Evaluation of Antioxidant Properties of Residual Hemp Leaves Following Optimized Pressurized Liquid Extraction

Abstract

Cannabis sativa, often called hemp, is a medicinal plant belonging to the Cannabaceae family and is widely recognized for its therapeutic applications. After the industrial supercritical CO₂ extraction method, hemp residue biomass was recovered, and a significant quantity of bioactive compounds was identified. Therefore, it is of paramount importance to study whether they can be further exploited using green techniques. In the present work, hemp leaf residues were treated using two extraction techniques: one conventional stirring extraction (STE) and one green pressurized liquid extraction (PLE). The latter technique is a promising and swift method for the efficient extraction of valuable molecules from natural sources. The two techniques were optimized through Response Surface Methodology, and the optimized parameters were the appropriate solvent, temperature, and extraction duration. The aim was to maximize the yield of bioactive compounds (polyphenols, flavonoids, and ascorbic acid) from hemp leaf residue and evaluate their antioxidant activity using the most appropriate technique. The results showed that after three 5 min PLE cycles, the recovered individual polyphenols were comparable (p > 0.05) to a 45 min STE (19.34 and 20.84 mg/g, respectively), as well as in antioxidant capacity assays and other bioactive compounds. These findings emphasize the efficacy of the rapid PLE approach as an effective extraction technique to enhance the value of hemp leaf residues while maximizing the concentration of high-added value molecules.

1. Introduction

Cannabis sativa L. (hemp) is among the oldest cultivated plants, utilized in various applications across the textile, construction, and paper industries, as well as in the nutritional, medicinal, and cosmetic sectors [1]. It is a dioecious plant that is a prominent member of the Cannabaceae family. It is one of the most cultivated and used species in the world, flourishing in a variety of habitats, soils, altitudes, and climatic conditions [2]. Currently, it is predominantly acknowledged that the genus comprises a single species, C. sativa L., which is divided into two subspecies: ssp. sativa (characterized by a low concentration of the psychoactive compound ∆⁹-tetrahydrocannabinol (THC), typically less than 0.3% in the dry weight of the upper third of flowering plants) and ssp. indica, referred to as drug (medical) cannabis (noted for its elevated THC levels) [3]. Throughout many periods, humans have cultivated it for diverse uses, including sustenance, recreation, seed oil production, industrial fiber, religious and spiritual rites, and medicinal uses [4].

Hemp contains a variety of bioactive compounds, including cannabinoids, flavonoids, phenolics, terpenes, and alkaloids [5]. Hemp also includes several polyphenols that confer considerable antioxidant properties [6,7]. It contains at least 554 identified compounds, among which are 150 phytocannabinoids, terpenophenolic active molecules almost unique to this plant, and other phytochemical compounds like terpenoids, flavonoids, and oxylipins [8]. The plant contains more than 400 bioactive compounds, including cannabinoids, terpenes, flavonoids, and other phenolic chemicals that exert advantageous effects on the human body [2,9,10,11]. Hemp has been subjected to many extraction techniques, such as supercritical fluid extraction, ultrasonication-assisted extraction, microwave-assisted extraction, pulsed electric field-assisted extraction, hydrodistillation, and stirring [4,6,12]. Among the innovative extraction techniques, pressurized liquid extraction has also been applied to hemp stalks and seeds.

Pressurized liquid extraction (PLE) has been employed in numerous research studies to extract contaminants from natural matrices, including soils, while others have focused on extracting metabolites from plants [13]. This technique exerts pressure to elevate the extraction solvent temperature beyond its boiling point. Usually, the applied temperatures vary between 50 and 200 °C, while the pressure level varies between 35 and 200 bar [4]. Consequently, it improves extraction efficiency by achieving elevated temperatures compared to traditional extraction methods. High pressure enhances mass transfer by increasing cell permeability. Conversely, elevated temperatures facilitate the diffusion of the solvent into the sample by decreasing the solvent’s viscosity. Additionally, PLE enhances the extraction rate by increasing the solubility of the targeted compound and facilitating mass transfer. This procedure is easy, provides safe and quick extraction, minimizes solvent use, reduces extraction duration, and offers high reproducibility and precision [4,14].

In recent years, considerable emphasis has been placed on the necessity of alleviating environmental impact. Numerous sectors are adopting novel, eco-friendly extraction methods to minimize the adverse effects on the environment. These techniques produce greater yields in reduced durations and with lowered solvent usage, which is frequently detrimental. Furthermore, utilizing by-products from the food industry has emerged as a contemporary trend aimed at minimizing environmental waste. Therefore, this research utilizes hemp var. Futura 75 by-products, specifically hemp leaves that have been previously treated with supercritical CO₂. Hemp residual leaves are subjected to PLE, and the appropriate solvent, temperature, and extraction duration are examined. Simultaneously, the optimization of the same factors in simple stirring extraction (STE) was carried out to compare the two techniques. The optimization was carried out via Response Surface Methodology with the Box–Behnken design to analyze the results of the extraction.

2. Materials and Methods

2.1. Chemicals and Reagents

Anhydrous sodium carbonate (≥99% w/w) and rutin hydrate (≥94% w/w) were obtained from Penta (Prague, Czech Republic). Folin–Ciocalteu reagent, gallic acid (≥99% w/w), and ethanol (≥99.8% v/v) were obtained from Panreac Co. (Barcelona, Spain). Hydrochloric acid (37% w/w), methanol (≥99.8% v/v), aluminum chloride (≥99% w/w), L-ascorbic acid (≥99% w/w), trichloroacetic acid (≥99% w/w), 2,4,6-tris(2-pyridyl)-s-triazine (TPTZ) (≥98% w/w), 2,2-diphenyl-1-picrylhydrazyl (DPPH^•) (≥90% w/w), and β-carotene analytical standard (≥95% w/w) were obtained from Sigma-Aldrich (Darmstadt, Germany). All chemical standards used for the HPLC analysis of polyphenols were purchased from MetaSci (Toronto, ON, Canada) and were of HPLC grade (≥99% w/w). Iron (III) chloride (≥99% w/w) was acquired from Merck (Darmstadt, Germany). A deionizing column was employed to generate deionized water for all experiments that were conducted.

