) and the oxidative hemolysis. The results of both in vitro assays are present in Table 5. As observed, the antioxidant activity of the extracts differed significantly (p < 0.05), and the highest EC 50 and IC 50 values were achieved with the normal extract. Thus, the extract obtained under optimal conditions stood out with greater ability to inhibit the TBARS formation and oxidative hemolysis, being suitable for lipid peroxidation inhibition in food products. This result can be explained by the higher concentration of cyanidin-3-O-glucoside found in the optimal extract; nevertheless, other non-identified compounds may also contribute to this bioactivity.  The antioxidant properties of passion fruit extracts have been reported by other authors, which measured this activity by different in vitro assays, such as ABTS, DPPH and FRAP [17][18][19], and pointed out a correlation between total phenolics and antioxidant capacity.

Antimicrobial Activity
The results of the antimicrobial activity of the two extracts are presented in Table 5. The optimal extract showed a slightly higher antimicrobial activity than the normal extract against Listeria monocytogenes and Escherichia coli, with a MIC of 4 mg/mL versus 8 mg/mL, respectively. Regarding the MBC values, 8 mg/mL of optimal extract was enough to kill the tested bacteria, while a concentration above 8 mg/mL of normal extract was generally required. The optimal extract was therefore the most effective one.
Among the six fungi tested, Penicillium ochrochloron was the most sensitive to the optimal extract, with 1 mg/mL being required to inhibit and kill this fungus (Table 5), followed by Trichoderma viride and Aspergillus niger (MICs = 4 mg/mL). The other strains had MIC values of 8 mg/mL for this extract, but still lower than those required of normal anthocyanin extract (>8 mg/mL).

Cytotoxic Activity
The cytotoxic activity of the tested passion fruit epicarp extracts against four tumor cell lines (NCI-H460, MCF-7, HepG2 and HeLa) and primary PLP2 cells are presented in Table 5. The normal extract showed no cytotoxic activity against the tumor cell lines at the tested concentrations (GI 50 > 400 µg/mL), while the anthocyanin extracted obtained under optimized processing conditions exhibited weak cytotoxicity to HepG2 cells, with a GI 50 value of 363 ± 15 µg/mL. Interestingly, none of the extracts showed hepatotoxicity up to a concentration of 400 µg/mL to non-tumor PLP2 cells. To the best of the authors' knowledge, there are no studies reporting the cytotoxicity or hepatotoxicity of passion fruit extracts.

Sample Preparation
Fresh passion fruit samples were supplied by the company KiwiCoop from Oliveira do Bairro, Portugal. The epicarp was separated from the pulp and then freeze-dried (FreeZone 4.5, Labconco, Kansas City, MO, USA) and reduced to powder (~20 mesh), which was homogenized to obtain a representative sample that was kept at −80 • C.

Determination of Colour in Fresh and Dried Epicarp
Color was measured on fresh and dried epicarps with a CR-400, Konica Minolta Sensing colorimeter (Osaka, Japan). Measurements were performed in the CIELAB color space as described by López et al. [20]. Data were processed with Spectra Magic Nx software (CM-S100W 2.03.0006, Konica Minolta).

Extraction and Analysis af Anthocyanins from Dried Epicarp
A conventional extraction was performed as formerly described by Jabeur et al. [12]. Powdered epicarp (1 g) underwent solid-liquid extraction twice (1 h + 1 h) with 80% ethanol acidified with 0.05% citric acid (20 mL). After filtering the supernatant, ethanol was separated in a Büchi R-210 rotary evaporator and water was freeze-dried.
For analysis of anthocyanins, the extract was dissolved in 20% ethanol at 5 mg/mL and filtered through a 0.2-µm syringe filter disk for injection in the HPLC-DAD-ESI/MS system formerly described by Bessada et al. [21], which also summarized the chromatographic conditions and the identification and quantification procedures. The calibration curve (y = 243287x − 1 × 10 6 ; r 2 = 0.9953) used in the quantification was constructed with the standard cyanidin-3-O-glucoside (Extrasynthèse, Genay, France). The results were given in mg/g extract.

Experimental Design
A CCD of 20 runs was applied to optimize the anthocyanin extraction from passion fruit epicarp using RSM. The natural and coded values of the independent variables X 1 (time (t), min), X 2 (temperature (T), • C) and X 3 (solvent (S) or ethanol proportion, % (v/v)) are presented in Table 2.

Extraction Method
Extractions were made as described by Pinela et al. [16]. Powdered epicarp samples (1 g) were mixed with 20 mL of solvent (ethanol/water mixtures acidified with 0.05% citric acid, until pH 3) and processed by stirring at 500 rpm according to the CCD design shown in Table 2, which combines five levels of t (5-85 min), T (20-90 • C), and S (0-100%). After processing, the mixtures were centrifuged (Centurion K24OR, West Sussex, UK) at 500 rpm for 20 min and the supernatants were filtered through filter paper to determine the extraction yield and anthocyanin content.

