Optimization of process conditions for ionic liquid-based ultrasound-enzyme-assisted extraction of resveratrol from Polygonum Cuspidatum

Highlights

• •Ionic liquids were efficient in extracting resveratrol from P. Cuspidatum.
• •[C4mim]⁺Br⁻ has proven to be more effective in increasing resveratrol yield from P. Cuspidatum than conventional organic solvents.
• •Extraction of resveratrol from P. Cuspidatum was modeled using RSM and ANN-GA.
• •ANN-GA provided both higher precision and yield in resveratrol extraction from P. Cuspidatum via UEAE than RSM.

Abstract

This work offered a productive technique for resveratrol extraction from Polygonum Cuspidatum (P. Cuspidatum) using ionic liquids in synergy with ultrasound-enzyme-assisted extraction (UEAE). Firstly, ionic liquids with different carbon chains and anions were evaluated. Subsequently, a comprehensive investigation was carried out to evaluate the effect of seven crucial parameters on the resveratrol yield: pH value, enzyme concentration, extraction temperature, extraction time, ultrasonic power, concentration of ionic liquid (IL concentration) and the liquid–solid ratio. Employing the Plackett-Burman Design (PBD), the critical factors were effectively identified. Building upon this foundation, the process was further optimized through the application of Response Surface Methodology (RSM) and an Artificial Neural Network-Genetic Algorithm (ANN-GA). The following criteria were determined to be the ideal extraction conditions: an enzyme concentration of 2.18%, extraction temperature of 58 °C, a liquid–solid ratio of 29 mL/g, pH value of 5.5, extraction time of 30 min, ultrasonic power of 250 W, and extraction solvent of 0.5 mol/L 1-butyl-3-methylimidazolium bromide. Under these conditions, the resveratrol yield was determined to be 2.90 ± 0.15 mg/g. Comparative analysis revealed that the ANN-GA model provided a better fit to the experimental data of resveratrol yield than the RSM model, suggesting superior predictive capabilities of the ANN-GA approach. The introduction of a novel green solvent system in this experiment not only simplifies the extraction process but also enhances safety and feasibility. This research paves the way for innovative approaches to extracting resveratrol from botanical sources, showcasing its significant potential for a wide range of applications.

1. Introduction

Polygonum Cuspidatum (P. Cuspidatum) is a perennial herb in the Polygonaceae family, is notably abundant in resveratrol and polydatin [1]. Resveratrol, a polyphenolic compound derived from plants, is an antitoxin that is secreted in response to environmental stress or pathogen assault. It is ubiquitous across various plant species, with particularly high concentrations observed in medicinal plants such as grapes, P. Cuspidatum, and peanuts. The chemical structure of resveratrol is characterized as (E)-3,5,4′-trihydroxystilbene, with a molecular formula of C₁₄H₁₂O₃ and a relative molecular mass of 228.25. Resveratrol exists in both cis and trans configurations, which can be conjugated with glucose to form resveratrol glycosides. These glycosides are subsequently hydrolyzed by glycosidase in vivo to release resveratrol [2]. Resveratrol has been recognized for its potent antioxidant, neuroprotective, cardiovascular, and hepatoprotective properties [3], [4], underscoring its high health and nutritional benefits. Consequently, the extraction of resveratrol from natural plant sources is of significant importance, with broad applications across the food, health care, and pharmaceutical industries.

The extraction of resveratrol from botanical sources has been the focus of numerous methodologies, each with its unique advantages and limitations. Traditional methods, such as organic solvent reflux extraction [5], are effective but require prolonged periods and substantial quantities of solvents. Conversely, microwave-assisted extraction [6] expedites the process but poses a risk to enzyme integrity due to the high temperatures employed, potentially leading to denaturation and loss of enzymatic function. To mitigate these drawbacks, ultrasound-assisted extraction (UAE) [7] and ultrasound-enzyme-assisted extraction (UEAE) [8] stand out for their ability to address the temperature sensitivity issue. UEAE, in particular, optimizes the extraction process by maintaining an appropriate temperature, allowing cellulase to effectively hydrolyze the cell wall without compromising its activity. This method not only conserves enzyme function but also shortens the extraction duration.

Recent studies have extensively explored the advantages of enzyme-assisted extraction of natural compounds from plants, highlighting its user-friendliness, efficiency, and eco-friendliness [9]. The majority of research in this field employs β-glucosidase pectinase, and cellulase to hydrolyze and break down the components of plant cell walls, thereby facilitating the release of intracellular substances [10]. Ultrasound’s potent thermal effect and the mechanical action of cavitation bubbles contribute to the efficient breakdown of the cell wall, enhancing the solubility of target compounds [11]. The instantaneous collapse of these bubbles releases shock waves that further disintegrate the cell wall structure, facilitating the release of bioactive ingredients. Moreover, ultrasonic extraction offers the added benefits of reduced extraction time, minimized reagent consumption, and lowered operational temperatures [12]. These attributes are particularly advantageous for preserving the structural integrity and biological activity of heat-sensitive compounds, ensuring that the extracted resveratrol retains its therapeutic potency.

Moreover, the process of chemical component extraction from natural flora is significantly influenced by the selection of appropriate extractants. Ionic liquids (ILs) are a class of molten salts that have organic cations and either inorganic or organic anions, and they stay liquid at temperatures lower than 100 °C. The outstanding thermal stability and low saturated vapor pressure of these ILs set them apart, negligible volatility, non-flammability, and their capacity to dissolve a wide spectrum of polar and non-polar substances [13], [14]. Particularly noteworthy are imidazolium-based ionic liquids, which exhibit remarkable solubility for cellulose. During the dissolution phase, the anionic and cationic components of these ILs can interact with cellulose to form intricate complexes. This interaction effectively disrupts the hydrogen bonding network that maintains the structural integrity of the plant cell wall, thereby facilitating the release of the extract into solution and enhancing the overall extraction efficiency [15]. When compared to conventional solvents, imidazolium ionic liquids have been shown in numerous studies to produce better outcomes when used to extract bioactive chemicals from plants. To date, there have been no documented instances of enzymatic extraction of resveratrol from P. Cuspidatum facilitated by ionic liquid ultrasound.

