Ultrasound-assisted deep eutectic solvent extraction of polysaccharides from Melastoma dodecandrum: RSM optimization, ANN prediction, kinetics, structural characterization and cellular antioxidant activity

Graphical abstract

Abstract

Natural polysaccharides have attracted much attention due to their biocompatibility, biodegradability, low toxicity and various biological activities. However, the extraction of polysaccharides from Melastoma dodecandrum Lour. (MD) in a sustainable and effective way remains a challenge. In the present study ultrasound-assisted deep eutectic solvent (UA-DES) extraction strategy was developed to improve the yield and bioactivity of MD polysaccharides (MDPs). Among the twenty-five prepared DES systems, DES-3.2 (choline chloride:citric acid, 1:2) showed the best extraction performance. Response surface methodology (RSM) was applied for optimization of the extraction parameters and artificial neural network (ANN) model was established to predict the extraction behavior and performance. In addition, extraction kinetics was successfully described by second-order kinetic model. Under the optimized conditions of liquid–solid ratio of 30 mL/g, ultrasonic temperature of 60 °C, ultrasonic power of 360 W and ultrasonic time of 90 min, the MDPs yield reached 1.38 %, which was increased by 51.65 % compared with conventional HWE. Furthermore, ANN outperformed RSM in prediction accuracy and effectively captured the complex nonlinear interactions inherent to the extraction process. Structural characterization (SEM, FT-IR, XRD, HPGPC and TGA) analysis revealed that UA-DES extraction caused significant disruption of plant cell walls without changing the characteristic functional groups and amorphous structure of MDPs accompanied by higher molecular weight and thermal stability. Monosaccharide composition analysis identified galactose, galacturonic acid, and glucose as the predominant components with UA-DES extraction significantly enrich the galacturonic acid content compared with HWE. Further, UA-DES extracted MDPs had stronger antioxidant capacity in chemical assays and exerted higher antioxidant and anti-inflammatory effects in cells by significantly decreasing intracellular reactive oxygen species (ROS) and nitric oxide (NO) levels in LPS induced RAW264.7 macrophages without cytotoxicity. Additionally, DES-3.2 maintained over 91.25 % recovery after five recycling cycles, with only a 5.78 % loss in extraction efficiency of MDPs. Despite the modest absolute yield of 1.38 %, the excellent recyclability of DES-3.2 provides a strong economic counterbalance by reducing solvent costs and waste generation. Overall, UA-DES extraction is an effective, green, high-performance approach for extracting structurally distinct and bioactive MDPs for applications in functional foods, nutraceuticals and pharmaceuticals.

1. Introduction

Polysaccharides are naturally occurring carbohydrates with high molecular weight. They are made up of monosaccharide units joined by glycosidic linkages. They are abundantly found in plants, algae, fungi, and microorganisms [1]. Different compositions of monosaccharide units, types of linkages, and molecular structures lead to broad spectrum of physicochemical and biological characteristics of polysaccharides [2]. Natural polysaccharides have attracted considerable interest for antioxidant, immunomodulatory, anti-inflammatory and antitumor activity due to their biocompatibility, biodegradability and low toxicity. They are frequently used as thickeners and stabilizers in the food industry, bioactive components in pharmaceuticals and nutraceutical industries, while in biomedical applications they are used for tissue engineering, wound healing, and drug delivery [3].

MD is a plant species of the Melastaceae family, widely consumed in South China and used in traditional practices as a food and medicinal source. It is generally regarded as a safe natural remedy for disorders related to inflammation [4]. Ethnopharmacological studies described its use in the treatment of ailments such as laryngitis, enteritis, and nephritis [5]. Along with these traditional claims, modern pharmacological studies demonstrated that MD extracts possess multiple biological activities, including hypoglycemic, hypolipidemic, antioxidant, and anti-inflammatory [4], [6]. Phytochemical studies found that MD exhibits a wide range of bioactive substances, including tannins, flavonoids, fatty acids, and triterpenoids [5]. Despite these promising pharmacological properties, research on MDPs is relatively limited. In our previous work we reported a novel acidic polysaccharide MDP2-1 from MD using conventional HWE [7]. Structural analysis revealed that MDP2-1 (29.234 kDa) is a homogeneous polysaccharide primarily composed of galactose, arabinose, rhamnose, glucose, glucuronic acid and galacturonic acid with a backbone of → 4)-α-D-GalpA-(1→,→2)-α-L-Rhap-(1→,→3,4)-α-D-GalpA-(1→, and → 2,4)-α-D-GlcpA-(1 → . MDP2-1 exerts significant anti-inflammatory effects by inhibiting the TLR4/NF-κB/MAPK signaling pathway in LPS induced acute lung injury [7]. Furthermore, we recently found that MD polysaccharide effectively alleviates ovalbumin induced allergic rhinitis in mice by modulating the NLRP3 inflammasome and IL-17 signaling axis, reducing Th17 cell differentiation and pyroptosis in nasal mucosa [8]. These findings confirm MDPs as promising candidates for treating inflammatory and allergic disorders. However, all these previous studies relied on HWE, a method known for prolonged extraction times, high consumption of energy, relatively low yield efficiency and compromising the bioactivity [9], [10], which severely limits MDPs practical application and scalability. Consequently, there is an urgent need to develop a more efficient, sustainable and mild extraction technology that can preserve or even enhance the structural integrity and bioactivity of MDPs. To the best of our knowledge no study has yet explored the application of green extraction techniques, such as UA-DES extraction for the recovery of bioactive polysaccharides from MD.

Extraction technology plays an important role in the research and development of natural polysaccharides. It significantly influences the yield, stability and biological activities of polysaccharides. To improve the efficiency of polysaccharide recovery a range of extraction assisted techniques have been developed, including ultrasound [11], microwave [12], enzyme [13] and other methods [14]. These approaches increased the disruption of plant cell wall and promote solute transportation. This covers the major constraints of traditional extraction and enhances the yield of polysaccharide. Furthermore, combining HWE with dilute acids, alkalis, or organic solvents can improve the polysaccharides yield. But sometimes such combinations damage the structural integrity and biological functionality of polysaccharides and raise issues about operator safety and environmental sustainability [15], [16]. As a result, the development of extraction technologies that are both highly efficient and environmentally friendly remains a crucial research priority. Ultrasound-assisted extraction (UAE) is distinguished by its lower energy requirements, shorter processing times and increased polysaccharides yield, while minimizing structural damage. Its unique physical mechanism generates the propagation of ultrasonic waves and acoustic cavitation, causing microbubbles to rapidly grow and disintegrate. These causes severe localized mechanical forces and efficiently break the cell wall structures, promote solvent penetration into the plant matrix and greatly increase mass transfer via microstreaming and turbulence [17].

Selection of suitable extraction solvent is important for successful and environmentally sustainable extraction. DES system has recently gained attention due to its excellent physicochemical properties and tunability. DES consists of two components a hydrogen bond acceptor (HBA) and additional hydrogen bond donors (HBD). Both combine under specific conditions to generate hydrogen bonds [18]. Deep eutectic solvent systems (DESs) have several major benefits that include ease of preparation, low volatility and minimum environmental and toxicological impacts. These DESs properties together with their tunable polarity and high solvation capacity increased the attention for the extraction of bioactive compounds from natural sources [17], [19]. Recently, many studies integrated UAE with DES system to improve the yield and bioactivity of plant polysaccharides [20], [21]. Recently, Zhang et al. [22] applied UAE approach using DES to isolate polysaccharides from Indocalamus tessellatus leaves. This method not only achieved a substantially higher polysaccharides yield as compared with conventional HWE but also produced polysaccharides with enhanced antioxidant activity. These findings highlight the strong potential of UA-DES extraction as an effective and promising strategy for the extraction of bioactive polysaccharides.