Extraction Process Optimization by Response Surface Methodology
The extraction yield and anthocyanin content in the filtrates were analyzed and used as response criteria. Anthocyanins were quantified as described in Section 2.3. and the extract weight (%, w/w) was determined gravimetrically by evaporating the solvent from the supernatants [16]. Three response criteria were used for RSM optimization, namely: Y 1 , anthocyanin content in the dried epicarp (mg/g dried epicarp); Y 2 , anthocyanin content in the extract (mg/g extract); and Y 1 /Y 2 , extraction yield (g extract/g dried epicarp).
The response surface models were fitted to the experimental data using the second-order polynomial equation described by Pinela et al. [22]. All fitting procedures, coefficient estimates and statistical analysis were performed as explained by Prieto and Vázquez [23].

Analysis of the Solid-to-Liquid Ratio
The solid-to-liquid ratio (R S/L ) was studied after optimizing the extraction conditions for the other three variables. The R S/L effects were depicted using a linear equation in which a positive slope value indicates an increase in extraction efficiency .

Production of Anthocyanin Extract under Optimal Extraction Conditions and Verification of the Predictive Models
An anthocyanin-rich extract was produced from passion fruit epicarp applying the optimized processing conditions previously established. The sample (1 g) was mixed with 100% water (20 mL) acidified with 0.05% citric acid (until pH 3) and processed at 20 • C for 37.5 min. After centrifugation, the supernatant was filtered and the solvent eliminated to obtain a dry extract. Thereafter, the anthocyanin content was determined to evaluate the model predictive capacity.

Antioxidant Activity
The TBARS formation inhibition and oxidative hemolysis inhibition (OxHLIA) assays were performed as described by Lockowandt et al. [24]. The extract's capacity to inhibit the formation of TBARS was assessed using porcine brain cells as oxidizable substrates and the results were given as EC 50 values (µg/mL). In turn, the extracts capacity to inhibit the oxidative hemolysis was tested using sheep erythrocytes and the results were given as IC 50 values (µg/mL) for a ∆t of 60 min. Trolox was the positive control.

Antimicrobial Activity
The antibacterial activity was tested against Escherichia coli, Salmonella typhimurium, Enterobacter cloacae, Staphylococcus aureus, and Listeria monocytogenes as described by Soković et al. [25], using ampicillin as positive control. The antifungal activity against Aspergillus fumigatus, Aspergillus versicolor, Aspergillus niger, Penicillium funiculosum, Penicillium ochrochloron, and Trichoderma viride was evaluated as described by Soković and van Griensven [26], using ketoconazol as positive control. The results were given as minimum inhibitory, bactericidal, and/or fungicidal concentrations (MIC, MBC, and MFC, respectively, mg/mL).

Cytotoxic Activity
The cytotoxicity against HeLa (cervical carcinoma), HepG2 (hepatocellular carcinoma), MCF-7 (breast adenocarcinoma), and NCI-H460 (non-small-cell lung cancer) cells was tested by the sulforhodamine B assay [27]. Non-tumor PLP2 cells were used to evaluate the hepatotoxic activity. Ellipticine was the positive control. The results were given as GI 50 values.

Statistical Analysis
Color and bioactivity analyses were made in triplicate. SPSS Statistics (23.0, IBM Corp., Armonk, NY, USA) was used for data analysis. A Student's t-test was performed to evaluate statistical differences between the two anthocyanin extracts (p < 0.05).

Conclusions
The coloring and bioactive potential of a bio-based ingredient produced from the dark purple epicarp of passion fruit was evaluated. The dehydrated epicarp used for recovery of anthocyanins had a more intense red color (higher a* value) than the fresh material, which demonstrated the suitability of freeze-drying. For extraction optimization, t, T, and S were combined in a CCD design coupled with RSM, using the extraction yield and cyanidin-3-O-glucoside levels as dependent variables. The developed models were successfully fitted to the experimental data and used to determine the optimal processing conditions that lead to maximum responses, namely anthocyanin content recovered from the epicarp and obtained in the extract. The R S/L analysis showed that 50 g/L is a suitable ratio. The extract obtained by the optimized method revealed a higher bioactivity (TBARS inhibition, antihemolytic, antimicrobial, and antifungal activities and cytotoxicity against cervical carcinoma), in agreement with the higher cyanidin-3-O-glucoside content. Therefore, it is concluded that passion fruit epicarp is a sustainable source of cyanidin-3-O-glucoside and can be used to produce bioactive colorants. Table S1: Anthocyanin identification and content in passion fruit epicarp. The retention time (Rt), wavelength of maximum absorption in the visible region (λ max ), and mass spectral data are presented, Figure S1: Passion fruit epicarp anthocyanin profile extracted at 520 nm. The identified peak corresponds to cyanidin-3-O-glucoside, Figure S2: Diagram of the different steps carried out to optimize the extraction conditions for recovery of anthocyanins from passion fruit epicarp.