This study introduces a novel approach, employing a “green solvent” ionic liquid solvent system in conjunction using an enzyme-assisted ultrasonic technique to extract resveratrol. Utilizing the Plackett-Burman Design (PBD), the factors that significantly influence the extraction process were screened. Subsequently, the extraction parameters were optimized by RSM and ANN-GA methods. Our objective is to augment the yield of resveratrol extraction from P. Cuspidatum without compromising environmental integrity or incurring excessive costs. We aspire to contribute a novel perspective and valuable insights to the comprehensive utilization of this pharmacologically rich component within the medical and healthcare sectors.

2. Materials and methods

2.1. Experiment materials

The raw material of P. Cuspidatum was purchased from Yulin City, Shanxi Province, China in October 2023 and identified by Associate Professor Xueying Chen of Northeast Agricultural University. The dried drug was crushed into uniform size and passed through a 40-mesh sieve. A voucher specimen (voucher no. 20240401) was deposited in the herbarium of Veterinary Medicine, Northeast Agricultural University for future reference.

Cellulase was purchased from Gaider Chemical Industry, resveratrol with 99.99% purity (Lot: C16195355) was obtained from MackLin Biochemical Technology Co., Ltd. (Shanghai), all ILs used in this test were purchased from Aladdin Reagent Co. and were used as is without modification. Acetonitrile, formic acid and methanol (all chromatography grade reagents) were purchased from Komiou Chemical Reagent Co., Ltd. (Tianjin). Pure water for testing was obtained by filtration using a Milli-Q IQ 7000 water purification system (Sigma Aldrich Shanghai Trading Co., Ltd.). Before loading into the HPLC column, all solutions and samples destined for chromatographic analysis were meticulously filtered through a 0.22 μm nylon membrane to ensure clarity and consistency.

2.2. Experiment instrumentation

The HPLC system (Shimadzu International Trading Co., Ltd.) consists of a SHIMADZU manual sample handling system series. HPLC system equipped with LC-20 AR Bin Pump and SPD-20 A UV Detector (Shimadzu, Japan); chromatographic separation was performed using a Diamonsil C18 reversed-phase column (4.6 mm × 150 mm, 5 µm, Dikma Technologies, Beijing China). The KQ-250DB Ultrasonic Cleaner (Kunshan Ultrasonic Instrument Co., Ltd.) consists of a 23.5 cm by 13.3 cm by 10.2 cm rectangular vessel with a maximum power of 250 W, equipped with an annealed transducer operating at a frequency of 50 kHz and attached to a base.

2.3. Experiment methods

2.3.1. Preparation of resveratrol standard curve

HPLC was used for the standard curve and yield determination of resveratrol based on the Li et al. technique [16]. The mobile phase consisted 0.1% of formic acid aqueous solution (solvent A) and acetonitrile (solvent B). The gradient elution program at 1 mL/min was as follows: 0–10 min, 30%-30% B; 10–11 min, 30%-90% B; 11–15 min, 90%-90% B; 15–17 min, 90%-30% B, and then held for 20 min. The wavelength was 306 nm and the temperature used for HPLC was 25 °C. And then, 9 mg of resveratrol standard was accurately weighed in a 5 mL brown volumetric flask, and the solution was fixed using methanol as solvent to obtain 1.8 mg/mL of resveratrol reserve solution. The reserve solution was prepared into 0.03, 0.06, 0.09, 0.12, 0.15 and 0.18 mg/mL standard solution, and the mobile term had an injection volume of 20 μL. The calibration curve obtained for this experiment ranged from concentrations of lycopene between 0.03 mg/mL to 0.18 mg/mL (R² = 0.9978), following the equation (1),

$y=2\times {10}^{8}X+273740,$ (1) where y represents peak area and X denotes sample concentration.

After the ultrasonic-assisted extraction, the resulting extract solutions were subjected to a 10-minute centrifugation at 3000 rpm to precipitate the solids, followed by filtration to collect the supernatants. An injection volume of 20 μL was utilized for the analysis. Identification of resveratrol was achieved by comparing its retention time with that of a reference standard. The calibration curves established for the substance facilitated the quantification of resveratrol using the external standard method.

2.3.2. IL ultrasound-enzyme-assisted extraction of resveratrol from P. Cuspidatum

In a round-bottom flask, 0.5 g of P. Cuspidatum powder was weighed and combined with a specific mass ratio of ionic liquid aqueous solution and cellulase, then using UEAE and the yield was determined according to the method of 2.3.1. Anion type, cation side chain length, pH (4.0, 4.5, 5.0, 5.5, 6.0), enzyme concentration (1.6%, 1.8%, 2.0%, 2.2%, 2.4%), extraction temperature (40, 45, 50, 55, 60 °C), extraction time (30, 60, 90, 120, 150 min), ultrasound power (100, 150, 200, 250 W), IL concentration (0.25, 0.50, 1.00, 1.50, 2.00, 2.50, 3.00 mol/L), and liquid–solid ratio (10, 15, 20, 25, 30, 35, 40 mL/g) were used to optimize for obtaining the highest resveratrol yield.