Although UA-DES extraction shows great potential in isolation of bioactive polysaccharides from plants, its application to MD remains largely unexplored. MDPs extracted through HWE, commonly limited by low efficiency. HWE demands high energy that may partially degrade the polysaccharides during extraction process. Currently, no study has been conducted to optimize the extraction conditions of UA-DES for MDPs. Therefore, effects of UA-DES on monosaccharide profile and bioactivity of MDPs are still unclear. Optimization of extraction process is useful for achieving conditions for maximum yield and efficiency. RSM is frequently used to optimize and evaluate the interactions between process factors and identify the conditions that deliver the best extraction performance. However, RSM can identify general trends, but their standard quadratic models often struggle to represent the nonlinear and complex interactions of characteristic [23]. This highlights the need for a strategy that finds nonlinear and complex interactions of variables and plays role in prediction of optimized yield for MDPs. ANN consist of interconnected neurons with nonlinear activation functions, which enable them to learn complex relationships between multiple variables [20], [24]. RSM provides statistical interpretation of factor interactions and ANN with RSM improved the predictive accuracy. A kinetic model is essential for understanding the dynamic behavior of polysaccharide extraction, including mass transfer and diffusion mechanisms [25]. Polysaccharides gradually diffuse from plant cells to solvent. During extraction process decreasing the concentration gradient creates a non-steady state system. Previous studies showed that this dynamic behavior is effectively described by using second-order kinetic models, which capture both rapid initial release and the subsequent approach toward equilibrium [26]. The obtained MDPs were characterized by FTIR, XRD, SEM and TGA. Monosaccharide composition, molecular weight distribution (HPGPC), uronic acid content, total phenolic content and protein content were also evaluated. Further, antioxidant activity of MDPs was determined by both chemical assays (DPPH, hydroxyl radical scavenging, lipid peroxidation inhibition, and reducing power) and cellular assays using RAW264.7 macrophages (cytotoxicity, intracellular ROS scavenging, and NO inhibition). DES recyclability was evaluated over five consecutive cycles to assess sustainability. The results demonstrate that UA-DES is an effective, efficient and environmentally sustainable strategy for recovering bioactive polysaccharides from MD.

2. Material and methods

2.1. Plant materials

A fresh MD aboveground part of plant sample was procured from Lishui, Zhejiang, China. Initially, plant material was washed with distilled water to cleanse the surface impurities. After this samples were dried at 40 °C until a constant weight was achieved. The dried material was then pulverized, sieved through a 60-mesh screen to ensure uniform particle size, and stored at 4 °C for later experimental use.

2.2. Chemicals and reagents

Choline chloride, DL-malic acid, lactic acid, glycerol, ascorbic acid, DPPH, salicylic acid, soy lecithin, glucose, trichloroacetic acid, thiobarbituric acid, potassium ferricyanide, D-galacturonic acid, and sodium tetraborate were purchased from Shanghai Macklin Biochemical Technology Co., Ltd, sulphuric acid from Lingfeng Chemical Co., Ltd. and absolute ethanol from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China).

2.3. Preparation and screening of DES

The DES in this investigation was synthesized using choline chloride (ChCl) as the standard HBA. Corresponding HBDs included DL-malic acid, lactic acid, citric acid, glycerol, and 1,4-butanediol. Each DES was synthesised by combining the HBA and HBD at predetermined molar ratios (Table S1) under continuous heating and agitation at 80 °C, continuing until a transparent, single-phase liquid formed, as described [27]. Prior to polysaccharide extraction from the MD substrate, each solvent was adjusted to the desired concentration by adding deionized water. The final DESs selection was based on maximizing polysaccharide recovery and carried forward for further study.

2.3.1. Recycling studies of DES

After polysaccharide precipitation and filtration, the remaining mixture contain DES, water and ethanol was processed for solvent recovery [28]. The collected polysaccharide extraction solution was enriched by rinsing the container with a small volume of anhydrous ethanol. This washed ethanol was combined with the stock polysaccharide extraction solution and then concentrated using a rotary evaporator at 65 °C to remove water and ethanol. Process allows the DES to be recovered and reused in subsequent extractions. The process efficiency was assessed by measuring the recovery of DES-3.2 and the yield of extracted polysaccharides. The extraction of polysaccharide content was evaluated by optimized extraction conditions (Table S4). The recovery rate of DES-3.2 was determined as presented in Eq. (1).

$$\mathrm{D}\mathrm{E}\mathrm{S}-3.2\mathrm{R}\mathrm{e}\mathrm{c}\mathrm{o}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{y}(\%)=\frac{V_1}{V_0}\times 100$$ (1)

where V₁ is the post evaporation recovered volume (mL) and V₀ is the initial DES-3.2 vol (mL) used for extraction.

2.4. Ultrasound-assisted extraction process

For the extraction, 1 g of MD powder was combined with 30 mL of DES in a stoppered Erlenmeyer flask. UAE was performed using an ultrasonic processor (20 kHz, probe Ø12 mm) model 20 K LS 1200B, Wuxi YuanshengTe Intelligent Technology Co., Ltd., China, attached with circulating water bath and continuous temperature monitor. The probe tip was immersed approximately 3–4 cm below the liquid surface and positioned away from the flask walls to ensure efficient heat and circulation of solvent around the tip. The UA-DES extraction employed a solvent with 30 % (w/w) water content at a liquid–solid ratio of 30 mL/g, 60 °C, 360  W and 90  min. For comparison ultrasonic-assisted water extraction (UA-WE) used the same parameters with distilled water instead of DES, and conventional HWE was conducted without ultrasonication, at same liquid–solid ratio (30  mL/g) and using distilled water at 90 °C for 60 min. The 30 mL solvent volume per 1 g of MD powder in UAE served as an effective heat sink, absorbing localized energy and dispersing it through convection. After extraction, separation of the solid phase was accomplished via centrifugation (4000 rpm, 5 min). Following centrifugation, the clear upper phase was harvested and precipitated by adding anhydrous ethanol at a 1:4 (v/v) ratio and refrigerated at 4 °C for 12 h to induce polysaccharide precipitation. The precipitated polysaccharides were recovered via centrifugation (4000 rpm, 5 min), rinsed with ethanol, and again dissolved in deionized water. Protein impurities were removed by mixing the polysaccharide solution with Sevag reagent (chloroform:n-butanol, 4:1 v/v) at a 4:1 vol ratio in a separation funnel. After agitation and following the phase separation, polysaccharide rich upper aqueous phase was carefully recovered. Dialysis was performed using a membrane with molecular weight cut-off (MWCO) of 3000 Da against distilled water for 72 h and water was replaced at regular intervals to ensure the effective removal of low molecular weight contaminants. The final polysaccharide extract was freeze dried to obtain crude MDPs. Each experiment was performed in triplicate to ensure statistical reliability. Phenol sulfuric acid method was used to determine the yield of MDPs [29].

2.5. Determination of polysaccharide contents

The total polysaccharide content was measured with phenol sulfuric acid method [30]. A calibration was prepared with glucose stock solution (1 mg/mL), which was volumetrically diluted to provide working standards concentrations of 0, 0.02, 0.04, 0.06, 0.08, and 0.1 mg/mL. Each 200 µL aliquot of glucose standard was mixed with 100 µL of a 5 % (w/v) phenol solution and 500 µL of concentrated sulfuric acid. The mixture was incubated at 100 °C for 30 min. After cooling with cold water for 10 min, absorbance was measured at 490 nm. The absorbance data was used for linear regression analysis and obtained this equation Y = 9.39429X + 0.01962. Which exhibited a strong linear fit (R² = 0.99933), confirming the reliability of the calibration for determination of MDPs yield. MDPs extract (200 µL) from each experimental run underwent identical processing conditions and absorbance was taken at 490 nm. Polysaccharide concentrations (C, mg/mL) were determined from the calibration curve. The MDPs yield % was calculated by using Eq. (2).

$$\mathrm{P}\mathrm{o}\mathrm{l}\mathrm{y}\mathrm{s}\mathrm{a}\mathrm{c}\mathrm{c}\mathrm{h}\mathrm{a}\mathrm{r}\mathrm{i}\mathrm{d}\mathrm{e}\mathrm{s}\mathrm{e}\mathrm{x}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{y}\mathrm{i}\mathrm{e}\mathrm{l}\mathrm{d}(\%)=\left(\frac{\mathrm{C}\times \mathrm{V}\times \mathrm{N}}{\mathrm{M}}\right)\times 100$$ (2)

where C = polysaccharide concentration (mg/mL), V = extract volume (mL), N = dilution factor, and M = mass of the raw plant material (g).

