Glycerol-Based Deep Eutectic Solvents for Simultaneous Organosolv Treatment/Extraction: High-Performance Recovery of Antioxidant Polyphenols from Onion Solid Wastes

Abstract

Onion solid wastes (OSW) are a food side-stream with high polyphenolic load and distinguished polyphenolic profile. This study was conducted in order to explore a novel methodology of production of polyphenol-enriched extracts with antioxidant properties from OSW, using glycerol and green deep eutectic solvents (DES), composed of glycerol/citric acid and glycerol/sodium acetate. The organosolv process developed was appraised by establishing models based on severity, but also response surface methodology. Using a linear model, it was, for the first time, proposed that there is a direct correlation between the yield of the process in total polyphenols and the combined severity factor. Furthermore, response surface optimization enabled the establishment of linear models to predict the effects of time and temperature on the total polyphenol extraction yield. Out of the solvents tested, the DES composed of citric acid and glycerol was found to provide the highest yield in total polyphenols (87.90 ± 3.08 mg gallic acid equivalents per g dry mass) at significantly higher combined severity. However, the extraction efficiency of this solvent was virtually equal to that of the two other solvents tested. On the other hand, the polyphenolic composition of the extract obtained with the glycerol/citric acid DES was characterized by exceptionally high quercetin concentration. This extract also displayed the highest antioxidant activity. Based on the evidence emerged, it was proposed that OSW polyphenol extraction with the DES glycerol/citric acid could be used for production of extracts enriched in the bioactive flavonoid quercetin, with enhanced antioxidant activity. Moreover, using this green methodology, 27.59 ± 0.09 g of pure quercetin could be recovered out of 1 kg OSW. Thus, this methodology could be employed as a sustainable means of producing quercetin, through valorization of food wastes in a biorefinery context.

1. Introduction

Based on the latest Food and Agricultural Organization of UN (FAO) estimates, almost 1.3 billion tons of food waste is generated globally per annum, which represents roughly a third of all food produced. This residual bio-material derives mainly from losses of food, production residues, and postharvest rejected tissues, as well as from discarded foods (unsold or spoiled) and consumption leftovers (household, restaurants, catering facilities, etc.). Dumping of such materials, which are rich in organic substances, at landfills may create severe hygiene problems and pose important economic, environmental, and social pressure [1]. On the other hand, food waste biomass is now largely acknowledged as a bio-resource, which, upon appropriate management, could be used to produce energy, platform chemicals, fertilizers, and high value-added substances. Such an option would contribute towards establishing circular economy models for supporting sustainable development by enabling food waste valorization and minimizing waste disposal [2].

Onion (Allium cepa) is one of the commonest consumed vegetables around the globe. The total onion world production has been estimated to be approximately 93.23 million tons, and onion processing, either at a domestic or industrial level, results in the generation of a significant amount of waste. This waste material may include worthless/damaged/malformed bulbs, apical trims, and dry/semi-dry outer layers. Only in Europe it has been reported that onion production and processing give rise to about 0.6 million tons of waste on an annual basis [3]. This residual biomass is rather underutilized because it is not suitable for fodder production due to the strong pungent smell, but it cannot be used as a fertilizer either. However, onion solid wastes (OSW) are a source rich in bioactive substances [4], such as quercetin and quercetin conjugates, including quercetin 4′-O-glucoside [5]. OSW also contain a range of quercetin derivatives, arising from peroxidase-catalyzed oxidations of quercetin and its glycosides. These unique OSW constituents were shown to exhibit peculiar bioactivities [6,7] and, therefore, OSW may be considered as a distinguished polyphenol-bearing biomaterial.