2.2. Instrumentation

The dried waste was grounded using an electric milling machine to reduce the grain diameter and expand the exposure contact in the extraction procedure. Then, the grounded powder was sieved using an Analysette 3 PRO (Fritsch GmbH, Oberstein, Germany) sieving machine. To conduct the extraction experiments, a stirring hotplate from Heidolph Instruments GmbH & Co. KG (Schwabach, Germany) and a pressurized liquid extraction (PLE) apparatus from Fluid Management Systems Inc. (Watertown, MA, USA) were employed. A NEYA 16R centrifuge from Remi Elektrotechnik Ltd. (Palghar, India) was utilized to separate the supernatant from the solid residue after the stirring extraction process. The supernatant extracts used in assays demanding thermostatic temperature were kept in an Elmasonic P70H ultrasonic bath from Elma Schmidbauer GmbH (Singen, Germany), whereas spectrophotometric analyses were done in a Shimadzu UV-1900i UV/Vis spectrophotometer (Kyoto, Japan). Finally, a Shimadzu CBM-20A high-performance liquid chromatography system employed with an SPD-M20A diode array detector (Shimadzu Europa GmbH in Duisburg, Germany) was used to quantify individual polyphenols. The latter was separated through a Phenomenex Luna C18(2) column (100 Å, 5 μm, and 4.6 mm × 250 mm) from Torrance (California, USA), in which the temperature was maintained at 40 °C.

2.3. Hemp Leaf Material

For all experiments, the residue of hemp leaves (Cannabis sativa var. Futura 75), excluding the flowering and fruiting tops, was used after an industrial supercritical CO₂ extraction method. The initial raw material was gathered and donated by CBD Extraction I.K.E from Farsala (Thessaly region, Central Greece) at 39°18′22″ N and 22°22′11″ E, based on Google Earth version 9.185.0.0. Hemp residue was grounded to fine powder with an average diameter of <400 μm. The material was finally stored at −40 °C until further analysis.

2.4. Hemp Leaf Extraction

To explore the optimal conditions and isolate bioactive compounds from hemp residue, a conventional extraction process was conducted using a constant liquid-to-solid ratio (i.e., 20:1 mL/g) in several hydroethanolic mixtures (0, 50, and 100% v/v ethanol in water) at temperatures of 20–80 °C and for a total duration of 30–90 min. More details can be found in

Table 1. Experimental findings for the three independent variables investigated and the responses of the dependent variables to the pressurized liquid extraction (PLE) technique.

Design Point	Independent Variables			Responses
				TPC (mg GAE/g dw)		FRAP (μmol AAE/g dw)
	X1 (C, %)	X2 (T, °C)	X3 (t, min)	Actual	Predicted	Actual	Predicted
1	0 (50)	0 (100)	0 (15)	12.18 ± 0.57	12.05	57.36 ± 1.38	59.39
2	−1 (0)	−1 (40)	0 (15)	7.84 ± 0.58	7.95	44.84 ± 1.08	52.36
3	0 (50)	−1 (40)	1 (25)	12.10 ± 0.64	10.76	64.19 ± 4.69	54.41
4	0 (50)	−1 (40)	−1 (5)	11.63 ± 0.35	10.76	64.91 ± 4.74	54.41
5	1 (100)	1 (160)	0 (15)	7.76 ± 0.45	5.92	29.72 ± 0.92	21.41
6	1 (100)	0 (100)	1 (25)	4.33 ± 0.24	4.63	15.82 ± 0.44	16.42
7	1 (100)	−1 (40)	0 (15)	2.16 ± 0.12	3.34	6.73 ± 0.42	11.44
8	0 (50)	0 (100)	0 (15)	12.09 ± 0.64	12.05	56.43 ± 2.77	59.39
9	1 (100)	0 (100)	−1 (5)	4.27 ± 0.27	4.63	13.41 ± 0.50	16.42
10	0 (50)	1 (160)	−1 (5)	13.24 ± 0.74	13.34	61.94 ± 3.72	64.38
11	−1 (0)	0 (100)	−1 (5)	8.57 ± 0.50	9.24	55.70 ± 3.01	57.34
12	0 (50)	1 (160)	1 (25)	11.12 ± 0.32	13.34	53.49 ± 3.32	64.38
13	0 (50)	0 (100)	0 (15)	11.98 ± 0.53	12.05	57.44 ± 1.26	59.39
14	−1 (0)	0 (100)	1 (25)	8.64 ± 0.54	9.24	53.42 ± 3.10	57.34
15	−1 (0)	1 (160)	0 (15)	11.92 ± 0.24	10.53	75.40 ± 3.47	62.33

Values represent the mean of triplicate determinations ± standard deviation.

and

Table 2. Experimental findings for the three independent variables investigated and the responses of the dependent variables to the stirring extraction (STE) technique.

Design Point	Independent Variables			Responses
				TPC (mg GAE/g dw)		FRAP (μmol AAE/g dw)
	X1 (C, %)	X2 (T, °C)	X3 (t, min)	Actual	Predicted	Actual	Predicted
1	0 (50)	0 (50)	0 (60)	19.50 ± 0.88	19.54	92.56 ± 2.04	92.19
2	−1 (0)	−1 (20)	0 (60)	12.44 ± 0.29	11.98	83.78 ± 3.27	85.91
3	0 (50)	−1 (20)	1 (90)	18.13 ± 1.21	18.51	97.80 ± 3.42	93.02
4	0 (50)	−1 (20)	−1 (30)	16.81 ± 0.50	17.41	98.74 ± 6.32	97.25
5	1 (100)	1 (80)	0 (60)	7.34 ± 0.40	7.74	40.90 ± 0.94	38.29
6	1 (100)	0 (50)	1 (90)	7.25 ± 0.41	7.49	20.57 ± 0.84	20.51
7	1 (100)	−1 (20)	0 (60)	8.40 ± 0.32	7.75	11.87 ± 0.88	16.00
8	0 (50)	0 (50)	0 (60)	19.46 ± 1.36	19.54	92.34 ± 5.26	92.19
9	1 (100)	0 (50)	−1 (30)	6.14 ± 0.24	6.16	26.20 ± 0.86	24.74
10	0 (50)	1 (80)	−1 (30)	19.90 ± 0.54	19.46	96.73 ± 5.03	100.41
11	−1 (0)	0 (50)	−1 (30)	12.85 ± 0.64	12.68	75.29 ± 2.86	75.51
12	0 (50)	1 (80)	1 (90)	19.78 ± 0.45	19.11	92.65 ± 3.24	96.18
13	0 (50)	0 (50)	0 (60)	19.53 ± 1.11	19.54	92.63 ± 2.96	92.19
14	−1 (0)	0 (50)	1 (90)	12.05 ± 0.54	12.10	69.02 ± 4.00	71.28
15	−1 (0)	1 (80)	0 (60)	14.06 ± 0.35	14.64	74.54 ± 3.06	69.93

Values represent the mean of triplicate determinations ± standard deviation.