2.6. Experimental design for yield optimization of UA-DES extraction

2.6.1. Single-factor experiment

Initially, we performed single-factor analysis to identify the suitable operating ranges of key parameters to provide a foundation for RSM optimization. The effects of independent variables, such as DES water content (20 %, 30 %, 40 %, 50 %, 60 % w/w), liquid–solid ratio (10, 20, 30, 40, 50 mL/g), ultrasonic temperature (30, 40, 50, 60, 70 °C), ultrasonic power (200, 280, 360, 440, 520 W), and ultrasonic time (30, 60, 90, 120, 150 min), on MDPs yield was investigated. One factor is being tested in experiment across its designated levels, while other four factors were maintained at constant baseline conditions. DES water content of 40 %, liquid–solid ratio of 30 mL/g, an ultrasonic temperature of 50 °C, a power of 360 W, and duration of 90 min was constant baseline conditions (Table S2). To ensure statistical reliability, each experiment was performed in triplicate and averaged yield was determined. The resulting data were analysed to establish the ranges of parameters that maximize the MDPs recovery.

2.6.2. RSM based experimental design and optimization

After screening the single-factor, extraction parameters were optimized through a response surface methodology (RSM) approach, implemented with a Box-Behnken design (BBD). The experimental design was constructed and analysed using Design-Expert software (Version 13, Stat-Ease Inc., Minneapolis, MN, USA). The water content of the deep eutectic solvent was fixed at 30 % (w/w), while four independent variables liquid–solid ratio (A), extraction temperature (B), ultrasonic power (C), and ultrasonic time (D) were evaluated at three coded levels (−1, 0, and + 1), as summarized in Table S3. A total of 29 experimental runs were conducted in a randomized sequence to account for variability. MDPs yield was selected as response variable and the data were regressed to a second-order polynomial model as shown in the following Eq. (3).

$$Y={\beta }_0+\sum _{i=1}^4{\beta }_i{X}_i+\sum _{i=1}^4{\beta }_{\mathit{ii}}{X}_i^2+\sum _{i=1}^4\sum _{j=i=1}^4{\beta }_{\mathit{ij}}{X}_i{X}_j$$ (3)

In this model, Y denotes the predicted response. The term β₀ represents the constant (intercept), β_i corresponds to the first-order (linear) coefficients, quadratic effects are represented by β_ii, and interaction effects by β_ij. The symbols X_i and X_j represent the coded levels of the independent variables.

The model robustness and predictive capability were assessed through a set of regression derived statistical indicators such as coefficient of determination (R²), adjusted R², adequate precision and the coefficient of variation (CV). Interactive variable effects and optimal conditions were interpreted through three dimensional response surface plots and corresponding two-dimensional contour diagrams. Model reliability was further validated through examining the agreement between the experimentally obtained MDPs yields and corresponding values predicted by the model.

2.7. Artificial neural network-based modeling and prediction of MDPs yield

An ANN model was built by using the three-layer feedforward neural network fitting environment available in MATLAB R2025b to predict the extraction of MDPs yield. The network architecture consisted of four input neurons (A–D), one hidden layer containing 10 neurons for nonlinear learning, and a single output neuron representing MDPs yield (Fig. S2). The number of hidden neurons was selected after preliminary optimization trials based on achieving the minimum prediction error and highest regression performance. The ANN was trained through the Levenberg–Marquardt backpropagation method known as ‘trainlm’. This approach was selected for its fast convergence and good precision in nonlinear regression applications. During the training stage, network synaptic weights and neuron biases were repeatedly adjusted to minimize the gap between ANN predicted and experimental values. The entire dataset was randomly divided into training (70 %), validation (15 %), and testing (15 %) subsets using MATLAB’s default randomization procedure to ensure smooth learning, evaluate predictive performance, and minimize overfitting. The validation subset was specifically used to monitor network generalization during training and prevent overtraining of the model. Tan-sigmoid (‘tansig’) functions were applied to activate neurons in the hidden layer, while a linear (‘purelin’) function was selected for the output layer. To improve the robustness, stability and reproducibility of the ANN model, repeated training runs were conducted under the same network architecture and training conditions. During each training run the network weights and biases were randomly initialized and the final optimized ANN model was selected based on the lowest validation mean squared error (MSE) and highest regression coefficient (R²).

2.8. RSM and ANN predictive performance comparison

To evaluate the accuracy of comparative models, performance metrics such as R² (coefficient of determination), MSE (mean squared error), RMSE (root mean square error), MAD (mean absolute deviation), and MAPE (mean absolute percentage error) were calculated for RSM and ANN outputs. The chosen metrics provide a comprehensive analysis of the models validity, prediction accuracy and related error.

The R2 measures the proportion of variance in the MDPs yield data that was captured by the model, calculated as

$$R^2=1-{\sum }_{i=1}^n\frac{{(y_i-y_x)}^2}{{(y_i-y_z)}^2}$$ (4)

$$\mathit{MES}=\frac{1}{n}\sum _{i=1}^n{(y_i-y_x)}^2$$ (5)

$$\mathit{RMES}=\sqrt{\frac{1}{n}\sum _{i=1}^n{(y_i-y_x)}^2}$$ (6)

$$\mathit{MAD}=\frac{1}{n}\sum _{i=1}^n\left|\begin{array}{c}{y}_i-y_x\end{array}\right|$$ (7)

$$\mathit{MAPE}=\frac{100}{n}\sum _{i=1}^n\left|\begin{array}{c}\frac{y_i-y_x}{y_i}\end{array}\right|$$ (8)

In the above equations, $y_i$ represents the experimentally measured actual MDPs yield, $y_x$ indicates the corresponding predicted yield, $y_z$ represents the mean of experimentally determined MDPs yields and n denotes the total count of experimental runs. Progression of ANN training was evaluated by observing the MSE over successive epochs. To prevent overfitting, if validation error continuously increased for six consecutive iterations, the training process was automatically terminated. This approach ensured that the final model retained optimal predictive accuracy on unseen data. Lower values of MSE, RMSE, MAD and MAPE together with higher R² values indicate better predictive performance and stronger agreement between experimental and model predicted results.

2.9. Extraction kinetics simulation of MDPs

2.9.1. Kinetic model building

To investigate the extraction kinetics of MDPs, kinetic experiments were performed under the optimal UA-DES extraction conditions that were obtained from RSM and ANN optimization (Table S4). These optimum extraction conditions, liquid–solid ratio (30 mL/g), ultrasonic temperature (60^◦C), and ultrasonic power (360 W) were maintained constant, while extraction time was varied from 5 to 120 min. At time intervals of 5 min, aliquots of the extraction solution were collected and the polysaccharide concentration (Ct, mg·mL⁻¹) was measured by the phenol sulfuric acid method.

The second-order kinetic model, experimentally determined polysaccharide concentrations (Ct, mg·mL⁻¹) at different extraction times were fitted using the linearized second-order kinetic equation (t/C_t versus t). The slope and intercept of the linear regression were used to derive key kinetic parameters: the saturation concentration (C_s), second-order rate constant (k) and initial extraction rate (h).

2.9.2. Second-order kinetic modeling

The extraction of polysaccharides from plant material can often be described accurately using a second-order kinetic model [26], which accounts for the dynamics of liquid–solid mass transfer. In this context, dissolution kinetics of polysaccharides from the biomass into the extraction medium is modeled using Eq. (9).

$$\frac{d{C}_t}{\mathit{dt}}=k{(C_s-C_t)}^2$$ (9)

where Ct (mg/mL) is the polysaccharide concentration at time t (min), Cs (mg/mL) denotes the equilibrium concentration, and k is the second-order rate constant.

2.9.3. Calculation of kinetic parameters (h, k, and C_s)

The parameter h, representing the initial extraction rate (mg·L⁻¹·min⁻¹), corresponds to the rate at the beginning of extraction when C_t = 0 and can be defined as Eq. (10).

$$h=k{C}_s^2$$ (10)

To obtain the kinetic parameters $k$ and Cs, Eq. (9) is integrated under the boundary conditions t = 0 to t = t and C_t = 0 to C_t = C_t, yielding the integrated form Eq. (11).

$$C_t=\frac{C_S^{2}kt}{1+C_{s}kt}$$ (11)

This can also be rearranged into a linearized expression suitable for graphical analysis as Eq. (12).

$$\frac{t}{C_t}=\frac{1}{h}+\frac{t}{C_s}$$ (12)

By plotting t/Ct against t, the slope and intercept provide experimental values of Cs and $k$ allowing the calculation of the initial extraction rate $h$.