2.3.3. Plackett-Burman design to screen for significance factors

To identify the significant factors influencing the test results and further optimize them, which can drastically cut down on the amount of time and optimizations required, it is necessary to screen the significant factors affecting the test results based on a single factor when the response results are influenced by multiple variables. To this end, the PBD method was employed to evaluate the impact of seven key parameters on the resveratrol yield. These parameters encompass the pH value, enzyme concentration, temperature, time, IL concentration, ultrasonic power, and the liquid–solid ratio (Table 1).

Table 1.

Factors and levels of Plackett-Burman design for resveratrol extraction.

Factors	Coded symbols	Levels
		−1	+1
Enzyme concentration (%)	A	1.8	2.2
pH	B	4.0	5.5
Extraction temperature (°C)	C	35	50
Time (min)	D	30	60
Ultrasonic power (W)	E	150	250
IL concentration (mol/L)	F	0.5	1.0
Liquid-solid ratio (mL/g)	G	15	25

2.3.4. Analysis of conditions for optimising resveratrol yield by Box-Behnken design

In this study, three pivotal variables—enzyme concentration (A), temperature (B), and liquid–solid ratio (C)—were meticulously optimized using the BBD method to explore the intricate interplay of these factors on the yield of resveratrol. A three-factor, three-level BBD test was conducted using Design-Expert 13.0 software, with fluctuation ranges of 1.8–2.2 %, 40–60 °C and 20–30 mL/g for A, B and C, respectively. Based on the five major experiments of the design, the BBD design included 51 experimental sites to estimate the pure error sum of squares, which effectively ensured the stability and reproducibility of the test results [17], and the levels of the variables were shown in Table 2.

Table 2.

Variables and experimental design levels for response surface.

Independent variables	Coded symbols	Levels
		−1	0	1
Enzyme concentration (%)	A	1.8	2.0	2.2
Extraction temperature (°C)	B	40	50	60
Liquid-solid ratio (mL/g)	C	20	25	30

2.3.5. Analysis of conditions for optimizing resveratrol yield by ANN-GA

This study employed an ANN to elucidate the impact of various factors on the yield of resveratrol. The ANN model was constructed using the Neural Net Fitting toolbox within Matlab R2019b software, employing a three-layer backpropagation (BP) neural network architecture. The network was designed with an input layer, a hidden layer, and an output layer. Key input variables encompassed enzyme concentration, temperature, and the liquid–solid ratio, with the measured resveratrol yield serving as the output variable for model establishment. Prior research has demonstrated that with a sound experimental design, even a modest dataset can be leveraged to create an accurate ANN model [18]. In this investigation, we utilized 51 BBD test samples to construct the ANN model, aligning with the sample size used for RSM model development. The ANN model development was executed by randomly partitioning the dataset into three distinct subsets: 70% for training, 15% for testing, and 15% for validation. This approach facilitated the ANN’s self-adaptation, self-learning, and predictive capabilities, culminating in the identification of an optimal model configuration. Subsequently, the developed ANN model was integrated with GA for the optimization of resveratrol yield parameters. Utilizing Matlab R2019b’s Optimization tool, we selected GA as the solver, with three variables for optimization. The Plot function was set to “best fit,” and the maximum number of genetic iterations was set to 100, with all other parameters retaining their default values. Following the execution, we obtained a plot depicting the average fitness change across iterations, which was instrumental in determining the optimal resveratrol yield value.

2.3.6. Comparison with conventional extraction methods and scanning electron microscopy analysis

In the current investigation, we compared two different extraction methods to the IL-UEAE approach. Pure water was chosen as a reference solvent for UEAE of resveratrol from P. Cuspidatum. All other components of the extraction experiment were carried out under ideal circumstances, with the exception of the kind of solvent. Specifically, 0.5 g powder, a solution pH of 5.5, an enzyme concentration of 2.2%, an extraction time of 30 min, a temperature of 50 °C, an ultrasonic power of 250 W, and a liquid–solid ration of 25 mL/g. The resveratrol yield was ascertained by the method outlined in section 2.3.1, which was subsequently calculated using a specified formula.

For heat reflux extraction (HRE), utilizing 0.5 g powder with 80% aqueous ethanol as the solvent, under reflux conditions for 1 h at a liquid–solid ration of 25 mL/g. Consistent with the water-UEAE procedure, the analysis was conducted using the method outlined in section 2.3.1.

Utilizing a tungsten filament scanning electron microscope (SEM), we examined the microstructural alterations on the surface of variously treated powder samples. The SEM was operated under a magnification range of 150 to 1000 times, with an acceleration voltage of 5.00 kV and a working distance of 13.5 mm. The comparative analysis was conducted on four distinct samples.

2.3.7. Influence of the extraction cycles on resveratrol yield

In the aforementioned tests, the samples were extracted just once, so in order to maximise the resveratrol yield, this study investigated the effect on resveratrol yield using [C4mim]Br ionic solvent systems with extraction cycles of 1, 2, 3, 4, and 5 times, respectively. Using the previously established ideal conditions, five sample groups were created.

2.3.8. Statistical analysis

The experimental data were statistically analyzed using Design Expert 13 software, MATLAB R2019b software and GraphPad Prism statistical software 7.0. In each experiment, the data were presented as the mean value ± the standard deviation (SD), with each experiment being conducted at least three times. Significant differences at the p < 0.05 level were taken into account using one-way analysis of variance (ANOVA) and Tukey’s multiple comparisons check. Design Expert 13 was used to statistically design and optimize the RSM for evaluation of the response to regression equations, determination of the contribution and significance of the parameters, generation of the response surface plots, and determination of the best extraction strategy. MATLAB R2019b was used to construct, test, and validate the ANN-GA.