. The extraction optimization was conducted using 50 mL screw-capped glass bottles and a stirring hotplate at 500 rpm, examining a single factor each time. After extraction, the samples were centrifuged at 10,000× g for 10 min to isolate the supernatant, whereas the solid residue was discarded. The dried mass was also extracted by means of a PLE apparatus as an alternative and sustainable extraction method. The pressure of the PLE apparatus was kept constant at 1700 psi. The liquid-to-solid ratio was kept constant at 40:2 mL/g, as was the case in STE. Extracts from PLE did not require centrifugation. Finally, supernatants of both extraction techniques were kept at −40 °C. All experiments using conventional stirring or PLE with their respective combinations are listed in Table 1 and Table 2, where the coded levels are also shown.

2.5. Optimization with Response Surface Methodology (RSM) and Experimental Design

Response Surface Methodology (RSM) through a Box–Behnken design with three factors and three levels was utilized to determine the optimal extraction conditions for total polyphenol content (TPC) and ferric reducing antioxidant power (FRAP). The Box–Behnken method is widely used to optimize extraction parameters with a reduced number of experiments as a means of minimizing processing variation and maintaining high-quality results [15]. This was applied to two extraction techniques: (I) pressurized liquid extraction (PLE) and (II) stirring extraction (STE) from C. sativa by-products post-supercritical CO₂ extraction. The independent variables examined were ethanol concentration in water (C, 0–100% v/v) as X₁, temperature (T, 40–160 °C for PLE and 20–80 °C for STE) as X₂, and extraction time (t, 5–25 min for PLE and 30–90 min for STE) as X₃. These variables were coded at three levels: low (−1), medium (0), and high (+1). To assess the method’s repeatability, fifteen experimental runs were designed, each including three center points. These runs were conducted in triplicate, and the average response values were recorded.

Stepwise regression was chosen to enhance the predictive accuracy of the model by minimizing variance due to unnecessary term estimation. The subsequent second-order polynomial equation modeled the interrelations among the three independent variables as follows:

$$Y_k={\beta }_0+\sum _{i=1}^2{\beta }_i{X}_i+\sum _{i=1}^2{\beta }_{\mathrm{ii}}{X}_i^2+\sum _{i=1}^2\sum _{j=i+1}^3{\beta }_{\mathrm{ij}}{X}_i{X}_j$$

(1)

where the independent variables are denoted by X_i and X_j, and the predicted response variable is defined by Y_k. In the model, the intercept and regression coefficients β₀, β_i, β_ii, and β_ij represent the linear, quadratic, and interaction terms, respectively.

2.6. Determination of Bioactive Compounds

2.6.1. Spectrophotometric Evaluation of Total Polyphenolic Content (TPC)

TPC was evaluated using a previous Folin–Ciocalteu methodology [16]. A gallic acid calibration curve (10–100 mg/L of gallic acid, R² = 0.9996) in methanol was used in a photometric assay at 740 nm. Total polyphenol yield results were expressed as mg gallic acid equivalents (GAE) per g of dry weight (dw).

2.6.2. Spectrophotometric Evaluation of Total Flavonoid Content (TFC)

TFC was employed to determine a specific subgroup of polyphenols and was evaluated using a previously established methodology [17]. A quercetin 3-O-rutinoside, also known as rutin, calibration curve (30–300 mg/L in methanol, R² = 0.9966) was employed to quantify TFC at 415 nm, whereas the results were calculated as mg rutin equivalents (RtE) per g of dry weight (dw).

2.6.3. Chromatographic Measurement of Individual Phenolic Compounds

High-Performance Liquid Chromatography (HPLC) was utilized to detect and quantify individual phenolic compounds in sample extracts. The analysis of C. sativa extracts was conducted using a Shimadzu CBM-20A liquid chromatograph and a Shimadzu SPD-M20A diode array detector (DAD), both acquired from Shimadzu Europa GmbH (Duisburg, Germany). Separation of the compounds occurred in a Phenomenex Luna C18(2) column, sourced from Phenomenex Inc. (Torrance, CA, USA) and maintained at 40 °C (100 Å, 5 μm, 4.6 mm × 250 mm). The mobile phase consisted of 0.5% aqueous formic acid (A) and 0.5% formic acid in acetonitrile (B). The gradient program for non-flavonoid identification started at 5% and increased to 12% B over 15 min, then to 55% B over 35 min, reaching 100% B in 1 min, holding steady for 3 min, and finally returning to 5% B over 40 min. For flavonoid identification, the program began at 5% and rose to 30% B in 3 min, then to 68% B over 34 min, peaked at 100% B in 1 min, remained constant for 3 min, and reverted to 5% B over 40 min. The mobile phase flow rate was maintained at 1 mL/min. Compounds were identified by matching their absorbance spectra and retention time to those of known standards and quantified using calibration curves ranging from 0 to 500 μg/mL.

2.6.4. Evaluation of Ascorbic Acid Content (AAC)

AAC was determined photometrically in each extract at 760 nm using 10% (v/v) aqueous Folin–Ciocalteu reagent and trichloroacetic acid, as described in our previous study [16]. A calibration curve was required for the determination (50–500 mg/L of ascorbic acid in 10% trichloroacetic acid aqueous solution, R² = 0.9980).