Ever since OSW were explicitly proposed as a rich source of polyphenolic phytochemicals [8], numerous extraction processes have been deployed for effective retrieval of OSW polyphenols, including methanol, water/ethanol or acetone mixtures, supercritical fluid extraction, microwave/ultrasound-assisted extraction, and subcritical water extraction [9], but also water/glycerol mixtures [10] and deep eutectic solvents [11]. These technologies may be considered as sustainable processes, but solvents under tight control by state laws (ethanol) or methodologies of high operating cost (subcritical/supercritical) may pose serious drawbacks in potent industrial applications. Thus, the search for alternative approaches is paramount.

Aside from eco-friendly solvents, the deployment of processes that could enable increase in extraction yield is also a key issue related to polyphenol extraction. In several instances, decomposition of complex polysaccharides and other bio-polymer networks encountered in plant tissues has been demonstrated to enable significant increase in polyphenol extraction yield. This is because of polyphenol liberation from complex matrices and entrainment into the liquid phase [12,13]. On this basis, treatments with solvents at relatively high temperatures, termed as organosolv treatments [14], have recently been proposed as a very effective means of recovering olive leaf polyphenols [15]. This being the case, this study was undertaken with the aim of examining the application of organosolv treatment on the polyphenol extraction from OSW. The solvents used were green glycerol-based deep eutectic solvents, and process appraisal was based on both extraction efficiency and severity, in conjunction with the polyphenolic profile of the extracts and their antioxidant properties. To the best of the authors’ knowledge, such an approach for polyphenol recovery from OSW is heretofore unreported.

2. Materials and Methods

2.1. Chemicals

Quercetin and spiraeoside (quercetin 4′-O-glucoside) were from Sigma-Aldrich (Darmstadt, Germany). Iron(III) chloride hexahydrate (FeCl₃) was obtained from Merck (Darmstadt, Germany). 2,4,6-Tris(2-pyridyl)-s-triazine (TPTZ) was purchased from Fluka (Steinheim, Germany). Citric acid and L-ascorbic acid were obtained from Carlo Erba (Milano, Italy). Glycerol anhydrous, sodium acetate anhydrous, and sodium carbonate anhydrous were from Penta (Prague, Czech Republic). 2,2-Diphenyl-1-picrylhydrazyl (DPPH) was from Alfa Aesar (Karlsruhe, Germany). Folin-Ciocalteu regent, absolute ethanol, and gallic acid monohydrate were from Panreac (Barcelona, Spain). All solvents used for chromatographic purposes were HPLC grade.

2.2. Deep Eutectic Solvent Production

The deep eutectic solvents (DES) synthesized were glycerol/citric acid [15] and glycerol/sodium acetate [16], with molar ratios of 4:1, termed as GL-CA and GL-SA, respectively. DES synthesis was accomplished by mixing exact masses of the hydrogen bond donor (glycerol) and hydrogen bond acceptor (citric acid or sodium acetate) and heating the mixture at 70 °C under stirring at 500 rpm for nearly 60 min. The transparent liquids formed were left to acquire ambient temperature and stored for a period of several weeks, over which stability (crystal appearance) was visually inspected.

2.3. Onion Solid Waste (OSW)

OSW composed essentially of the dry/semi-dry layers of the bulb and the apical trims was obtained from a catering facility (Karditsa, Greece) and further processed within 2 h from the receipt. The OSW used were from red onions (Allium cepa), free from infections and damages. Upon receipt, OSW were placed over aluminum discs to form layers of almost 1 cm thickness and dried into an oven (Binder BD56, Bohemia, NY, USA) at 75 °C for 8 h [17]. A table domestic mill (Tristar, Tilburg, the Netherlands) was used to grind the dried material, which was then sieved to produce a powder with average particle diameter of 0.120 mm. This material was stored at −40 °C until used.