3. Results and discussion

3.1. Effect of ionic liquid anion type and cation side chain length on resveratrol yield

The comparative resveratrol yields from the [C4mim]⁺ ionic liquid system with varying anions are depicted in Fig. 1a. Analysis of the data presented in Fig. 1a revealed that the extraction rates of resveratrol from P. Cuspidatum, when ranked from highest to lowest, are as follows: Br^-, [OAC]^-, Cl^-, [BF₄]^-, [CF₃SO₃H]^-, and [HSO₄]^-. Correspondingly, the resveratrol yields were arranged in descending order as 2.34 ± 0.03, 2.16 ± 0.05, 2.03 ± 0.06, 1.88 ± 0.03, 1.79 ± 0.02, and 1.54 ± 0.03 mg/g, respectively. It is indicated that the efficacy of the ionic liquid system in extracting chemical constituents is predominantly influenced by the anions' capacity to form hydrogen bonds [19]. Notably, a higher propensity for hydrogen bond formation correlates with an enhanced extraction performance. Among the ionic liquids studied, [C4mim]Br exhibits the most favorable water solubility and is thus predisposed to establish a hydrogen bonding network, thereby augmenting the extraction efficiency of the system [11]. Furthermore, a comparative analysis between Br^- and Cl^- anions with respect to resveratrol yield demonstrates that an increase in the ionic radius of Br^-, and consequently its polarizability, leads to a proportional increase in the extracted resveratrol yield. These observations were congruent with the experimental data reported by Peng et al. [12].

Fig. 1.

Effects of anion (a) and alkyl chain lengths of cation (b) on the extraction yield of resveratrol from P. Cuspidatum. Yields are the mean ± SD (n = 3 replicates).

Fig. 1b illustrated the resveratrol yields obtained from Br-ionic liquid systems with varying lengths of cationic alkyl side chains. Notably, an augmentation in the alkyl side chain length correlates positively with the resveratrol yield extracted from P. Cuspidatum, as evidenced by a rise from an average yield of 1.83 ± 0.04 mg/g to 2.25 ± 0.08 mg/g. This enhancement is attributed to the diminished polarization between anions and cations within the ionic liquid solvent system as the alkyl side chain lengthens. The reduced polarization facilitates a more rapid ionic mobility within the aqueous phase, thereby enhancing the transfer rate of resveratrol and other constituents from the plant matrix to the solution [20]. Consequently, this accelerates the overall production of resveratrol. Additionally, the hydrophobicity of the ionic liquids increases as the length of the alkyl side chain grows [12]. Based on the principle that “similar substances tend to dissolve each other,” the less water-soluble resveratrol is more readily solvated in a more lipophilic environment. Consequently, the yield of resveratrol was augmented in the presence of 1-butyl-3-methylimidazole bromide, which provides a more lipophilic solvent milieu.

3.2. Analysis of single factors affecting resveratrol yield in P. Cuspidatum

3.2.1. Effect of solution pH on resveratrol yield

As depicted in Fig. 2a, the yield of resveratrol exhibited a biphasic trend, initially increasing and subsequently decreasing with the rise in solution pH. The apex of this trend was observed at a pH of 5.5, where the resveratrol yield peaked at 2.06 ± 0.03 mg/g. This pattern can be attributed to the optimal pH conditions for the enzymatic activity of cellulase, which catalyzes the most effective breakdown of plant components within the solvent system. Excessive acidity in the solution pH can denature the cellulase, leading to its inactivation, while a neutral or alkaline environment may impair the enzyme’s efficiency and rate of cell wall decomposition [21]. Consequently, to maximize the extraction yield of resveratrol, a pH range of 4.0–5.5 was deemed appropriate for further experimental optimization.

Fig. 2.

Effects of pH (a), extraction temperature (b), enzyme concentration (c), time (d), ultrasonic power (e), IL concentration (f) and liquid–solid ratio (g) on the the extraction yield of resveratrol from P. Cuspidatum. Yields are the mean ± SD (n = 3 replicates).

3.2.2. Effect of extraction temperature on resveratrol yield

As depicted in Fig. 2b, the production of resveratrol initially rose and then fell as the temperature for enzyme digestion increased. The apex of resveratrol yield was observed at a digestion temperature of 50 °C, amounting to 2.16 ± 0.03 mg/g. Raising the temperature not only reduces the viscosity of the ionic liquid solvent system, thereby enhancing its interaction with the powder, but also aids in lowering the activation energy required for cellulase reactions [22]. This enhancement in temperature facilitates the activity of cellulase and accelerates the mobility of ions within the solvent, culminating in an increased yield of resveratrol. Conversely, an excessive rise in temperature can impair the activity and efficacy of cellulase, hindering the thorough cleavage of the cell wall. This inadequacy can prevent the full release of cellular components and may even lead to the oxidative degradation of resveratrol, thereby damaging its molecular integrity [23]. Consequently, the optimal temperature range for enzymatic digestion, based on these considerations, was suggested to be between 35 and 50 °C for further experimental optimization.

3.2.3. Effect of enzyme concentration on resveratrol yield

As depicted in Fig. 2c, there was a progressive enhancement in resveratrol production with an incremental enzyme concentration, which subsequently plateaus. Saturation is achieved upon reaching an enzyme concentration of 2.0%, yielding a resveratrol output of 2.31 ± 0.01 mg/g. At first, the low amount of cellulase added poses a challenge for the breakdown of the plant cell wall by the cellulase and ionic liquid system, hindering the effective dissolution of plant components with the ionic liquid. Further escalation in enzyme concentration leads to a more complete degradation of the cell wall of powder, allowing for a more substantial release of the bioactive components, which in turn results in a significant surge in resveratrol yield. When an excessive amount of cellulase is added, the cell wall breakdown is already fully realized [24], and the cellular constituents have been fully released, leading to a saturation point beyond which the yield no longer increases, and the reaction reaches a state of completion. To mitigate the inefficiency and waste associated with an excessive amount of cellulase, a subsequent optimization process will focus on the enzyme concentration range of 1.8–2.2 %.