2.10. Structural characterization

2.10.1. Scanning electron microscopy

Surface morphology of untreated and treated MD biomass as well as HWE and UA-DES extracted MDPs were analysed by using scanning electron microscopy (SEM). Prior to observation, samples were mounted on stubs and vacuum coated with gold to ensure surface conductivity for SEM observation. Surface morphology was then recorded using a Hitachi SU-8010 microscope (Tokyo, Japan) operated at an accelerating voltage of 3.0 kV.

2.10.2. Fourier transform infrared (FTIR) spectroscopy

The FT-IR spectra for the UA-DES and HWE MDPs were acquired on a Nicolet iS50 spectrometer (Thermo Fisher Scientific, USA). Polysaccharides samples were finely pulverized and thoroughly mixed with dry KBr powder, after which the mixtures were pressed into uniform pellets (approximately 1 mm thick). Spectral data were collected over the wavenumber range of 4000–400 cm⁻¹.

2.10.3. X-ray diffraction

The crystalline properties of the samples were analyzed via X-ray diffraction (XRD) using a Shimadzu XRD-700 instrument with Cu Kα radiation (λ = 1.5406 Å). The system was operated at 40 kV and 40 mA. Data were collected across a 2θ range from 5° to 80° at a scan speed of 4° per minute and a step size of 0.02^◦. A crystallinity index was derived by comparing the integrated intensities of the crystalline diffraction peaks with those of the broad, amorphous features.

2.10.4. Monosaccharide composition analysis

Monosaccharide composition analysis of UA-DES and HWE of MDPs was performed by following a modified procedure as reported by [31]. Hydrolysis was initiated by treating 5 mg samples of the MDPs extracts (UA-DES and HWE) with 1 mL of 2 M trifluoroacetic acid (TFA) at 121 °C for 2 h. After hydrolysis, TFA was removed by evaporating the mixture under nitrogen. The residue was washed with methanol and dried again under nitrogen. This process was repeated 2 to 3 times to ensure complete TFA removal. The purified hydrolysate was then dissolved in 1 mL of deionized water, filtered through a 0.22 μm membrane and prepared for chromatographic analysis.

High-performance anion-exchange chromatography analysis was performed on a Thermo ICS 5000 + system fitted with a Dionex™ CarboPac™ PA-20 column (3 × 150 mm, 10 μm; Thermo Fisher Scientific, USA) with pulsed amperometric detection (HPAEC-PAD). The mobile phase during analysis was ternary gradient of deionized water (A), 0.1 M NaOH (B) and 0.1 M NaOH with 0.2 M sodium acetate (C) at a flow rate of 0.5 mL/min. The gradient elution program of A/B/C (v/v) was set as 95:5:0 at 0 min, 85:5:10 at 26 min, 85:5:10 at 42 min, 60:0:40 at 42.1 min, 60:40:0 at 52 min, 95:5:0 at 52.1 min, and 95:5:0 at 60 min. A 5 μL aliquot was injected for each run. Resulting data were obtained through Dionex ICS 5000 + system and subsequently processed with Chromeleon 7.2 CDS (Thermo Fisher Scientific, USA). MDPs monosaccharide identification was based on retention time compared with standards and quantification was performed using external calibration curves.

2.10.5. Determination of protein, uronic acid and total phenolic contents

Protein content determination was performed by using a commercial bicinchoninic acid (BCA) assay kit (Biosharp Life Sciences, China). Bovine serum albumin (BSA) was used as standard for preparation of calibration curve.

The uronic acid content in the MDPs samples was determined with carbazole sulfuric acid method established by Bitter & Muir [32]. A calibration curve was prepared using D-galacturonic acid standards across a concentration range of 0–100 µg/mL. Sodium tetraborate was dissolved in concentrated sulfuric acid (4.78  mg/mL), and carbazole was prepared in anhydrous ethanol (1.5  mg/mL) and stored in the dark. Aliquots of standard and sample solutions (100  µL) were mixed with 0.5  mL of the sulfuric acid reagent, heated in a boiling water bath for 20  min, cooled and then reacted with 20  µL of carbazole solution at room temperature for 0.5–2 h. After incubation for colour development, 200  µL of each mixture was transferred to a 96-well plate and absorbance was recorded at 530  nm. Experiment was performed in triplicate, and uronic acid content was calculated from the standard curve.

Total phenolic content (TPC) was determined using the Folin–Ciocalteu method by Singleton et al. [33], 20 μL of MDPs solution (1 mg/mL) or gallic acid standard (0–500 μg/mL) was mixed with 75 μL of water, 15 μL of Folin–Ciocalteu reagent and 175 μL of 7.5 % Na₂CO₃ in a 96-well plate. After shaking for 3 min, the plate was incubated in the dark for 30 min at room temperature. Absorbance was measured at 750 nm. TPC was calculated by using gallic acid standard curve and results expressed as mg gallic acid equivalents per gram of MDPs (mg GAE/g).

2.10.6. Molecular weight distribution analysis

The molecular weight distribution of MDPs was assessed by using high performance gel permeation chromatography (HPGPC). Analysis was conducted on a Thermo U3000 HPLC system fitted with a Shimadzu RI-20A refractive index detector. Separation was performed through tandem BRT gel columns (BRT105-103–101, 8 mm × 300 mm; Borui Saccharide Biotech Co., Ltd., China). The mobile phase was 0.2 M NaCl (0.7 mL/min). The temperature of the column was 40 °C and 50 μL of sample was injected. 5 mg/mL of samples and standards concentration were prepared and filtered through a 0.22 μm membrane. Calibration was prepared using dextran standards (P5, P10, P20, P50, P100, P200, P400, P800; Shodex, Japan) along with maltose heptose. The calibration curves were plotted (LgMp-RT: y = −0.2233x + 11.869 (R² = 0.9966); LgMw-RT: y = −0.223x + 11.858 (R² = 0.9963); LgMn-RT: y = -0.2216x + 11.803 (R² = 0.9962)) with logarithm of molecular weight against retention time. The peak molecular weight (Mp), weight average molecular weight (Mw), number average molecular weight (Mn) and the polydispersity index (PDI = Mw/Mn) were calculated.

2.10.7. Thermogravimetric analysis

5 mg sample was applied to a sample plate and measured using a thermogravimetric analyzer (Pyris Diamond TGA, Perkin Elmer, USA). The test conditions were as follows: heating rate of 10 °C/min and temperature range of 25–800 °C. The thermogravimetric analysis (TGA) graph between weight loss (%) vs temperature (°C) was drawn and result was examined.

2.11. Antioxidant activity assays

2.11.1. Determination of DPPH radical scavenging ability

The capacity of UA-DES and HWE extracted MDPs to neutralize DPPH radicals was assessed by an established procedure [34] with minor modifications. Samples were diluted in a series of 10 test tubes and prepared concentrations, ranging from 10 to 100 ug/mL. A volume of 1 mL of freshly prepared 0.1 mmol/L DPPH in ethanol was dispensed into all tubes and vortexed. Mixtures were placed in incubation for 46 min in the absence of light. Absorbance readings were taken at 517 nm. Vitamin C at equivalent concentrations was served as the positive control, while distilled water was used as the blank. The DPPH radical scavenging activity was calculated by using Eq. (13).

$$\mathrm{D}\mathrm{P}\mathrm{P}\mathrm{H}\mathrm{R}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{S}\mathrm{c}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}(\%)=\left[1-\frac{({\mathrm{A}}_1-{\mathrm{A}}_2)}{{\mathrm{A}}_0}\right]\times 100$$ (13)

A₀: Absorbance of the blank (DPPH solution alone), A₁: Absorbance of the reaction mixture (sample + DPPH solution), A₂: Absorbance of the sample baseline (sample in solvent only).