2.7. Evaluation of Antioxidant Activity of the Extracts

2.7.1. Ferric Reducing Antioxidant Power (FRAP) Assay

A typical electron-transfer method (i.e., FRAP) in which iron oxidation state reduction from +3 to +2 was detected at 620 nm was employed to determine the antioxidant capacity of C. sativa residue extracts, the methodology of which is comprehensively described in our previous study [17]. A calibration curve of ascorbic acid (50–500 μM in 0.05 M HCl, R² = 0.9997) was utilized, and the results were expressed as μmol of ascorbic acid equivalents (AAE) per gram of dw.

2.7.2. DPPH Scavenging Activity Assay

A slightly modified method from Shehata et al. [18] was used to further evaluate the antioxidant activity of C. sativa residue extracts. A photometric method of scavenging the synthetic DPPH^• probe at 515 nm was employed for that reason. In summary, 50 μL of a suitably diluted sample was combined with 1950 μL of a 100 μM methanolic DPPH^• solution. That mixture was allowed to stand for 30 min at ambient temperature in the absence of light, and the final absorbance (i.e., A_515(f)) was measured at the same wavelength. The % scavenging of the radical after a 30 min exposure to antioxidants from the extract was calculated using a calibration curve (100–1000 μM of ascorbic acid concentration, C_AA, in methanol, R² = 0.9926), as indicated in Equation (2). The results from antiradical activity (A_AR) were expressed as μmol AAE per g of dw, as shown in Equation (3). A blank solution including 100 μM of methanolic DPPH and methanol instead of extract was used to determine initial absorbance (A_515(i)). The extraction volume (V, in L) and solid weight (w, in g) were also taken into account.

$$\mathrm{Scavenging} \left(\%\right)={\displaystyle \frac{A_{515\left(\mathrm{i}\right) }-A_{515\left(\mathrm{f}\right)}}{A_{515\left(\mathrm{i}\right)}}} \times 100$$

(2)

$$A_{\mathrm{AR}}\left(\mathsf{\mu }\mathrm{mol} \mathrm{AAE}/\mathrm{g} \mathrm{dw}\right)={\displaystyle \frac{C_{\mathrm{AA}} \times V}{w}}$$

(3)

2.8. Color Evaluation of the Extracts

A colorimeter was employed to determine the CIELAB color coordinates of C. sativa residue extracts, as reported earlier [19]. The low and high values of L*, a*, and b* coordinates were required to measure the lightness–darkness, redness–greenness, and yellowness–blueness of the extracts, respectively.

2.9. Statistical Analysis

JMP^® Pro 16 software (SAS, Cary, NC, USA) was utilized for statistical analysis in Response Surface Methodology (RSM) and distribution analysis. Each set of C. sativa residue extracts underwent the extraction process at least twice, and quantitative analyses were performed in triplicate. The Kolmogorov–Smirnov test assessed the normality of the data. Significant differences were evaluated using one-way analysis of variance (ANOVA) with the Tukey HSD multiple comparison test. Results were expressed as averages with measures of variability. Additionally, JMP^® Pro 16 facilitated the statistical evaluation of multiple factor analysis (MFA), multivariate correlation analysis (MCA), and partial least squares (PLS).

3. Results and Discussion

3.1. Extraction Parameter Optimization

It is imperative to optimize the extraction of bioactive compounds from C. sativa residues using PLE in order to attain the highest possible yield of bioactive compounds. It is important to mention that carotenoid assays were also conducted on all of the extracts; however, the results indicated that none of them contained carotenoids. This can be attributed to the fact that the C. sativa residues were previously treated with supercritical CO₂, which resulted in the removal of any carotenoids they may have contained. Supercritical CO₂ is a crucial nonpolar solvent; therefore, it allows for the extraction of a diverse array of lipophilic compounds, such as carotenoids [20]. The extraction procedure may be complex due to the presence of several bioactive compounds, resulting in variances in solubility and polarity [21]. Furthermore, the extraction technique and other processing parameters highly impact the extract yield and antioxidant capacity. Recently, there has been notable progress in extraction technologies that diminish dependence on hazardous and toxic solvents, safeguard human health, and necessitate minimal energy consumption. Incorporating an eco-friendly solvent is crucial for the successful application of this technique [22]. Water is an easily obtainable and sustainable solvent owing to its remarkable ability to extract polar molecules, affordability, and non-toxicity to people. Moreover, organic solvents are also frequently utilized to improve the extraction process. Ethanol may be mixed with water to formulate an extraction solvent appropriate for application in the food sector [23]. Taking all of these into account, various parameters, such as solvent composition, extraction temperature, and duration, were tested.

Table 1 presents the values of PLE parameters along with the predicted TPC and FRAP values, whereas Table 2 shows the corresponding predicted TPC and FRAP values of the STE procedure. The results showed that a wide range of TPC values was observed (from 2.16–13.24 mg GAE/g dw), indicating that the examined extraction parameters had a vast impact on the recovery of polyphenolic compounds. It was revealed that design point 7 had the lowest value, whereas design point 10 had the highest value. A similar trend applied in FRAP values, which ranged from 6.73–75.40 μmol AAE/g dw and revealed an eleven-fold difference. Design point 7 was again found to be the weakest; however, design point 15 showed the highest antioxidant activity. At first glance, it seems that PLE at high temperatures was preferable. In addition, high concentrations of ethanol in the binary mixture were deemed to have detrimental effects on both TPC and FRAP. Finally, extraction duration did not seem to highly affect the recovery of targeted compounds. Regarding the results from the STE process, they were found to be slightly higher than those obtained from PLE in both assays. For instance, TPC values ranged from 6.14 to 19.90 mg GAE/g dw. It was observed that the three centroid design points (i.e., 1, 8, and 13) had similar values, which were also the highest. Moreover, it was again highlighted that moderate polarity seemed to be preferable for the efficient recovery of polyphenols, whereas extraction at ambient temperatures had a diminishing effect on the extraction of targeted molecules. Regarding FRAP values, the results showed that the range from 11.87 to 98.74 μmol AAE/g dw had a ~9-fold difference. Finally,

Table 3. Analysis of variance (ANOVA) for the response surface quadratic polynomial model for pressurized liquid extraction (PLE) and stirring extraction (STE) techniques.