2.4. Extraction Process

The solvents used in all experiments were glycerol, termed as GL, and the DES GL-CA and GL-SA. These solvents were always used as 90% (w/w) mixtures with water. The pH of these mixtures was 3.50, 1.40, and 8.07 for GL, GL-CA, and GL-SA, respectively. All extractions were carried out with the assistance of a hotplate (Witeg, Wertheim, Germany), operated at 500 rpm, and adjusted at the desired temperature. To perform the extractions, 1 g of OSW was mixed with 10 mL of solvent in a 25 mL Duran™ vial immersed in an oil bath. After the extraction, the mixtures were centrifuged at 10,000× g for 10 min. Control extractions were accomplished with water and 60% (v/v) ethanol for 60 and 180 min, respectively, at 70 °C [18].

2.5. Response Surface Methodology—Process Optimization

The organosolv treatment/extraction process with each of the solvents tested (GL, GL-CA, and GL-SA) was optimized by considering the two most influential process variables, temperature, T, and time, t [18]. Optimization was accomplished with response surface methodology and a central composite experimental design, which included, in total, 11 design points. Both variables (t, T) were codified in 3 levels, −1, 0, and 1, as dictated by the experimental design. The codification method and other details pertaining to the experimental design have been described analytically elsewhere [19]. The coded and actual levels of the variables are shown in

Table 1. Coded and actual levels of the independent variables used for performing response surface optimization.

Process Variables	Codes	Coded Variable Level
		−1	0	1
t (min)	X1	15	30	45
T (°C)	X2	105	120	135

.

Selection of the actual values (ranges) for the variables considered was based on literature data [15], as well as preliminary experiments. A minimum significance level of 95% was considered to evaluate the overall significance of the models (R², p) and the significance of each coefficient of the equations (models). The evaluation was based on lack-of-fit and analysis of variance (ANOVA) tests.

2.6. Total Polyphenol and Antioxidant Activity Determination

For the determination of total polyphenol content, the Folin–Ciocalteu methodology was used, as described elsewhere [20]. Using gallic acid as calibrating standard, the yield in total polyphenols (Y_TP) was given as mg gallic acid equivalents (GAE) per g dry mass (DM). The antioxidant activity was assessed by determining both the ferric-reducing power (P_R) and the antiradical activity (A_AR). The first test was performed using the stable radical DPPH and expressed as μmol DPPH per g DM. The second test was based on the use of TPTZ as the chromophore probe, and the results were given as μmol ascorbic acid equivalents (AAE) per g DM, using ascorbic acid as calibrating standard. The analytical protocols for both assays have been previously reported in detail [19].

2.7. Appraisal of Extraction Efficiency

The extraction efficiency was appraised using the extraction efficiency factor (F_EE), as recently proposed [18]. First, the extraction efficiency was determined as follows:

$$\mathrm{EE}=\frac{{\mathrm{Y}}_{\mathrm{TP}}}{t\times T}\mathrm{mg}{\mathrm{g}}^{-1}°{\mathrm{C}}^{-1}$$

(1)

where t and T are the values for optimal extraction time and temperature, and Y_TP the maximum total polyphenol yield (mg GAE g⁻¹ DM), as revealed by the response surface methodology. Then, the extraction efficiency factor (F_EE) was determined:

F_EE = −log(EE)

(2)

2.8. Appraisal of Process Severity

The severity of the extraction process was appraised using the severity factor (SF), and the combined severity factor (CSF) [21]. These factors provide a measure of the severity of organosolv processes, which results from combinations of temperature and time, and enable comparison of the different extraction conditions. Process severity may be calculated as follows:

$$R_0=t\times e^{\left(\frac{T-100}{14.75}\right)}$$

(3)

The severity factor may then be defined as:

SF = logR₀

(4)

The combined severity is calculated by the following expression:

$${R_0}^{\prime }=10-\mathrm{pH}\times \times e^{\left(\frac{T-100}{14.75}\right)}$$

(5)

Thus, the combined severity factor is determined by the equation:

CSF = logR₀′ = logR₀ − pH

(6)

where R₀ is the severity, R₀′ the combined severity, and 100 °C the reference temperature. The value 14.75 is an empirical parameter associated with temperature and activation energy.