2.11.2. Determination of hydroxyl radical scavenging ability

The hydroxyl radical scavenging capacity of the MDPs was evaluated following the procedure as described in [35]. A series of polysaccharide working solutions were prepared at concentrations of 0.125, 0.25, 0.5, 1, 2, 4, and 6 mg/mL, each with a final volume of 1 mL. For the reaction, 1 mL of each concentration was combined with 1 mL of a freshly prepared FeSO₄ solution (6 mmol/L), 1 mL of H₂O₂ (5 mmol/L), and 1 mL of salicylic acid (20 mmol/L). The combined mixtures were incubated at 37 °C in a thermostatic water bath for 30 min. After incubation the absorbance of each sample was recorded at 510 nm using a UV–visible spectrophotometer with a spectral scan performed from 400 to 600 nm for verification. Distilled water served as the reagent blank, and Vc was used as the positive control. All assays were executed in triplicate. The hydroxyl radical scavenging activity was calculated using Eq. (14).

$$\mathrm{H}\mathrm{y}\mathrm{d}\mathrm{r}\mathrm{o}\mathrm{x}\mathrm{y}\mathrm{l}\mathrm{r}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{s}\mathrm{c}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}(\%)=\left(\frac{{\mathrm{A}}_0-{\mathrm{A}}_1}{{\mathrm{A}}_0}\right)\times 100$$ (14)

where A₀ and A₁ denote the absorbance of the blank and the sample, respectively.

2.11.3. Determination of lipid peroxidation inhibition ability

The ability of the polysaccharides to inhibit the lipid peroxidation was assessed by reported method [36] with minor modifications. A range of concentrated MDPs solutions (0.125, 0.25, 0.5, 1, 2, 4, and 6 mg/mL) was prepared. Aliquots 2 mL of each concentration were placed into separate test tubes and treated with 1.8 mL of a soy lecithin emulsion (1 mg/mL) and 0.4 mL of a freshly prepared FeSO₄ solution (10 mmol/L). The reaction mixtures were incubated at 37 °C in a water bath for 30 min. Following the incubation, the samples were treated with 1 mL of 20 % trichloroacetic acid (TCA) and 1 mL of 0.8 % thiobarbituric acid (TBA). The test tubes were then heated in a boiling water bath for 15 min. After cooling, the mixtures were centrifuged to pellet the particulate matters and the supernatant was clarified. Absorbance readings were taken at 535 nm using a UV–visible spectrophotometer and a full wavelength scan from 400 to 600 nm was performed for validation. Distilled water was used as the reagent blank, and Vc was employed as the reference antioxidant. All experimental runs were performed in triplicate. The inhibitory effect on lipid peroxidation was determined according to Eq. (15).

$$\mathrm{A}\mathrm{n}\mathrm{t}\mathrm{i}-\mathrm{l}\mathrm{i}\mathrm{p}\mathrm{i}\mathrm{d}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{x}\mathrm{i}\mathrm{d}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{c}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}(\%)=\left[\frac{(B_0-B_1)}{B_0}\right]\times 100$$ (15)

In Eq. (15), the terms B₀ and B₁ correspond to the absorbance of the blank control and the sample containing reaction, respectively.

2.11.4. Determination of reducing capacity

Polysaccharide solutions were prepared at the concentrations of 0.125 to 6 mg/mL. To each solution, 1 mL of phosphate buffer (0.25 mol/L, pH 6.6) and 1 mL of potassium ferricyanide (K₃[Fe(CN)₆], 0.4 %) were incorporated. The mixtures were then incubated in a water bath at 54 °C for duration of 20 min. After cooling to room temperature, 1 mL of trichloroacetic acid (4 %) was added and the solutions were centrifuged to isolate the supernatant. In this supernatant, 1 mL of distilled water and 0.2 mL of FeCl₃ solution (0.4 %) were added and the reaction was permitted to proceed at room temperature for 15 min. The final solution was centrifuged to remove any precipitate and the absorbance of the clear supernatant was measured at 700  nm [37].

2.12. Assessment of cellular antioxidant and anti-inflammatory activity in RAW264.7 macrophages

2.12.1. Culture of RAW264.7 cells

Murine macrophage RAW264.7 cells were acquired from the Chinese Academy of Sciences' Cell Bank in Shanghai, China. Cells were grown in Dulbecco's Modified Eagle Medium (DMEM) with 10 % foetal bovine serum (FBS) and 1 % penicillin streptomycin. Cultures were maintained at 37 °C in a humid incubator with 5 % CO₂. Cells in the logarithmic growth phase were utilized for further studies.

2.12.2. CCK-8 cytotoxicity assay

The cytotoxicity of the extracted polysaccharides on RAW264.7 cells was determined with the CCK-8 test kit (Biosharp Life Sciences, Cat. No. BS350B, China). Cell suspensions (8 × 104 cells/mL) were seeded onto 96-well plates (100 μL per well) and incubated for 24 h in a humidified incubator at 37 °C and 5 % CO₂. Then, 10 μL of HWE or UA-DES MDPs at different doses (25, 50, 100, 200, 400, 600, or 800 μg/mL) were added to each well and incubated for another 24 h. Then, 10 μL of CCK-8 reagent was added to each well and the plates were incubated for 4 h in the dark. The optical density (OD) was measured at 450 nm with a microplate reader. The cell viability (%) was determined compared to the untreated blank control group.

2.12.3. Intracellular ROS and NO assessment in LPS-stimulated RAW264.7 cells

RAW264.7 cells were grown in plates and left to adhere overnight. Cells were pretreated with UA-DES or HWE MDPs at concentrations of 50, 100, and 200 μg/mL (designated as low, medium, and high dose groups) for 2 h. Dexamethasone (4 μg/mL) was used as a positive control. After pretreatment, cells were stimulated with 1 μg/mL lipopolysaccharide (LPS) for 24 h, except for the normal control group. The culture supernatant was collected to determine nitric oxide (NO), which was measured using a commercial NO assay kit (Beyotime, S0021S-3, China) according to the manufacturer's instructions. To measure intracellular reactive oxygen species (ROS), activated cells were stained with a commercial DCFH-DA fluorescent ROS test kit (Biosharp Life Sciences, Cat. No. BL714A, China). Intracellular ROS levels were detected with a fluorescence microscope and measured using relative fluorescence intensity.

2.13. Statistical analysis

All statistical analyses were performed through OriginPro 2017 (OriginLab Corporation, Northampton, MA, USA). The experimental outcomes are presented as the mean ± standard deviation (SD) derived from three independent replicates. In graphical presentations, error bars represent the SD values. For single-factor optimization, differences between groups were evaluated using one-way analysis of variance (ANOVA). ANOVA results revealed significant differences (p < 0.05). To identify the differences in groups, Tukey’s post-hoc test was applied. Before executing ANOVA, the data were assessed for normality through the Shapiro–Wilk test and for equal variances via Levene’s test to confirm the suitability of parametric testing. A p-value of less than 0.05 was considered statistically significant.

3. Results and discussion

3.1. Initial screening of various DESs based on MDPs yield

It is well established that the composition of DESs is important parameter that influences the polysaccharides extraction and various bioactive substances from plant sources. DES offers a tunable platform through manipulation of HBAs, HBDs and their molar ratios, polarity, viscosity and acidity for efficient extraction [20]. To identify the most effective DES for extraction of polysaccharides from MD, based on previous studies and preliminary trials, choline chloride (ChCl) was selected as the HBA. ChCl and citric acid are not only effective for polysaccharide extraction but also exhibit high biodegradability and low toxicity compared to conventional organic solvents. In particular, citric acid is highly biodegradable and environmentally sustainable, while choline chloride based DESs show significantly lower toxicity than traditional volatile organic solvents [38], [39]. A total of 25 DESs formulations were prepared by using ChCl, combining it with five different HBDs across molar ratios ranging from 1:1 to 1:5 [29]. These DESs were evaluated for their ability to extract MDPs (Table S1). Results revealed that citric acid based DESs had better MDPs yield than malic acid, lactic acid, and polyol based DESs. DES-3.2 (choline chloride/citric acid, 1:2) provided highest polysaccharide yield of 1.10 % (Fig. 1a).

Fig. 1.

(a) Extraction yield of MDPs with different DES; DES-1 (Choline chloride: DL Malic acid), DES-2 (Choline chloride: Lactic acid), DES-3 (Choline chloride: Citric acid), DES-4 (Choline chloride: Glycerol) and DES-5 (Choline chloride: 1, 4-Butanediol). (b-f) Single-factor optimization of extraction parameters: (b) DES-water content, (c) Liquid-solid ratio, (d) Ultrasonic temperature, (e) Ultrasonic power, (f) Ultrasonic time. Values are mean ± SD (n = 3). Different letters denote significant differences within each panel (one-way ANOVA, Tukey's test, p < 0.05).