Factor	PLE		STE
	TPC	FRAP	TPC	FRAP
Stepwise regression coefficients
Intercept	12.049 *	59.394 *	19.542 *	92.192 *
X1—solvent concentration	−2.306 *	−20.460 *	−2.784 *	−25.390 *
X2—temperature	1.2888 *	4.985	0.6625 *	1.5788
X3—extraction time	-	-	0.1888	−2.115
X12	−5.112 *	−22.510 *	−9.015 *	−44.180 *
X22	-	-	-	4.526 *
X32	-	-	−0.920 *	-
X1X2	-	-	−0.670	9.5675 *
X1X3	-	-	0.4775	-
X2X3	-	-	−0.360	-
ANOVA
F-value (model)	36.85	28.77	117.19	160.62
F-value (lack of fit)	6.49	18.57	481.01	788.03
p-Value (model)	<0.0001	<0.0001	<0.0001	<0.0001
p-Value (lack of fit)	0.0207	0.0014	0.0021	0.0013
R2	0.9095	0.8870	0.9936	0.9918
Adjusted R2	0.8848	0.8561	0.9852	0.9856
RMSE	1.1780	7.9394	0.6292	3.6797
PRESS	33.046	1504.01	25.015	712.4
CV	37.235	44.171	36.268	43.153
DF (total)	14	14	14	14

* The values significantly affected responses at a probability level of 95% (p < 0.05). TPC, total polyphenol content; FRAP, ferric reducing antioxidant power; F-value, test for comparing model variance with residual (error) variance; p-Value, probability of seeing the observed F-value if the null hypothesis is true; RMSE, root mean square error; PRESS, predicted residual error sum of squares; CV, coefficient of variation; DF, degrees of freedom.

provides the statistical details of the stepwise regression analysis of ANOVA for the TPC and FRAP techniques on PLE and STE, respectively. Variables that do not have a statistically significant effect (p > 0.05) have been removed from the equation. Moreover, the derived models fit the data adequately; however, it is distinguished that R² was better for the STE technique (>0.98).

3.2. Model Analysis

The results of the data analyzed from the surface-response model resulted in the generation of simplified polynomial equations, as variables that do not have a statistically significant contribution (p > 0.05) to optimizing the recovery of targeted bioactive compounds are missing. For this reason, Equations (4) and (5) below refer to the modeling for the PLE technique, while Equations (6) and (7) address the STE technique.

TPC = 7.095 + 0.158X₁ + 0.021X₂ − 0.002X₁²

(4)

FRAP = 49.032 + 0.491X₁ + 0.083X₂ − 0.009X₁²

(5)

TPC = 6.79 + 0.308X₁ + 0.068X₂ + 0.133X₃ − 0.004X₁² − 0.001X₃² − 0.0004X₁X₂ + 0.0003X₁X₃ − 0.0004X₂X₃

(6)

FRAP = 103.51 + 0.941X₁ − 0.769X₂ − 0.071X₃ − 0.018X₁² + 0.005X₂² + 0.006X₁X₂

(7)

The optimization of the extraction of bioactive compounds evaluated by TPC and FRAP methods through the specific parameters X₁, X₂, and X₃ is also possible by 3D graphical plots, as can be seen in Figure 1 and Figure 2. The interpretation of all 3D plots is based on the color depiction of the impact of the extraction parameters on each assay. In each experiment, it is preferable to employ the parameter conditions that yield maximum values (red color). Experimental conditions that produce blueish or greenish color in each assay are ineffective, as they yield low values. Figure 1 shows how the specific assays (i.e., TPC in Figure 1A and FRAP in Figure 1B) are affected by the parameters X₁ and X₂ in the PLE extraction technique. For example, in Figure 1A, one can observe that for TPC optimization, values between 20 and 40 were demanded for parameter X₁ (i.e., % v/v of ethanol in water), while for the corresponding X₂ (i.e., temperature), values higher than 150 were required, in which the red color is shown in the graph. The same rationale applies to Figure 1B and Figure 2A–F.

3.3. Impact of Extraction Parameters on Assays Through Pareto Plot Analysis

A normalized Pareto plot was used, with a statistical significance level of p < 0.05, to evaluate the impacts and interactions to STE and PLE procedures resulting from each extraction parameter (i.e., X₁, % solvent concentration; X₂, extraction duration; X₃, extraction temperature). Figure 3 displays the correlation between total polyphenols and antioxidant activity through the FRAP assay and the independent variables. Red color depicts a negative impact of a parameter, whereas blue color indicates a positive impact. Also included are the orthogonal coded estimates that were derived from the estimations utilizing the orthogonality transformation. The t-ratio on a Pareto plot was determined by dividing the parameter estimate by its standard error. This ratio is crucial for assessing the significance of each parameter estimate. Additionally, the cumulative line on the plot illustrates the total of the absolute t-ratios, providing insight into the relative explanatory power of each estimate. To that end, pink asterisks were inserted to declare the statistical significance (p < 0.05) of each parameter.

Based on the results of the Pareto plot, we can deduce that an X1 negatively affected TPC and FRAP when it came to the extraction parameters, as was previously noticed. This outcome means that increasing ethanol concentration in the binary mixture of ethanol–water was not preferable. Both extraction methods suited this trend as well. Regarding the X₂ parameter, it was revealed that an increasing temperature in each extraction technique was found to have an appreciable impact on the extraction of targeted compounds, particularly when PLE is carried out within a cell with the absence of light and without oxygen for a shorter duration than the lengthy conventional extraction process (i.e., STE). Nevertheless, this influence was not as substantial as the solvent composition of the binary mixture, suggesting the critical importance of solvent polarity. Finally, increased extraction duration was detrimental in STE, as indicated by the earlier discussion. Conversely, it appears to have no impact on the PLE technique.