2.9. Liquid Chromatography−Diode Array−Mass Spectrometry (LC−DAD−MS)

For both qualitative and quantitative chromatographic determinations, a published methodology was adopted [22]. The equipment employed was a Finnigan (San Jose, CA, USA) P4000 pump, a UV6000LP diode array detector, and a Finnigan AQA mass spectrometer. All analyses were carried out on a Fortis RP-18 column, 150 mm × 2.1 mm, 3 μm, a 10-μL injection loop, at a constant temperature of 40 °C. The eluents used for separation were (A) 0.5% aqueous formic acid and (B) 0.5% formic acid in MeCN/water (6:4) at a constant flow rate of 1 mL min⁻¹. The volume of the sample injected was 20 μL, and the elution program implemented was: 100% A to 60% A in 40 min; then, 60% A to 50% A in 10 min; and, finally, 50% A to 30% A in 10 min, which was kept constant for an additional 10 min. Mass spectra acquisition was carried out with electrospray ionization (ESI) in positive ion mode. Quantification was based on external standard calibration curves, constructed with spiraeoside (R² = 0.9966) and quercetin (R² = 0.9990). All standards were prepared in methanol and the ranges used for the calibration curves were from 0 to 50 μg mL⁻¹.

2.10. Statistical Processing

Response surface methodology, the design of experiment, all associated statistics (ANOVA and lack-of-fit), and distribution analyses were elaborated with JMP™ Pro 13 (SAS, Cary, NC, USA). SigmaPlot™ 12.5 (Systat Software Inc., San Jose, CA, USA) was employed for linear regressions. All treatments were carried out at least twice, and quantitative determinations in triplicate. Values given are means ± standard deviation (sd).

3. Results and Discussion

3.1. Effect of Process Severity on Polyphenol Recovery

The concept of adopting process severity for evaluating polyphenol extraction has recently been proposed [15,23,24]. But, while severity may be used as a means of assessing the combined effect of resident time and temperature on polyphenol recovery, it has been well documented that a severity expression also considering the pH of the treatment/extraction medium may provide a more representative image of severity effects [25] and could be used for both acidic and alkaline media [26]. On these ground, and in order to ascertain whether process severity may actually be correlated to polyphenol extraction yield, a series of experiments was performed under different sets of t and T (

Table 2. Time and temperature combinations for the organosolv treatments/extractions performed, along with the corresponding severity factors (SF), combined severity factors (CSF) and total polyphenol yields (Y_TP).

T (°C)	t (min)	SF	CSF			YTP (mg GAE g−1 DM)
			Solvent			Solvent
			GL	GL-CA	GL-SA	GL	GL-CA	GL-SA
105	15	1.32	−2.18	−0.08	−6.75	52.99	67.73	67.32
105	30	1.62	−1.88	0.22	−6.45	55.54	68.07	69.61
105	45	1.80	−1.70	0.40	−6.27	63.73	69.72	69.03
120	15	1.76	−1.74	0.36	−6.31	61.63	74.00	72.74
120	30	2.07	−1.43	0.67	−6.00	67.34	77.13	78.45
120	45	2.24	−1.26	0.84	−5.83	69.57	79.64	78.60
135	15	2.21	−1.29	0.81	−5.86	69.18	84.90	80.67
135	30	2.51	−0.99	1.11	−5.56	74.63	87.31	81.09
135	45	2.68	−0.82	1.28	−5.39	77.68	87.90	87.28

). For each set, the combined severity factor (CSF) was calculated to take into account the pH of the treatment/extraction solvent. The treatment/extraction performed with GL-SA was shown to be of low severity, giving CSF values that ranged from −6.75 to −5.39. On the other hand, the use of GL gave CSF within the range of −2.18 to −0.82, but treatments/extractions with GL-CA were of considerably higher severity, with CSF lying between −0.08 and 1.28. These values were significantly lower than 2.24 reported for polyphenol extraction from waste orange peels with a mixture of aqueous glycerol and HCl [27].