3.2.4. Effect of extraction time on resveratrol yield

As illustrated in Fig. 2d, the resveratrol yield exhibited a noticeable oscillatory trend, initially increasing with enzymatic hydrolysis time, then declining, and finally rebounding, reflecting a pronounced undulating pattern. This behavior can be attributed to several factors. Initially, the substrate concentration was notably high, facilitating the rapid hydrolysis of cell wall constituents by the ionic liquid and cellulase. This rapid hydrolysis, coupled with the abundant interaction between the ions and the particulate components within the solution, led to a marked increase in the concentration of resveratrol. However, as the reaction progressed and the temperature within the system rose, resveratrol was prone to oxidation, leading to the formation of resveratrol glycosides. This chemical transformation resulted in a diminished yield of resveratrol from the extraction process [1]. Upon the exhaustion of the air within the bottle, and given an ample reaction time, resveratrol glycosides are enzymatically transformed into resveratrol. This enzymatic conversion results in a subsequent increase in the concentration of resveratrol, ultimately reaching a state of saturation [25]. The findings underscore the relatively poor thermal stability of resveratrol, indicating that prolonged enzymatic hydrolysis could lead to significant variations in its yield. Notably, the concentration of resveratrol achieved within the initial 30 min of enzymatic hydrolysis was measured at 2.19 ± 0.02 mg/g. Further analysis revealed that extending the enzymolysis time beyond the initial phase did not yield a substantial increase in resveratrol yield. Therefore, to optimize the efficiency of subsequent testing procedures and to minimize both time and cost, it was determined that an extraction time of 30–60 min would be the most effective for test optimization in future studies.

3.2.5. Effect of ultrasound power on resveratrol yield

As depicted in Fig. 2e, there was a notable correlation between the resveratrol yield and the intensity of the applied ultrasound power. The apex of resveratrol yield, reaching 2.36 ± 0.02 mg/g, was observed at an ultrasound power setting of 250 W. This escalation in power facilitates a more thorough collapse of the powder within the confines of the vessel, thereby allowing for a more comprehensive outflow of plant fractions. This, in turn, ensures a more complete interaction with the solvent system, leading to an enhanced extraction yield. Furthermore, the augmentation of ultrasound power also intensifies the cavitational effect. An increase in power results in a more pronounced acoustic chemical effect, generating larger cavitation bubbles and a more robust sound pressure. Within a given volume, there is a threshold to the number of acoustic bubbles that can form, which means that a portion of the energy is inevitably transformed into heat. Within an optimal range, this elevation in temperature can accelerate the interaction between ions in the solution and the active constituents, thereby further augmenting the yield of resveratrol [26]. Therefore, the ultrasound power range between 150–250 W was identified as optimal and chosen for further experimental refinement.

3.2.6. Effect of concentration of ionic liquids on resveratrol yield

As depicted in Fig. 2f, resveratrol yield displayed a typical trend of initial rise followed by a decline as the concentration of ionic liquid increased. The peak of resveratrol yield was achieved at a concentration of 1.0 mol/L. This finding is supported by prior research [11], [12], confirming that higher concentrations of ionic liquids lead to an increase in the solvent system’s viscosity. This increase in viscosity adversely affects the solvent’s capacity to penetrate plant cellular structures, consequently diminishing the extraction efficiency, a finding that aligns with our results. Furthermore, the yields of resveratrol at concentrations of 0.5 mol/L and 1.0 mol/L were measured to be 2.42 ± 0.04 mg/g and 2.45 ± 0.06 mg/g, respectively. Given the marginal difference in yields between these two concentrations, and in consideration of the economic implications of the experimental procedure, the concentration range of 0.5–1.0 mol/L for [C4mim]Br was determined to be suitable for further experimental optimization.

3.2.7. Effect of liquid–solid ratio on resveratrol yield

As depicted in Fig. 2g, the yield of resveratrol exhibited a stabilizing trend following an initial increase with the rising liquid–solid ratio, ultimately reaching a plateau at a ratio of 25 mL/g, corresponding to a resveratrol yield of 2.41 ± 0.01 mg/g. Initially, the quantity of the added ionic liquid system was minimal, resulting in insufficient contact area with the powder, which hindered the reaction of some of the powder with the ionic liquid system, thereby leading to a low yield of resveratrol. However, as the quantity of the ionic liquid system was progressively increased, the volume expanded, allowing for more comprehensive contact and interaction between the powder and the solvent system [26]. Consequently, the yield of resveratrol demonstrated a consistent upward trajectory. Nonetheless, when the addition of the ionic liquid approached or surpassed the point of saturation, the yield of resveratrol began to level off. To prevent the unnecessary consumption of a substantial amount of ionic liquid solvent, a liquid–solid ratio of 15–25 mL/g was chosen for further optimization in subsequent experiments.

3.3. Plackett-Burman design to screen for significance factors

Utilizing the Plackett-Burman Design (PBD), this study meticulously optimized seven pivotal parameters—enzyme concentration (A), pH (B), extraction temperature (C), time (D), ultrasonic power (E), ionic liquid concentration (F), and liquid–solid ratio (G)—to enhance the yield of resveratrol. The PBD test matrix with resveratrol yield as response variable is listed as shown in Table 3.

Table 3.

Different combinations of PBD variables and average resveratrol yield.