3.4. Partial Least Squares (PLS) Analysis

The PLS model was used to estimate the impact of the extraction condition parameters (X₁, X₂, and X₃). Figure 4 shows an illustrative correlation loading plot. The extraction of bioactive compounds is greatly affected by several factors, the most significant of which are temperature, solvent composition, and extraction duration [24]. First of all, the extraction process may be complicated by the polyphenols’ varying solubility and polarity [21,25]. Regarding the PLE technique, it can be observed that the X₁ parameter had the most statistically significant (p < 0.05) impact among other parameters in the extraction process, as indicated by the Variance Importance Plot (VIP) in Figure 4C. In the same Figure 4A, the previous observations from the 3D models of the response surface were confirmed, as the optimum concentration was 32% v/v, and the optimum temperature was 160 °C. Regarding the extraction duration, it seemed that it had no effect on the extraction; therefore, the lowest possible was preferred since the specific technique is renowned for its rapidity and sustainability. Concerning the STE technique, it was found that an intermediate polarity solvent was optimal for more efficient recovery of polyphenols while obtaining extracts of hemp by-products with high antioxidant activity, as also verified by the Variance Importance Plot. Temperature and extraction duration had a minor impact on the extraction process; however, high temperature and short extraction time were preferred. For the PLE technique, substantial weakening of the solute–matrix connection due to van der Waals forces or hydrogen bonds can occur under conditions of high heat and pressure. This reduces solvent viscosity, increases solute molecular extraction efficiency, and minimizes energy requirements. This allows the solvent to more easily diffuse into the sample by lowering its resistance to the matrix [26]. The model provided a low extraction duration, as previous studies have demonstrated the efficacy of both brief [27] and extended [28] periods. Even though elevated temperature enhances the extraction of bioactive compounds by increasing their solubility in STE, numerous thermolabile compounds may undergo degradation at such elevated temperatures [29].

Experimental results and PLS model predictions show remarkable agreement, demonstrated by high correlation coefficients of 0.9948 and 0.9976 and significant coefficients of determination (R²) of 0.9897 and 0.9953 for PLE and STE techniques, respectively. Low p-values (<0.0001) for both PLE and STE suggest that the differences between the actual and predicted values are not statistically significant.

The desirability function was employed to determine the maximum predicted values for bioactive molecules and antioxidant assessment. The predicted responses were ideal when various extraction procedures and circumstances were employed for each experiment. For example, TPC necessitated only 5 min at 160 °C when utilizing the PLE technique, while STE demanded 45 min at 80 °C. Both extraction techniques demanded similar conditions, including high temperatures, low duration, and similar solvent polarity. It could be deduced that despite the PLE technique having significantly lower possible extracted polyphenols, it demanded only 5 min, indicating the possibility of several extraction cycles in a short time.

Table 4. The partial least squares (PLS) prediction profiler identified the conditions for maximum desirability across all variables under each optimal condition for pressurized liquid extraction (PLE) and stirring extraction (STE) techniques.

Technique	Independent Variables			PLS Model Values
	X1 (C, %)	X2 (T, °C)	X3 (t, min)	TPC (mg GAE/g dw)	FRAP (μmol AAE/g dw)
PLE	32	160	5	13.51	68.83
STE	40	80	45	20.44	100.75

contains comprehensive information regarding maximum predicted responses and optimal extraction conditions for both assays.

3.5. Analysis of Optimized Extracts

3.5.1. Extensive Bioactive Compound and Antioxidant Capacity Determination

PLE is recognized for its rapid extraction from plant material, categorizing it as an environmentally friendly and sustainable extraction method [24]. The authors deemed it suitable to employ three cycles of the PLE approach, highlighting the importance of rapid extraction, whilst a single STE cycle was utilized primarily for comparison purposes. The optimum extracts were thoroughly examined and analyzed in terms of their bioactive constituents and antioxidant efficacy. TFC and AAC were included in the investigation of bioactive compounds. To assess antioxidant activity, a free radical scavenging assay (i.e., DPPH) was integrated with the evaluation of the extract’s reducing power. These investigations increase understanding of the advantages of an extract obtained from a hemp by-product.

Table 5. Bioactive compounds and antioxidant activities were assessed under optimal conditions using pressurized liquid extraction (PLE), which involved three cycles of extracting by-products from C. sativa. The study compared the PLE and stirring extraction (STE) techniques.

Technique	TPC 1	FRAP 2	DPPH 3	TFC 4	AAC 5
1st cycle PLE	15.41 ± 0.52 c	79.20 ± 5.62 b	61.47 ± 1.23 c	9.68 ± 0.49 c	1.42 ± 0.09 a
2nd cycle PLE	3.65 ± 0.22 d	11.89 ± 0.36 c	11.98 ± 0.52 d	3.13 ± 0.11 d	n.d.
3rd cycle PLE	1.36 ± 0.03 e	4.39 ± 0.28 c	2.38 ± 0.11 e	0.64 ± 0.04 e	n.d.
PLE (SUM cycles)	20.42 ± 0.74 a	95.47 ± 2.10 a	75.83 ± 1.52 b	13.45 ± 0.43 a	1.42 ± 0.09 a
STE	18.81 ± 0.77 b	100.96 ± 4.85 a	85.64 ± 5.31 a	11.2 ± 0.53 b	0.62 ± 0.02 b

Statistical significance (p < 0.05) is denoted by different lowercase letters (e.g., a–e) within each column; n.d. means not detected. ¹ Values in mg GAE/g dw; ² values in μmol AAE/g dw; ³ values in μmol AAE/g dw; ⁴ values in mg RtE/g dw; ⁵ values in mg AA/g.

provides further information about the yielded bioactive compounds, along with an assessment of antioxidant capacity using two different techniques.

The results indicated that the extraction of hemp by-products could generate high-value products with plenty of bioactive compounds. Notably, TPC showed high levels (~20 mg GAE/g dw), comparable to those observed in our prior research [6], even though we processed hemp leaves (without flowering and fruiting tops) of the same variety (Futura 75). In another investigation by Ferrante et al. [30] utilizing inflorescences of the same cultivar, the dried material was extracted using water solvent and ultrasound-assisted extraction, resulting in a TPC of 8.1 mg GAE/g dw. Mazzara et al. [31] reported that the TPC of eleven determinate biomasses from commercial hemp varieties ranged from 9.54 to 23.90 mg/g dw. By-products from hemp oil were also examined in a study by Mourtzinos et al. [7]. The authors yielded 0.94–4.51 mg GAE/g in an extraction optimization process. Regarding TFC, the results suggested that these bioactive molecules account for over 50% of the total polyphenols. In the prior study by Ferrante et al. [30], TFC was quantified at 6.0 mg RtE/g dw, which is half of our results. The previously reported study by Mazzara et al. [31] identified an expansive range, with commercial hemp types exhibiting a range of 3.25–22.00 mg RtE/g dw. The PLE technique was found to yield significantly higher AAC than STE in a single cycle. Finally, antioxidant capacity assays revealed that the STE technique was slightly more preferable in terms of DPPH assay, wherein 85.64 against 75.83 μmol AAE/g dw were measured. Statistically non-significant (p > 0.05) results were obtained in a FRAP assay from both techniques.