In order to examine whether CSF was correlated with the extraction yield in total polyphenols (Y_TP), linear regression analysis was performed (Figure 1). These correlations provided the following linear equations (models):

Y_TP(GL) = 18.58CSF_GL + 93.23 (R² = 0.97, p < 0.0001)

(7)

Y_TP(GL-CA) = 17.55CSF_GL-CA + 66.42 (R² = 0.90, p < 0.0001)

(8)

Y_TP(GL-SA) = 14.78CSF_GL-SA + 165.47 (R² = 0.91, p < 0.0001)

(9)

This outcome provided proof of the strong linear correlation between severity of the process and the yield in total polyphenols, which was also in line with recent findings on polyphenol extraction from wheat bran [23]. As judged by R² and p, and considering a significance level of 95%, it could be argued that all three models had a high degree of reliability. Thus, using these models, credible predictions for Y_TP could be made based on CSF. Especially for the case of treatments/extractions with glycerol, the model had excellent adjustment to the experimental data (R² = 0.97). Taking into account that the pH of GL-CA, GL, and GL-SA were 1.40, 3.50, and 8.07, respectively, then it could be supported that the strong correlation found between Y_TP and CSF was independent of the pH of the solvent used.

On the ground of these results, the highest Y_TP for the treatment/extraction with GL could be achieved at 135 °C for 45 min (Table 2). However, the Y_TP attained at 135 °C for 30 min was lower by only 3.9%. Therefore, the latter set of conditions could be chosen as the most appropriate. The same held true for the treatment/extraction with GL-CA, where there was virtually no difference in the Y_TP found for the treatment at 135 °C, either at 45 or 30 min. In the case of GL-SA, the difference in Y_TP, when these two sets of conditions were employed, was 7.1%. Thus, a treatment/extraction at 135 °C for 45 min would give the highest performing combination. Under these conditions, the process with GL-SA had significantly lower severity compared to both GL and GL-CA (p < 0.05).

3.2. Process Optimization—Extraction Efficiency

With the aim of ascertaining the results drawn from the linear models, response surface optimization was also performed. The design of the process included t and T, to assess their effect on the response (Y_TP) but, also, to reveal possible cross effects between the independent (process) variables. Both the ANOVA and the lack-of-fit tests served as criteria for evaluating the model fitting, based on the closeness of the actual (measured) and predicted Y_TP values (

Table 3. The full experimental design (design points) employed for the response surface optimization and the corresponding predicted and measured values of the response.

Design Point	Process Variables		Response (YTP, mg GAE g−1 DM)
	t (min) (X1)	T (°C) (X2)	GL		GL-CA		GL-SA
			Measured	Predicted	Measured	Predicted	Measured	Predicted
1	15 (−1)	105 (−1)	52.99	52.26	67.73	66.74	67.32	67.31
2	15 (−1)	135 (1)	69.18	69.79	84.90	84.60	80.67	79.22
3	45 (1)	105 (−1)	63.73	62.44	69.72	69.93	69.03	69.58
4	45 (1)	135 (1)	77.68	77.73	87.11	87.90	87.28	86.39
5	15 (−1)	120 (0)	61.63	61.76	74.00	75.29	72.74	74.21
6	45 (1)	120 (0)	69.57	70.82	79.64	78.53	78.60	78.93
7	30 (0)	105 (−1)	55.54	57.57	68.07	68.79	69.61	69.07
8	30 (0)	135 (1)	74.63	73.98	87.31	86.71	81.09	83.43
9	30 (0)	120 (0)	66.95	66.51	75.96	77.36	77.80	77.20
10	30 (0)	120 (0)	65.96	66.51	78.31	77.36	79.00	77.20
11	30 (0)	120 (0)	68.03	66.51	78.02	77.36	76.60	77.20

). Details concerning the statistical analyses performed may be seen in Figures S1–S3 (Supplementary Materials) (

Table 4. Models derived from the response surface optimization.