Run	Factors							Response
	A (%)	B (pH)	C (°C)	D (min)	E (W)	F (mol/L)	G (mL/g)	Y (mg/g)
1	1.8	5.5	50	30	250	1.0	25	2.44 ± 0.02
2	1.8	4.0	50	30	250	1.0	15	1.78 ± 0.04
3	2.2	5.5	35	30	150	1.0	15	1.92 ± 0.05
4	1.8	5.5	50	60	150	0.5	15	2.01 ± 0.08
5	1.8	5.5	35	60	250	0.5	25	1.93 ± 0.05
6	2.2	4.0	50	60	250	0.5	15	2.24 ± 0.07
7	2.2	5.5	50	30	150	0.5	25	2.58 ± 0.11
8	2.2	4.0	50	60	150	1.0	25	2.77 ± 0.06
9	2.2	5.5	35	60	250	1.0	15	1.57 ± 0.03
10	1.8	4.0	35	30	150	0.5	15	1.33 ± 0.09
11	2.2	4.0	35	30	250	0.5	25	2.25 ± 0.06
12	1.8	4.0	35	60	150	1.0	25	1.86 ± 0.10

The PBD model was established by fitting the average resveratrol yield, and the first-order polynomial equation for the average resveratrol yield was obtained as shown in Equation (2), and the ANOVA of its PBD model was shown in Table 4.

$$Y=1.57+0.1618A+0.0203B+0.2377C+0.00046D-0.0235E+0.0012F+0.2462G$$ (2)

Table 4.

Results of the Plackett-Burman design.

Source	Sum of Squares	df	Mean Square	F Value	p-Value
Model	1.73	7	0.2474	9.00	0.0254*
A	0.3143	1	0.3143	11.43	0.0278*
B	0.0050	1	0.0050	0.1804	0.6929
C	0.6779	1	0.6779	24.65	0.0077**
D	0.0003	1	0.0003	0.0091	0.9285
E	0.0066	1	0.0066	0.2407	0.6494
F	0.0000	1	0.0000	0.0006	0.9816
G	0.7275	1	0.7275	26.46	0.0068**
Residual	0.1100	4	0.0275
Cor Total	1.84	11

Note: * p < 0.05; ^**p < 0.01.

As presented in Table 4, the calculated F value of 9.00 confirms the statistical significance of PBD test model. The influence of the seven parameters on the yield of resveratrol is ranked as follows: liquid–solid ratio (G), extraction temperature (C), enzymatic concentration (A), ultrasonic power (E), pH (B), time (D), and IL concentration (F). Specifically, enzyme concentration (A), extraction temperature (C), and liquid–solid ratio (G) were found to have significant effects on the model (p < 0.05). In contrast, the pH (B), extraction time (D), ultrasonic power (E), and IL concentration (F) were determined to have non-significant effects. Notably, the parameters of enzyme concentration (A), extraction temperature (C), and liquid–solid ratio (G) positively contribute to the resveratrol yield, with an observed trend of increased yield corresponding to higher values of these parameters.

The significance analysis plat for resveratrol yield was shown in Fig. 3. Normally, factors with t value higher than the limit of t value are considered as significant parameters, while factors not exceeding the t line are not significant parameters. From the figure, it can be seen that the selected significant parameters were enzyme concentration (A), extraction temperature (C) and liquid–solid ratio (G) and all of them had a positive effect on the resveratrol yield, which is in line with the results of Table 4, and to minimize resource waste from excessive optimization, the highest resveratrol yield obtained from the single factor test was used as a reference. The subsequent optimization phase of the BBD test, three critical parameters were focused on: enzyme concentration, extraction temperature, and liquid–solid ratio. Moreover, the optimized conditions included a pH value of 5.5, an extraction time of 30 min, an ultrasonic power of 250 W, and an IL concentration of 0.5 mol/L.

Fig. 3.

Pareto chart of the effect of seven variables on the extraction yield of resveratrol from P. Cuspidatum. Variables with t-values higher than the critical value (2.78) are regarded as statistically significant.

3.4. Analysis of conditions for optimizing resveratrol yield by Box-Behnken design

3.4.1. Establishment and statistical analysis of BBD model for resveratrol extraction

The model test matrix and results of resveratrol extraction were shown in Table 5. The non-linear regression equations between the design parameters enzyme concentration (A), extraction temperature (B) and liquid–solid ratio (C) and resveratrol yield (Y) with simultaneous linear, interaction and quadratic terms were obtained from the model analysis of BBD as shown in Equation (3).

$$\begin{array}{c}Y=2.36+0.1657A+0.1916B+0.187C+0.0799AB-0.0385AC\\ +0.1303BC+0.0481A^2-0.1070B^2-0.0386C^2\end{array}$$ (3)

Table 5.

Box-Behnken design for the experimental values for Resveratrol yield.

Run	A B C			Y
	Enzyme concentration (%)	Extraction temperature (°C)	Liquid-solid ratio (mL/g)	Resveratrol yield (mg/g)
1	1.8	50	30	2.02 ± 0.03
2	1.8	40	25	1.99 ± 0.04
3	1.8	50	20	2.21 ± 0.13
4	1.8	60	25	2.14 ± 0.03
5	2	60	20	2.29 ± 0.05
6	2	50	25	2.37 ± 0.07
7	2	50	25	2.02 ± 0.03
8	2	40	30	1.97 ± 0.10
9	2	40	20	2.72 ± 0.03
10	2	60	30	2.38 ± 0.08
11	2	50	25	2.43 ± 0.11
12	2	50	25	2.37 ± 0.19
13	2	50	25	2.74 ± 0.10
14	2.2	60	25	2.62 ± 0.17
15	2.2	50	30	2.25 ± 0.05
16	2.2	40	25	2.36 ± 0.20
17	2.2	50	20	2.02 ± 0.03

The significance, variance and regression equation of the BBD model were analyzed according to the non-linear regression equation as shown in Table 6.

Table 6.

Analysis of variance of regression equation for Box-Behnken Design.