3.5.2. Color Analysis

Pigments can make an extract more aesthetically pleasing, which increases its demand by consumers [32]. The pigmentation process of the hemp leaves could be a matter of carotenoids and chlorophylls; however, preliminary experiments revealed the absence of carotenoids, probably due to supercritical CO₂ extraction. The extraction of hemp leaf residue resulted in the generation of brownish extracts—the color with several hues is illustrated in Figure 5. The values of each of the L*, a*, and b* coordinates are also visible in the graph. It is obvious that the first PLE cycles result in darker extracts, and after each cycle, the color becomes lighter. That is a trend that explains the ~20-unit difference between the STE and the third PLE cycle. The low a* values illustrate the greenish hue, whereas higher b* values underline the yellow color of the extracts. Babiker et al. [33] investigated roasting effects on hemp seed color, in which all color coordinates were found to be slightly darker than our extracts (i.e., L* ~45, a* ~3, and b* ~15).

3.5.3. Individual Polyphenol Quantification

The authors quantified 15 individual polyphenols in the hemp by-product, classifying them into flavonoids and non-flavonoids to enhance the understanding of its bioactive molecules. A representative chromatogram with all these compounds is depicted in Figure 6. Additional details regarding these compounds are available in

Table 6. Optimal extraction conditions for phenolic compounds using pressurized liquid extraction (PLE) involve three cycles of C. sativa by-product extracts. The study compared the PLE and stirring extraction (STE) techniques.

Phenolic Compounds (mg/g)	1st Cycle PLE	2nd Cycle PLE	3rd Cycle PLE	PLE (SUM Cycles)	STE
Non-Flavonoids
Catechol	0.21 ± 0.01 a	n.d.	n.d.	0.21 ± 0.01 a	n.d.
Pyrogallol	0.23 ± 0.01 c	0.19 ± 0.01 c	0.10 ± 0.00 c	0.52 ± 0.02 b	2.44 ± 0.16 a
Phenol	0.32 ± 0.01 c	0.08 ± 0.00 d	0.06 ± 0.00 d	0.47 ± 0.03 b	0.75 ± 0.03 a
Pyrocatechuic acid	0.23 ± 0.01 b	0.17 ± 0.01 c	0.08 ± 0.00 d	0.53 ± 0.02 a	n.d.
Caffeic acid	0.76 ± 0.03 b	0.17 ± 0.01 c	0.07 ± 0.00 c	0.89 ± 0.03 b	2.28 ± 0.11 a
Homovanilinic acid	0.44 ± 0.01 c	0.06 ± 0.00 d	n.d.	0.57 ± 0.04 b	1.09 ± 0.06 a
Syringic acid	1.36 ± 0.09 b	0.27 ± 0.01 c	0.08 ± 0.00 c	1.60 ± 0.10 b	2.99 ± 0.22 a
p-Coumaric acid	0.29 ± 0.01 b	n.d.	n.d.	0.39 ± 0.01 a	n.d.
Ferulic acid	0.35 ± 0.02 b	0.09 ± 0.00 c	n.d.	0.53 ± 0.03 a	n.d.
SUM Non-Flavonoids	4.19 ± 0.24 c	1.03 ± 0.04 d	0.38 ± 0.02 d	5.72 ± 0.29 b	9.55 ± 0.57 a
Flavonoids
Cyanidin-3-glucoside chloride	0.06 ± 0.00 b	0.03 ± 0.00 b	0.01 ± 0.00 b	0.09 ± 0.01 b	1.58 ± 0.07 a
Rutin	3.40 ± 0.23 b	2.50 ± 0.14 c	0.19 ± 0.00 e	6.16 ± 0.19 a	1.99 ± 0.13 d
Luteolin-7-glucoside	3.80 ± 0.08 b	1.16 ± 0.08 d	0.24 ± 0.02 e	5.28 ± 0.21 a	2.65 ± 0.16 c
Apigenin-7-O-glucoside	0.73 ± 0.03 b	0.18 ± 0.01 c	0.01 ± 0.00 d	0.87 ± 0.04 a	0.76 ± 0.04 b
Fisetin	0.36 ± 0.03 c	0.15 ± 0.01 d	0.06 ± 0.00 d	0.58 ± 0.03 b	1.89 ± 0.11 a
Quercetin	0.37 ± 0.01 c	0.18 ± 0.00 d	0.08 ± 0.00 d	0.64 ± 0.02 b	2.43 ± 0.09 a
SUM Flavonoids	8.72 ± 0.43 c	4.20 ± 0.24 d	0.58 ± 0.03 e	13.62 ± 0.50 a	11.29 ± 0.61 b
Total identified	12.91 ± 0.67 b	5.22 ± 0.28 c	0.96 ± 0.05 d	19.34 ± 0.79 a	20.84 ± 1.18 a

Statistical significance (p < 0.05) is denoted by different lowercase letters (e.g., a–e) within each row; n.d. means not detected.

. The concentration of these polyphenols showed no statistically significant difference (p > 0.05) between the two extraction procedures (i.e., when three cycles of PLE were conducted), with ~20 mg/g dw determined. Nonetheless, a noteworthy observation was the variation in the recovery of the two classes of polyphenols based on the extraction method employed. For instance, in the PLE technique, a two-fold difference was observed in the concentration of flavonoids compared to non-flavonoids (13.62 against 5.72 mg/g, respectively). This finding could trigger interest in examining the extraction mechanism in each technique in a future study.