Solvent	Equations (Models)	R2	p
GL	YTP = 66.51 + 4.53X1 + 8.21X2	0.98	0.0003
GL-CA	YTP = 77.36 + 1.62X1 + 8.96X2	0.98	0.0002
GL-SA	YTP = 77.20 + 2.36X1 + 7.18X2	0.96	0.0018

).

All three models had R² equal to or higher than 0.96, and the p values were highly significant (based on a confidence interval of 95%). These data concurred that the models showed fine adjustment to the experimental values. The 3D plots that visualized the models are given in Figure 2 to depict at a glance the effect of the process variables on the response (Y_TP) and illustrate the differences among the solvents tested. For all three solvents, both t (X₁) and T (X₂) were significant, with positive effect on Y_TP, and, therefore, increases in both variables would result in increased Y_TP, in accordance with the models based on CSF (Equations (7)–(9)). No cross terms were shown to be significant, suggesting that the interactions of t and T did not affect Y_TP. Using the desirability function (Figures S1–S3), the optimal predicted conditions (t, T) and the maximum Y_TP values were calculated (

Table 5. Optimal conditions predicted to provide maximum response (Y_TP) and the corresponding extraction efficiency factors (F_EE).

Solvent	Maximum Predicted Response (mg GAE g−1 DM)	Optimal Conditions		FEE
		t (min)	T (°C)
GL	77.73 ± 3.46	45	135	1.89
GL-CA	87.90 ± 3.08	45	135	1.84
GL-SA	86.39 ± 3.99	45	135	1.85

). For all three solvents tested, the accomplishment of maximum Y_TP required a resident time of 45 min and a temperature of 135 °C. These values coincided with those that yielded maximum CFS (Table 2).

The determination of optimum conditions (t, T) also enabled the estimation of the extraction efficiency factor (F_EE), according to Equation (2), which were very close and varied from 1.84 to 1.89 (Table 5). Considering the categorization proposed by Morsli et al. [18], the extraction with any of the solvents tested was of moderate efficiency. Extraction with GL-CA and GL-SA afforded Y_TP that was higher by 11.6 and 10% compared to GL, and they were equally effective in OSW polyphenol extraction.

To have a more integrated image of the efficiency of the three solvents tested, a comparison with control solvents (water and 60% ethanol) was also considered (Figure 3). This comparison highlighted the efficiency of GL-CA, which was 70.4 and 43.8% higher than those of the aqueous and hydroethanolic extractions.

The highest Y_TP (87.90 ± 3.08 mg GAE g⁻¹ DM), achieved with GL-CA, was of comparable magnitude to that reported for ultrasound-assisted OSW extraction (88.03 mg GAE g⁻¹ DM) with a glycerol/sodium potassium tartrate deep eutectic solvent [16] and microwave-assisted extraction (80.45 mg GAE g⁻¹ DM) with a choline chloride/urea deep eutectic solvent [28]. Similar levels of 90.07 mg GAE g⁻¹ DM were attained with ultrasound-assisted aqueous glycerol extraction [10], but, significantly, a higher yield (137.50 mg caffeic acid equivalents g⁻¹ DM) has been reported for OSW extraction with a glycerol/sodium propionate deep eutectic solvent [11]. The use of hydroethanolic solvents at ambient temperature has been shown to afford Y_TP levels of approximately 81.48–93.42 mg GAE g⁻¹ DM [29,30]. However, the range of total polyphenols in OSW may vary largely, from 14.55 to 288.74 mg GAE g⁻¹ DM [31].