Source	Sum of squares	df	Mean square	F value	p-Value
Model	0.9560	9	0.1062	28.26	0.0001**
A	0.2196	1	0.2196	58.43	0.0001**
B	0.2938	1	0.2938	78.16	<0.0001**
C	0.2800	1	0.2800	74.48	<0.0001**
AB	0.0256	1	0.0256	6.80	0.0351*
AC	0.0059	1	0.0059	1.58	0.2492
BC	0.0679	1	0.0679	18.06	0.0038**
A2	0.0097	1	0.0097	2.59	0.1518
B2	0.0482	1	0.0482	12.83	0.0089**
C2	0.0063	1	0.0063	1.67	0.2377
Residual	0.0263	7	0.0038
Lack of fit	0.0167	3	0.0056	2.31	0.2179
Pure error	0.0096	4	0.0024
Cor Total	0.9823	16

Note: * p < 0.05; ^**p < 0.01.

As can be seen from Table 6, in the significance analysis of the model, the F value is 28.26 and the p value of the model is 0.0001, which represents that the model is significant with 99.99% confidence level used to explain the actual results of the test, and it can be found that the p-value of the linear term, the interaction term, and the secondary term in A, B, C, AB, BC, and B² are all less than 0.05, which is a significant variable factor.

The optimal extraction process conditions for resveratrol extraction by UEAE were obtained through single factors test, Plackett-Burman test for screening of significant factors and RSM for further optimization and analysis, with enzyme concentration of 2.171%, extraction temperature of 58.93 °C, and liquid–solid ratio of 28.934 mL/g. The analyzed conditions were adjusted according to the actual production. The process conditions were 2.17% enzyme concentration, extraction temperature of 59 °C, liquid–solid ratio of 29 mL/g, pH value of 5.5, extraction time of 30 min, ultrasonic power of 250 W, and IL concentration of 0.5 mol/L. The resveratrol yield obtained in the validation test was 2.78 ± 0.12 mg/g, which was highly close to the predicted value of 2.77 ± 0.08 mg/g from the response surface, indicating that the optimization results of the RSM are reliable and can be used to predict the real experimental results.

3.4.2. Response surface analysis of resveratrol extraction model

ANOVA and regression equation analysis of the BBD model were performed to compare the significance of the significant factors enzyme concentration (A), extraction temperature (B) and liquid–solid ratio (C) affecting resveratrol yield as linear, quadratic and interaction terms in the regression equation. Three-dimensional and two-dimensional response surface plots were drawn through the interaction of factors presenting the interaction effect between each two factors as shown in Fig. 4.

Fig. 4.

3D response surface graph (a and c) and contour plot (b and d) for the interactions of different extraction parameters (A: enzyme concentration, %; B: temperature, °C; C: liquid–solid ratio, mL/g) on the extraction yield of resveratrol from P. Cuspidatum.

As shown in Fig. 4a and b, the yield of resveratrol steadily increases with the elevation of enzyme concentration, eventually reaching a plateau. Additionally, an increase in the extraction temperature is associated with a positive trend in resveratrol yield, and the trend of the increase of extraction temperature accompanied by the rise of resveratrol yield was more obvious compared with the enzyme concentration, which may be due to the fact that in the interaction between the two parameters, the impact of extraction temperature on resveratrol yield was more significant than that of enzyme concentration, a finding that aligns with the screening results from the PBD model.

Fig. 4c and d illustrated the intricate interplay between extraction temperature and liquid–solid ratio, highlighting a pronounced and ascending trend in resveratrol yield. This relationship is characterized by a sharply inclined two-dimensional curve, with the variation in liquid–solid ratio showing a more distinct influence on the yield compared to the enzyme digestion temperature. These observations are in close agreement with the PBD screening outcomes. While enzyme concentration (A) and liquid–solid ratio (C) were scrutinized, their insignificance in the context of this study precludes a detailed discussion here.

3.5. Analysis of conditions for optimizing resveratrol yield by ANN-GA

Neural network training and simulation were carried out on the basis of response surface, and 51 sets of data were used for training, in which the input layer was 3, enzyme concentration, extraction temperature, and liquid–solid ratio; and the output layer was 1, resveratrol yield. The predictive model was established by ANN-GA, and the predictive model was used as the fitness function of the ANN-GA for individual screening and then for optimization of the resveratrol extraction process, whose training set, validation set, test set, and all the data are shown in Fig. 5a and b. To evaluate the fit quality, various parameters were computed, yielding an MAE of 0.03, an MSE of 0.002, an RMSE of 0.04, a MAPE of 17.2891%, and R² value of 0.98324, their fitting coefficients are all above 0.9 proving that the ANN-GA model is successfully established and combines tightly with the actual values as shown in Fig. 5c, which indicates that ANN-GA is more effective for process optimization of resveratrol yield. Combining ANN and GA for global optimization to obtain the best resveratrol extraction process, the optimization results are shown in Fig. 5d, which reaches the best fitness after 60 generations, after which its curve is smooth and the resveratrol yield has reached the maximum value.

Fig. 5.

Optimization results of ANN-GA. Regression analysis (a), performance (b), relationship between the actual values and predicted values (c) and fitness curve with evolutionary algebra (d).

ANN-GA optimized the optimal process conditions as 2.179% enzyme concentration, extraction temperature 58.475 °C, liquid–solid ratio 28.706 mL/g. The conditions were modified as 2.18% enzyme concentration, extraction temperature 58 °C, liquid–solid ratio 29 mL/g, and the rest of the influencing factors were the same as the response surface mentioned above, and the validation test yielded the yield of resveratrol 2.90 ± 0.15 mg/g, which highly matched with the neural network optimization result of 2.89 ± 0.06 mg/g indicating that the optimization results of the ANN-GA are reliable and could be used to predict the real test results.