Some of the individual polyphenols that were quantified in our previous study [6], in which we utilized leaves from the Futura 75 cultivar, were also quantified in the hemp residue. For instance, rutin (0.45 mg/g dw) and apigenin-7-O-glucoside (0.51 mg/g dw) were quantified; however, these molecules were yielded in higher amounts in hemp residue. Syringic acid (~0.05 mg/g dw), ferulic acid (~0.007 mg/g), and other similar polyphenols were identified by Babiker et al. [33]; however, significantly lower yields were obtained since the authors investigated hemp seed extracts. The same trend applies to Izzo et al. [34]; in their study, several similar polyphenols from inflorescences from commercial hemp cultivars (i.e., Kompolti, Tiborszallasi, Antal, and Carmagnola) ranged from 0.37 to 0.74 mg/g dw.

3.6. Correlation Analyses in Optimized Extracts

3.6.1. Multiple Factor Analysis (MFA)

A consensus map graph was also used for better data processing and interpretation from the samples and variables, as illustrated in Figure 7. The consensus map applies the average correlation between samples and variables under examination. All of the aforementioned variables are graphically shown in this plot as lines, whereas samples, including several PLE cycles, are depicted as dots. The degree to which various parameters were consistent or varied between the samples might also be shown. The results were analyzed using the concept of “inertia” values, which is a metric for measuring the degree of consistency across different clusters, widely known as the within-cluster sum of squares. Data points inside a cluster with a high inertia value are quite different from each other, according to El Khediri et al. [35]. This graph easily enables the classification of samples and evaluation of all variables’ interactions. As an example, a small grey arrow highlights the proximity of brown lines (which represent non-flavonoid compounds) between the samples’ second and third PLE cycles, as they were previously quantified at 1.03 and 0.38 mg/g, respectively. On the contrary, notable inertia was denoted with a larger grey arrow. It was observed in the third cycle and STE samples in flavonoid concentration (green lines), as these values significantly varied (p < 0.05) at 0.58 and 11.29 mg/g.

The RV coefficient serves as a multivariate extension of the Pearson correlation coefficient, evaluating the linear association between two sets of matrices. Within the scope of

Table 7. RV correlations measure the similarity between two sets of parameters.

Parameters	Bioactive Compounds	Antioxidant Activity	Color	Non-Flavonoids	Flavonoids	Centroid
Bioactive compounds	–	0.929	0.6363	0.8077	0.9716	0.9611
Antioxidant activity		–	0.7262	0.7565	0.8473	0.9381
Color			–	0.6016	0.6295	0.7876
Non-Flavonoids				–	0.8423	0.8889
Flavonoids					–	0.9511
Centroid						–

, a centroid denotes the central location within a data point cluster, signifying the mean position of all the points within that cluster. In MFA, the centroid aids in discerning the central tendency of data points associated with a specific group. An anticipated high correlation was observed between bioactive compounds and antioxidant activity (0.929). In addition, an interesting finding was that flavonoids contributed more to antioxidant activity than non-flavonoid compounds since they had a higher RV correlation (i.e., 0.8473 compared to 0.7565).

The Block Partial Contributions plot is an essential tool in our study as it assists in identifying which blocks, such as bioactive compounds, antioxidant activity, color, non-flavonoids, and flavonoids, contribute most to the variance explained by the principal dimensions (Dim1 and Dim2). Figure 8 facilitates the interpretation of the MFA results by illustrating the contribution of each block to the principal components. It is determined by multiplying the eigenvalue by 100 and then dividing it by each set of the contribution of a variable to each dimension. This makes it easier to examine how each block fits into the bigger picture of the data structure. The most prominent trend, represented by Dim 1, is responsible for 68% of the overall dataset variance. In contrast, the second most significant trend, which is orthogonal to the first, is described by Dim 2 and accounts for 19.6% of the variance. Variable loadings illustrate the correlations between the variables and the components, while the plot displays the contribution of each original parameter to the principal component. The relative importance of the data points is shown by the dots of varying sizes. The size of the dots indicates the relative importance of the values; larger dots indicate more significance and lower ones less significance. For instance, it was previously mentioned that flavonoid concentration was higher than non-flavonoids, which may describe the difference in their size. Color variables had little effect on the overall value of the extracts, which explains the small dot size.

3.6.2. Multivariate Correlation Analysis (MCA)

MCA was conducted to elucidate the correlation between the measured parameters, including individual polyphenols, all bioactive compounds, color analysis, and antioxidant capacity assays, as shown in the plot in Figure 9. The correlation between all these parameters could be quantified on a scale from −1 (negative correlation, red color) to 1 (positive correlation, green color), as shown in the heat map. To start with, most individual polyphenols had a positive correlation with each other and with antioxidant capacity, as previously stated. Rutin did not correlate at all with parameters such as pyrogallol, fisetin, quercetin, homovanilinic acid, or a*. An interesting finding could be the negative correlation of AA with polyphenols, which could indicate that the specific chemical might be recovered with different extraction conditions than those used for the majority of polyphenols. The same rationale could underlie the lightness of the samples (i.e., L*) and the extraction conditions, as it was observed that the higher the concentration of polyphenols, the darker the color of the extracts. Finally, the negative correlation of catechol with any other polyphenol or antioxidant capacity might occur due to the small amount of data for the specific polyphenol, as it was not quantified in most cases.

4. Conclusions

Since supercritical CO₂ extraction removed several lipid-soluble components from C. sativa leaf residue, this study aimed to seek further valorization of these by-products by applying a sustainable extraction method. To ensure that the final product is suitable for exploitation in a variety of industries, including the food sector, this study was conducted using environmentally friendly and food-grade solvents (i.e., water and ethanol). The results of this study revealed the significant high-added value of extracts from C. sativa residual leaves, as they yielded polyphenol levels comparable to those of the leaf extract material, approximately 20 mg/g dw. In addition, the utilization of PLE was highlighted as a relatively rapid technique, as a single cycle required only 5 min. A higher percentage of flavonoids, specifically luteolin-7-glucoside—which accounted for ~25% of the total polyphenols (19.34 mg/g)—was recovered compared to traditional extraction procedures. Finally, this approach could be utilized as a foundation for additional research involving other green extraction techniques and focused on maximizing the bioactive compounds of various plant material residues.