The removal of residual DES is crucial for UA-DES extracted polysaccharides in order to ensure MDPs purity. Low molecular weight DES components (choline chloride; 139.62 Da; citric acid: 192.123 Da) are successfully eliminated by extensive dialysis (3000 Da MWCO, 72 h) (Section 2.4). FTIR (Fig. 8a), shows no extra peaks typical of free citric acid or choline chloride. Further, choline chloride is a vitamin B precursor and Generally Recognized as Safe (GRAS) by the FDA, while citric acid is a natural food additive widely used in beverages and pharmaceuticals [40]. MDPs extracted with UA-DES extraction are free of substantial DES contamination and suitable for biological applications.

Fig. 8.

Comparative structural analysis of MDPs, extracted by HWE and UA-DES extraction (a) Fourier-transform infrared (FTIR) spectroscopy profiles; (b) X-ray diffraction (XRD) pattern.

The better efficiency of DES-3.2 may be due to the fact that citric acid is a tricarboxylic acid, which provides more carboxyl groups (Hydrogen Bond Donors) compared to malic (dicarboxylic) or lactic (monocarboxylic) acid. This leads to a stronger interaction with the plant cell wall matrix and enhanced the extraction of MDPs yield [41], [42]. Under similar ultrasound-assisted conditions (section 2.4), water (UA-WE) was employed as a conventional solvent to assess the relative performance of DES. The yield of MDPs with UA-WE was 1.01 %, which is 8.18 % less than the yield with UA-DES (1.10 %). This demonstrates the functional advantage of DES in promotion of MDPs solubilisation. Significantly, both ultrasonic techniques performed better than traditional HWE (0.91 %), with UA-DES extracted 20.87 % more and UA-WE extracted 10.99 % more than the HWE. Extraction comparison of UA-DES with UA-WE indicates that the increase in yield of MDPs is due to DESs performance rather than ultrasound assisted process conditions, underscore the potential of DES-3.2 as effective solvent for extraction.

3.2. Single-factor experimental analysis

The effects of key extraction parameters on MDPs yield were first evaluated through single factor experiments. The factors selected for this preliminary screening were DES water content (%), liquid–solid ratio (mL/g), ultrasonic temperature (^◦C), ultrasonic power (W), and ultrasonic time (min). These studies were carried out by altering one factor while leaving the other four factors constant. This technique provides a full understanding of each variable mechanistic impacts, as well as preliminary suggestions for multi-factor optimization. DES water content controls viscosity, polarity, and hydrogen-bond strength of DESs. The increase in water content from 20 % to 30 % enhanced the MDPs yield from 0.95 % to 1.37 % (Fig. 1b), likely due to improved solvent diffusion and mass transfer resulting from reduced viscosity. Due to viscosity and polarity, which might impair the solubilization effectiveness, higher water content (40 to 60 %) slightly decreased the yield of MDPs (1.30 to 1.01 %) [43]. These findings support earlier findings that DES performance for polysaccharide extraction is maximized by modest water input [29]. Regarding the liquid–solid ratio, the maximum MDPs yield (1.36 %) was observed at 30 mL/g (Fig. 1c). Lower liquid–solid ratio (10–20 mL/g) reduced solvent penetration, whereas greater liquid–solid ratio (40–50 mL/g) caused to a decline in MDPs yield. This variation may be directly related to two important variations. Initially, the greater proportion of solvent increased the osmotic pressure, which accelerated the diffusion of intracellular components. Additionally, the higher solvent volume extended the effective region for cavitation, which resulted in significantly improvement of ultrasonic assisted MDPs extraction efficiency [44]. Beyond the optimal liquid–solid ratio of 30 mL/g, the excessive dilution reduces the intermolecular collisions and attenuates the transmission of ultrasonic energy through the medium. The extraction temperature (60 °C) exhibited a considerable favorable effect (Fig. 1d), resulting in MDPs yield of 1.38 %. Temperatures below 50 °C appeared insufficient for cell wall breakdown. Rising temperatures didn't improve the extraction considerably, revealing that a balanced control of temperature and sonication energy is required to break down the cell walls without destroying the polysaccharides [26]. Similarly, ultrasonic power (360 W) yields maximum MDPs of 1.36 % (Fig. 1e). The lower power (200–280 W) resulted in insufficient cavitation. Extreme ultrasound power of 440–520 W slightly decreased the MDPs yield due to mechanical degradation of polysaccharides from high shear forces [45]. This highlights that balanced ultrasonic intensity is important for efficient extraction. Optimal ultrasonic time of 90 min yielded 1.36 % of MDPs (Fig. 1f). Shorter extraction time of 30–60 min was insufficient for complete recovery of MDPs. Conversely, prolonged ultrasonic time of 120–150 min slight degrade the polysaccharide and reduced the MDPs yield of UA-DES method [29]. These findings suggest that MDPs recovery is influenced by the combined effect of solvent chemistry and ultrasonic extraction parameters. DES-3.2 created favorable hydrogen bonding, whereas ultrasonic promotes the cell wall breakdown and mass transfer, resulting in higher MDPs yield. Based on single factor findings, preliminary optimum conditions for multi factor RSM optimization was determined as 30 % DES water content, 30 mL/g liquid–solid ratio, 60 °C ultrasonic temperature, 360 W ultrasonic power, and 90 min of sonication duration. These conditions provide a rational starting point for exploring the interactions of factors and predicting the optimum MDPs yield.

3.3. BBD–RSM optimization of MDPs extraction

3.3.1. Experimental design and model development

After the selection of DES-3.2 as the optimal extraction solvent with DES water content fixed at 30 %, A Box-Behnken design (BBD) was implemented to systematically evaluate the effects of ultrasonic extraction parameters on MDPs yield. Based on preliminary single factor tests, four independent variables were chosen, liquid–solid ratio (A), ultrasonic temperature (B), ultrasonic power (C), and ultrasonic time (D). Each factor was evaluated at three coded levels (−1, 0, +1), as shown in Table S3. In total 29 experiments, containing five replicate runs at the central point, were performed to ensure accurate estimation of experimental error and model robustness (Table 1). The experimentally obtained yield of MDPs varied from 0.79 % to 1.43 %, indicating that MDPs yield was highly sensitive to changes in ultrasonic and solvent related parameters within the investigated domain. However, relatively narrow yield range suggests that the selected DES system operates efficiently under ultrasound assistance, while further optimization is governed by complex interactions among the process variables rather than drastic parameter changes. The influence of the four independent variables on polysaccharide yield was modelled through a second-order polynomial regression (Eq. (16). This fitted quadratic model expressed in coded units is presented below.

$$\mathrm{Y}\mathrm{i}\mathrm{e}\mathrm{l}\mathrm{d}(\%)=1.36+0.1863A+0.0408B+0.1308C+0.0671D-0.0049AB-0.0375AC+0.0463AD-0.06BC-0.0225BD+0.11CD-0.1033A^2-0.1602B^2-0.0527C^2-0.2158D^2$$ (16)

Table 1.

Experimental design for independent variables and response MDPs yield.