3.6. Experimental verification of RSM and ANN-GA optimization results

As can be seen from Table 7, the relative error of the ANN-GA optimisation (0.11%) is lower than that of the RSM optimisation (0.29%), and the resveratrol yield (2.90 ± 0.15 mg/g) is higher than that of the RSM optimisation results (2.78 ± 0.12 mg/g), it can be concluded that the ANN-GA model has a more accurate fit and estimation and prediction ability than the RSM model, which indicates that the results of the ANN-GA optimisation are better compared with the RSM, and it proves that this method is suitable for the optimisation of the extraction of resveratrol yield in P. Cuspidatum feasible and the experimental results are reliable.

Table 7.

Comparison of optimization results.

Optimistic method	Enzyme concentration (%)	Extraction temperature (°C)	Liquid-solid ratio (mL/g)	Predicted value (mg/g)	Measured value (mg/g)	Relative error (%)
RSM	2.17	59	29	2.77 ± 0.08	2.78 ± 0.12	0.29
ANN-GA	2.18	58	29	2.89 ± 0.06	2.90 ± 0.15	0.11

3.7. Comparison with conventional extraction methods and scanning electron microscopy analysis

The resveratrol yield from reflux extraction using 80% ethanol, at 2.03 ± 0.09 mg/g, was lower compared to the 2.90 ± 0.15 mg/g yield achieved through UEAE with ionic liquids. This indicates that the IL-UEAE method is more efficient, convenient, and faster, enhancing extraction yield and reducing experimental time. The yield of resveratrol extracted using UEAE with pure water was significantly lower at 0.87 ± 0.12 mg/g, compared to the IL-UEAE method, thereby highlighting the substantial impact of ionic liquids on enhancing extraction efficiency.

The microscopic changes in the surface of P. Cuspidatum powder after different treatments were observed using a scanning electron microscope with a tungsten filament at a magnification of 150–1000 times, an accelerating voltage of 5.00 KV and a working distance of 13.5 mm. As can be seen from Fig. 6 compared to the untreated powder, the ethanol reflux-treated material shows a slightly wrinkled surface, but its spatial structure is roughly the same as the untreated one (Fig. 6aa₁a₂ and bb₁b₂). The UEAE with pure water images (Fig. 6cc₁c₂) show that the cavitation effect of ultrasound disrupts the cell wall and damages the cellular organisation, leading to cell rupture to form a hollow structure. In contrast, more cell tissue was damaged and cavity structures were more pronounced in the IL-UEAE pictures (Fig. 6dd₁d₂), illustrating that the combined action of the IL and UEAE methods can lead to increased plant cell fragmentation. The results indicate that employing ionic liquids in conjunction with UEAE effectively disrupts the microstructure of plant tissues, facilitating solvent penetration and dissolution. Nonetheless, the incorporation of ionic liquids escalates the extraction costs. For a seamless integration into large-scale manufacturing, it is imperative to delve into more efficient strategies for the recovery and recycling of ionic liquids.

Fig. 6.

Scanning electron microscopy of P. Cuspidatum powder. Untreated powder (a:×150; a₁:×300; a₂:×1000), dried powder after ethanol reflux extraction (b:×150; b₁:×300; b₂:×1000), dried powder after UEAE with pure water (c:×150; c₁:×300; c₂:×1000) and dried powder after UEAE with ionic liquid (d:×150; d₁:×300; d₂:×1000).

3.8. Influence of the extraction cycles on resveratrol yield

The corresponding resveratrol yield changes in the [C4mim]Br system with extraction cycles of 1, 2, 3, 4 and 5 times were shown in Fig. 7. It can be seen that with the increase of the cycle, the resveratrol yield showed a trend of firstly increasing and then levelling off, and the resveratrol yield reached an equilibrium of 3.37 ± 0.06 mg/g when the extraction cycle was 3 times, and the active ingredient had been sufficiently crushed and flowed out, and in order to save the cost of the test and the time, the extraction cycle was selected to be 3 times.

Fig. 7.

Effect of extraction cycles on the extraction yield of resveratrol from P. Cuspidatum.

4. Conclusion

This research presents an efficient protocol for the extraction of resveratrol from P. Cuspidatum, leveraging the synergistic application of ultrasonic-enzyme-assisted extraction and ionic liquids. The experimental design was meticulously optimized through the RSM complemented by ANN-GA. The optimal extraction parameters were identified as follows: utilization of the [C4mim]Br ionic liquid system, an enzyme concentration rate of 2.18%, an enzymatic digestion temperature of 58 °C, a liquid-to-solid ratio of 29 mL/g, a pH value of 5.5, an enzymatic digestion duration of 30 min, an ultrasonic power of 250 W, and an ionic liquid concentration of 0.5 mol/L. Under these conditions, the resveratrol yield was maximized, achieving a yield of 2.90 ± 0.15 mg/g. Compared to alternative techniques, the proposed method offers enhanced eco-friendliness, superior extraction efficiency, and a considerably shortened extraction timeframe. This study introduces a novel, environmentally benign solvent system that simplifies the extraction process, enhances safety, and ensures feasibility. It offers a contemporary reference and theoretical underpinning for the application of resveratrol in the medical and health care sectors.

Author contributions

Hongyi Zhao and Zunlai Sheng conceived and designed the study. Junping Wang and Hongyi Zhao performed the experiments. Yutong Han acquired the data. Xin Wang analyzed the data. Hongyi Zhao drafted the manuscript. Zunlai Sheng critically revised the article and proved the final manuscript.

CRediT authorship contribution statement

Hongyi Zhao: Writing – original draft. Junping Wang: Methodology. Yutong Han: Formal analysis. Xin Wang: Formal analysis. Zunlai Sheng: Writing – review & editing, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.