3.3. Polyphenolic Composition

The polyphenolic composition of the extracts produced with GL, GL-CA, and GL-SA under optimized conditions and the control extracts obtained with 60% ethanol and water was profiled by HPLC, considering the two major polyphenolic metabolites, spiraeoside (quercetin 4′-O-glucoside) and quercetin [32,33]. The chromatographic traces recorded at 360 nm (Figure 4) for all extracts tested were indeed dominated by the peaks corresponding to these two compounds. The highest yield in spiraeoside was achieved by extracting with GL (9.88 mg g⁻¹ DM), which was very close to that obtained with aqueous ethanol (9.77 mg g⁻¹ DM) (

Table 6. Quantitative information on the major flavonols detected in the extracts obtained. The values are average of three determinations (±standard deviation).

Peak #	Compound	Yield (mg g−1 DM)
		Water	AqEt 1	GL	GL-CA	GL-SA
1	Spiraeoside	3.62 ± 0.10	9.77 ± 0.03	9.88 ± 0.03	2.71 ± 0.01	0.32 ± 0.00
2	Quercetin	1.69 ± 0.04	7.19 ± 0.03	14.25 ± 0.01	27.59 ± 0.09	0.49 ± 0.02
	Sum (flavonols)	5.31	16.96	24.13	30.30	0.81
	Spiraeoside/Quercetin ratio	2.14	1.34	0.69	0.10	0.65

¹ Denotes extraction with 60% (v/v) aqueous ethanol; values assigned with different letters within rows are statistically different (p < 0.05).

). On the other hand, extraction with GL-CA afforded the highest yield in quercetin (27.59 mg g⁻¹ DM), which was 73.9 and 48.3% higher than those found for the extractions with aqueous ethanol and GL, respectively.

Overall, GL-CA was shown to be the most efficient solvent, providing a total flavonol yield of 30.30 mg g⁻¹ DM. By contrast, GL-SA provided extracts with particularly low concentration of both spiraeoside and quercetin. This finding led to the assumption that the alkaline pH of GL-SA (8.07) was rather detrimental for both substances, while the hump on the baseline in the chromatogram (Figure 4) manifested the formation of unresolved material, most probably of polymeric nature. It should be highlighted that quercetin is very unstable under alkaline conditions and it decomposes, yielding a spectrum of breakdown products [34]. Thus, it would be reasonable to presume that both spiraeoside and quercetin suffered extensive degradation during extraction with GL-SA, owed to both the alkaline pH and high temperature.

Furthermore, it was also noteworthy that important differences were seen for the relative amounts of spiraeoside and quercetin in the extracts examined. Thus, while the ratio of spiraeoside/quercetin concentration in the aqueous extract was 2.14, in the extract produced with GL-CA, this ratio was 0.10. Considering that the pH of GL-CA was 1.40, then it could be presumed that hydrolysis of spiraeoside to quercetin took place. Such a phenomenon has been observed in acidic DES made of choline chloride and various carboxylic acids, where simultaneous extraction and deglycosylation of quercetin glycosides was accomplished [35].

On the other hand, evidence for flavone glycoside hydrolysis has been provided by organosolv treatment of olive leaves with slightly alkaline DES, composed of glycerol and sodium citrate [15]. Acidic pH favors hydrolysis of various flavonoid glycosides to their corresponding aglycones, and such reactions have been used for flavonoid determination [36]. However, flavonoid glycoside hydrolysis has also been performed at near neutral pH upon hydrothermal treatment [37]. Moreover, high processing temperatures (130–165 °C) have been proven an effective means of converting quercetin glycosides into quercetin [38]. Based on such evidence, it would be supported that the combination of a relatively high treatment temperature (135 °C) and the acidic pH of the GL-CA contributed in spiraeoside hydrolysis and quercetin formation. The content of pure quercetin thus obtained was 27.59 mg g⁻¹ DM (Table 6) and represented 93% of the total flavonols. This level was equal to that attained with microwave-assisted extraction and hydroethanolic solvent [39] and significantly higher than those reported in other studies, varying from 5.32 [40] to 17.32 mg g⁻¹ DM [41].