No.	A	B	C	D	Polysaccharide Yield (%)
					Actual	RSM		ANN
						Predicted	Residual	Predicted	Residual
1	20	60	360	60	0.79	0.83	−0.04	0.80	−0.01
2	30	70	280	90	1.09	1.12	−0.03	1.08	0.01
3	30	50	440	90	1.33	1.30	0.03	1.33	0.00
4	20	60	440	90	1.20	1.19	0.01	1.21	−0.01
5	40	60	440	90	1.43	1.48	−0.05	1.37	0.06
6	30	60	440	60	1.06	1.05	0.01	1.07	−0.01
7	30	70	440	90	1.24	1.26	−0.02	1.24	0.00
8	30	50	360	60	0.85	0.85	−0.00	0.88	−0.03
9	30	50	360	120	0.96	1.03	−0.07	1.01	−0.05
10	30	60	360	90	1.35	1.36	−0.01	1.35	0.00
11	30	60	360	90	1.39	1.36	0.03	1.35	0.04
12	40	50	360	90	1.28	1.25	0.03	1.27	0.01
13	20	70	360	90	0.94	0.95	−0.01	1.01	−0.07
14	30	60	280	120	0.92	0.92	0.00	0.92	0.00
15	20	60	280	90	0.88	0.85	0.03	0.86	0.02
16	40	60	280	90	1.26	1.30	−0.04	1.23	0.03
17	30	50	280	90	0.94	0.91	0.03	0.95	−0.01
18	40	70	360	90	1.35	1.32	0.03	1.35	0.00
19	40	60	360	60	1.09	1.11	−0.02	1.10	−0.01
20	30	60	280	60	1.01	1.00	0.01	1.01	0.00
21	40	60	360	120	1.39	1.34	0.05	1.36	0.03
22	30	60	360	90	1.37	1.36	0.01	1.35	0.02
23	30	70	360	120	1.05	1.07	−0.02	1.04	0.01
24	30	60	440	120	1.41	1.40	0.01	1.40	0.01
25	30	60	360	90	1.33	1.36	−0.03	1.35	−0.02
26	30	60	360	90	1.36	1.36	0.00	1.35	0.01
27	30	70	360	60	1.03	0.98	0.05	1.03	0.00
28	20	50	360	90	0.85	0.86	−0.01	0.91	−0.06
29	20	60	360	120	0.91	0.88	0.03	0.90	0.01

A: Liquid-Solid Ratio (mL/g), B: Ultrasonic Temperature (°C), C: Ultrasonic Power (W), D: Ultrasonic Time (min).

Here, A, B, C, and D represent the coded values for the liquid–solid ratio, ultrasonic temperature, ultrasonic power, and ultrasonic time, respectively.

3.3.2. Model fitting and ANOVA

ANOVA results demonstrated the high statistical significance of the quadratic model (P < 0.0001), with a substantial F-value of 43.92, demonstrating that the regression effectively captured the experimental trends (Table 2). The models predictive performance was validated by the coefficient of determination (R² = 0.9777), which showed good agreement between expected and observed responses.

Table 2.

Regression analysis of polysaccharides extraction model and regression coefficients.

Source of variance	Sum of Squares	df	Mean Square	F-value	P-value	Significant
Model	1.1871	14	0.0848	43.92	< 0.0001**	Significant
A- Liquid-Solid Ratio	0.4163	1	0.4163	215.63	< 0.0001**
B- Ultrasonic Temperature	0.0200	1	0.0200	10.36	0.0062*
C-Ultrasonic Power	0.2054	1	0.2054	106.40	< 0.0001**
D-Ultrasonic Time	0.0540	1	0.0540	27.97	0.0001**
AB	0.0001	1	0.0001	0.0518	0.8233 ns
AC	0.0056	1	0.0056	2.91	0.1099 ns
AD	0.0086	1	0.0086	4.43	0.0538 ns
BC	0.0144	1	0.0144	7.46	0.0162*
BD	0.0020	1	0.0020	1.05	0.3231 ns
CD	0.0484	1	0.0484	25.07	0.0002**
A2	0.0693	1	0.0693	35.88	< 0.0001**
B2	0.1665	1	0.1665	86.24	< 0.0001**
C2	0.0180	1	0.0180	9.33	0.0086*
D2	0.3022	1	0.3022	156.52	< 0.0001**
Residual	0.0270	14	0.0019
Lack of Fit	0.0250	10	0.0025	5.01	0.0672	not significant
Pure Error	0.0020	4	0.0005
Cor Total	1.214	28
R2 = 0.9776,	R2Adj = 0.9555,	R2Pre = 0.8787
C.V. % = 3.85	Adeq Precision = 20.5633

ns: not significant (P > 0.05).

^*indicates significant (P < 0.05).

^**Indicates highly significant (P < 0.001).

The modified R² (0.9555) and predicted R² (0.8787) indicate the models robustness and predictive capacity. The strong agreement between adjusted and anticipated R² (difference < 0.2) confirms the models predictive reliability. Furthermore, the high adequate precision rate (20.56) suggested an acceptable signal-to-noise ratio. The lack of fit test was non-significant (P = 0.0672), indicating that the model error was mostly due to random experimental variability rather than model inadequacies. The liquid–solid ratio (A) had the most significant impact on polysaccharide yield (F = 215.63, P < 0.0001), followed by ultrasonic power (C) and ultrasonic time (D). The quadratic factors D² and B² were extremely significant, exhibited strong curvature effects and optimal operating regions rather than monotonic trends. Notably, the interaction of ultrasonic power and ultrasonic time (CD) demonstrated a considerable synergistic effect (P = 0.0002), emphasizing the relevance of collective ultrasonic energy input in determining the MDPs yield. These findings are align with earlier studies on UAE of polysaccharides [46], where RSM revealed that the liquid–solid ratio, ultrasonic time, and ultrasonic power significantly influence the polysaccharides yield and their interactions played a critical role in defining the optimal conditions for maximum recovery of polysaccharides.

3.3.3. Response surface methodology (RSM) analysis

Three-dimensional response surface plots and related contour diagrams (Fig. 2 and S1) show how each factor and their interactions affect the MDPs extraction. The liquid–solid ratio (A) significantly increased the polysaccharide extraction up to an optimal level. Beyond which the yield plateaued or slightly dropped (Fig. 2a, b). This behavior indicates that mass transfer and solubilisation improved at larger liquid–solid ratio. But excessive dilution might lower the effective cavitation intensity per unit mass [47]. Ultrasonic temperature (B) had a significant quadratic effect on optimising MDPs extraction. Its interaction with the liquid–solid ratio, ultrasonic power, and ultrasonic time (Fig. 2a, d, e) peaked at about 60 °C, revealing a significant thermal window for MDPs yield optimization. Excessive heat most likely causes polysaccharide breakdown or affects the solvent characteristics [48]. Ultrasonic power (C) increased MDPs yield linearly from 280 to 440 W, with a minor reduction at the maximum power (Fig. 2b, d). This indicates an upper limit for effective power application, which should be optimized rather than maximized within this range. Similarly, ultrasonic time (D) showed a trend of declining results by plateauing about 90 min (Fig. 2c, e, f). The large curvature for both parameters shows that the benefits of greater cavitation and longer extraction are eventually counteracted by accumulative shear and possible oxidative damage [49]. Power and interaction (CD) time shows that overall acoustic energy, rather than individual parameters is important aspect in extraction yield of MDPs. Additionally, the yield plateau found at high liquid–solid ratios under maximum ultrasonic power demonstrates the critical sensitivity in balancing mechanical energy and solvent capacity. Overall, the response surface analysis determines the operating boundaries for optimum MDPs extraction. To get optimum yield of MDPs, an appropriate volume of liquid–solid ratio with the temperature around 60 °C should be used to enhance mass transfer. While ultrasonic power should be enough to disrupt cell walls without degradation of MDPs, ultrasonic time is limited to 90 min to avoid any adverse effects caused by cavitation. This arrangement is ideal for maximizing mass transfer while minimizing damage caused by unwanted heat and energy.

Fig. 2.

(a-f). Three-dimensional response surface and contour plots showing interactions of factors influencing MDPs extraction.

3.3.4. Verification of RSM predictive model

The regression model was used to establish the best extraction parameters and to assess the RSM models accuracy and prediction performance. MDPs yield was 1.36 % using a liquid–solid ratio of 30 mL/g, ultrasonic temperature of 60 °C, ultrasonic power of 360 W, and ultrasonic time of 90 mins. Experimental run to verify these optimum extraction conditions resulted in a polysaccharide recovery of 1.38 %, which was close to the model predicted value. The difference between predicted and experimental findings was 0.02 %, indicating the excellent agreement and supports the robustness and reliability of the model. A comparison of the optimal MDPs yield with previously reported yields of polysaccharides extracted from different plant materials using UA-DES methods is presented in Table S6. This verification indicates that the RSM model can accurately estimate the MDPs yield within the experimental domain and give a solid foundation for process optimization. The discovered optimum balance between ultrasonic power and ultrasonic time ensures maximal polysaccharide release without degradation caused by excessive sonication.