3.4. Antioxidant Properties

To ascertain whether the compositional differences of the extracts could impact their antioxidant properties, the ferric-reducing power (P_R) and the antiradical activity (A_AR) were determined. The extract obtained with GL-CA displayed the highest A_AR, which was significantly higher (p < 0.05) than the corresponding values of the extracts produced with glycerol, water, and aqueous ethanol (Figure 5A). However, there was no statistical difference between GL-CA and GL-SA extracts. Likewise, the GL-CA extract exhibited the highest P_R, which was significantly higher than the corresponding values of the GL-SA, aqueous, and hydroethanolic extracts but not from the GL extract (Figure 5B).

This outcome implied that GL-CA provided extract with increased antioxidant activity, most probably owing to the higher concentration of quercetin, which is a powerful antioxidant [42]. To test this hypothesis, the ratio of spiraeoside/quercetin of each extract was plotted against A_AR and P_R (Figure 6). From the linear regressions established, it was demonstrated that this ratio was significantly linked to A_AR (R² = 0.90, p = 0.0145) but correlation with P_R was exceptionally high (R² = 0.97, p = 0.0019). Thus, extracts where quercetin concentration was higher showed enhanced A_AR and P_R. Therefore, hydrolytic reactions leading to quercetin liberation from spiraeoside may reinforce the antioxidant activity of the extracts.

Such a tendency has been highlighted by early studies, where quercetin formation from its glycosides through onion fermentation enhanced the antioxidant activity [43]. Deglycosylation of rutin (quercetin 3-O-rutinoside) and isoquercetin (quercetin-3-O-glucoside) and formation of quercetin also promoted higher antioxidant activity in poly(L-lactic acid) films [38], while apple pomace extracts produced with deep eutectic solvents showed higher antioxidant activity due to quercetin glycoside hydrolysis and generation of quercetin aglycone [35].

The enhancement of antioxidant activity as a result of glycoside hydrolysis and quercetin liberation would be anticipated, in the light of previous examinations which demonstrated the antioxidant supremacy of quercetin over several of its glycosides [42]. Thus, it could be suggested that the use of processes that enable quercetin formation through quercetin glycoside hydrolysis might be an effective methodology to produce extracts with enhanced antioxidant activity.

Several DES have been found to display outstanding performance with regard to flavonoid extraction; yet, an issue that would hamper potential industrial applications is solvent removal to separate the solvent from the extract. Although several methodologies (i.e., microporous resins, anti-solvents, back extraction, etc.) have been proposed, the increase in unit operations, with the associated increases in energy and time, dictates that such steps could be circumvented. This could be accomplished by using the system in its entirety [44], without the need for extract isolation or purification. In such a case, solvent would serve as an integral part of the final product formulation, and solute (or solvent) recovery could be efficiently bypassed. Thus, polyphenol-enriched extracts could be directly incorporated into foods/pharmaceuticals/cosmetics. Considering that current trends suggest ingredient production based on functionality rather than purity, the extracts produced could be used for specific applications rather than for general use. Research advancements in such a framework could offer innovative opportunities and development of processes with high scaling-up potential [45].

4. Conclusions

The simultaneous organosolv treatment/extraction may be regarded as a new trend in the recovery of phytochemicals from agri-food wastes. In this framework, the present study demonstrated that onion solid waste polyphenols may be very effectively recovered using a glycerol/citric acid deep eutectic solvent. The process, approached both with a model implicating the combined severity factor and response surface methodology, was shown to depend on temperature and time in a linear manner. The total polyphenol yield achieved was very satisfactory, providing extracts enriched in quercetin, with enhanced antioxidant activity. This was probably enabled through hydrolytic reactions favored by the low pH of the glycerol/citric acid deep eutectic solvent. Currently, work is in progress to ascertain whether acidic solvents could contribute towards producing extracts with increased amounts of quercetin, which is a multifunctional biomolecule.