3.4. ANN model prediction and performance analysis

Architecture of the neural network model used for MDPs yield prediction consists of 4 input variables, 10 hidden processing layers, and 1 output response layer (Fig. S2). ANN calibrated by using the experimental dataset generated from the RSM design, showed strong predictive performance for MDPs yield extraction. Performance metrics (Table 3) indicate the ANN model attained a MSE of 0.00073 and a RMSE of 0.02695. Additional accuracy indicators, including a MAD of 0.01851, a MAPE of 1.719 %, and R² of 0.9827, collectively affirmed the models reliability. The model capability was further validated by correlation coefficients for individual data subsets, R² of 0.9925 for training, 0.9892 for validation, and 0.9985 for testing. Culminating in an overall R² of 0.9862 (Fig. 3) confirmed a close agreement between observed and predicted values.

Table 3.

Performance comparison of RSM and ANN models in predicting MDPs yield.

Metric	RSM Model	ANN Model
R2	0.9791	0.9827
MSE	0.00088	0.00073
RMSE	0.02961	0.02695
MAD	0.02431	0.01851
MAPE (%)	2.185	1.719

Fig. 3.

Performance evaluation of the ANN model using the Levenberg–Marquardt (TRAINLM) algorithm. Regression plots illustrate the correlation between predicted and actual yield of MDPs for the (a) training, (b) validation, (c) testing, and (d) complete datasets.

Analysis of the learning progression (Fig. 4a, c) revealed the concurrent decreases in training and validation errors prior to convergence with best validation performance (MSE = 0.001692) reported at epoch 5. Training was stopped to prevent model overfitting. The convergence was confirmed by a final gradient value of 2.063 × 10⁻¹⁰. The error histogram (Fig. 4b) showed that prediction error approaches to zero with a negligible skew. Extraction results revealed that 86.2 % of the residuals were within ± 0.03 yield percentage point interval, indicating the robustness of model. The ANN model predicted a MDPs yield of 1.35 % (Table S4) under optimal extraction parameters (liquid–solid Ratio 30 mL/g, ultrasonic temperature 60 °C, ultrasonic power 360 W, ultrasonic time 90 min). This prediction showed excellent concordance with the actual experimental yield of 1.38 %, verifying the model predictive accuracy. Compared with RSM model, ANN model offered enhanced resolution in the prediction that aligned more precisely with the experimental data. As detailed in Table 3, a direct model comparison revealed the ANN enhanced performance.

Fig. 4.

Comprehensive evaluation of the Levenberg–Marquardt (TRAINLM) neural network: (a) Mean squared error trends for training, validation, and testing subsets showing optimal validation performance; (b) Histogram of prediction errors. (c) Evolution of training parameters (gradient magnitude, mu adaptation, validation failures) during iterative learning. (d) Experimental run-wise comparison of measured polysaccharide yield with RSM and ANN model predictions.

The ANN model attained a higher R² = 0.9827 and consistently lower errors across all metrics RMSE, MAD, and MAPE. Both RSM and ANN models validated statistical strength in their respective frameworks (Fig. 5a, b). ANN exhibited a better ability to model the nonlinear interactions among extraction parameters and predict the extraction yield of MDPs. While RSM model showed excellent statistical capability (R² = 0.9776, Adjusted R² = 0.9555), endorsing its ability for process optimization. Over all, RSM and ANN are complementary methodologies. RSM enables statistical evaluation and mechanistic insight into effects of parameters, while ANN improves the predictive performance for complex and nonlinear extraction processes.

Fig. 5.

Comparative evaluation of RSM and ANN models for predicting the extraction yield of MDPs (a) Correlation between experimental yields and RSM predictions. (b) Correlation between experimental yields and ANN predictions.

3.4.1. Validation of ANN model predictions

The ANN model performed well, with R² = 0.989 values surpassed for all data subsets (Fig. 3). ANN predictions were more closely matched with experimental values than those from the RSM model (Fig. 4d), a conclusion supported by residual data (Table 1). The error distribution displayed in the histogram (Fig. 4b), approximated a normal distribution cantered on zero, demonstrating no systematic bias in the predictions. The RSM model constrained by its second order polynomial structure produced response surfaces through its mathematical form. Whereas, ANN model outperformed the RSM model not only in MDPs yield prediction accuracy but also in its ability to reveal the complex nature of parameter interactions through its advanced fitting algorithm [50]. ANN-generated surfaces were created directly from data allowing them to capture intricate nonlinear system behaviors that may be inherent in the extraction process. This suggests that the ANN not only improved prediction accuracy [26] but also provided a more realistic depiction of the mechanisms that underlie dynamics.

3.5. Kinetic model for ultrasonic extraction of MDPs

A second-order kinetic model was used to investigate the extraction behaviour of MDPs under optimum US-DES conditions. Table 4 shows the polysaccharide concentrations (Ct) at different extraction times. A clear linear relationship between t/Ct and extraction time (t) was revealed that the second-order model is suitable for MDPs yield extraction. Linear regression yielded Eq. (17) with a coefficient of determination R² = 0.9590, indicating a strong correlation between experimental observations and model results.

$$\frac{t}{C_t}=267.9+8.867\mathrm{t}$$ (17)

Table 4.

Comparison of experimental and predicted MDPs concentrations (C_t) during extraction, second-order kinetic model and ANN predictions.

Time (min)	Actual Ct (mg·mL⁻1)	t/Ct	Equation Predicted Ct (mg·mL⁻1)	Equation Residual	ANN Predicted Ct (mg·mL⁻1)	ANN Residual
5	0.021	238.10	0.016	−0.005	0.017	−0.004
10	0.026	384.62	0.028	0.002	0.028	0.002
15	0.033	454.55	0.037	0.004	0.033	0.000
20	0.040	500.00	0.045	0.005	0.041	0.001
25	0.046	543.48	0.051	0.005	0.045	−0.001
30	0.052	576.92	0.056	0.004	0.051	−0.001
35	0.057	614.04	0.061	0.004	0.058	0.001
40	0.062	645.16	0.064	0.002	0.060	−0.002
45	0.063	714.29	0.068	0.005	0.065	0.002
50	0.071	704.23	0.070	−0.001	0.069	−0.002
55	0.076	723.68	0.073	−0.003	0.075	−0.001
60	0.079	759.49	0.075	−0.004	0.079	0.000
65	0.083	783.13	0.077	−0.006	0.081	−0.002
70	0.086	813.95	0.079	−0.007	0.086	0.000
75	0.089	842.70	0.080	−0.009	0.089	0.000
80	0.089	898.88	0.082	−0.007	0.090	0.001
85	0.091	934.07	0.083	−0.008	0.091	0.000
90	0.091	989.01	0.084	−0.007	0.090	−0.001
95	0.088	1080.68	0.086	−0.002	0.089	0.001
100	0.086	1162.79	0.087	0.001	0.087	0.001
105	0.089	1179.78	0.088	−0.001	0.087	−0.002
110	0.085	1294.12	0.089	0.004	0.085	0.000
115	0.083	1385.54	0.089	0.006	0.084	0.001
120	0.081	1481.48	0.090	0.009	0.081	0.000

The second-order model fits the extraction data well because it inherently captures the biphasic nature of the UA-DES process, an initial rapid desorption phase driven by ultrasonic cavitation that disrupts cell walls and releases surface polysaccharides, followed by a slower diffusion-controlled phase in which polysaccharides migrate from the dense matrix interior. This dual desorption diffusion mechanism is consistent with the second-order model premise that the extraction rate is proportional to the square of the concentration driving force [51]. Slope and intercept of Eq. (17) were used to calculate kinetic parameters. The saturation concentration (Cs) was determined to be 0.1128 mg/mL and the initial extraction rate (h) and second-order rate constant (k) were estimated to be 0.00375 mg/mL·min⁻¹ and 0.2945 mL·mg⁻¹·min⁻¹, respectively. While the DES system encourages strong hydrogen-bond interactions with polysaccharides, ultrasonic cavitation is responsible for the high initial extraction rate by disrupting cell wall structures and increasing penetration of solvent [20]. As extraction proceeds, the rate gradually decreases due to reduction in readily accessible polysaccharides and increased internal diffusion resistance, eventually equilibrium approached as reflected by the saturation concentration (Cs). Similar diffusion controlled kinetic behaviour has been reported for ultrasound-assisted DES extraction of plant polysaccharides from Cercis chinensis bark [52].

3.5.1. Comparison of kinetic and ANN model predictions

To further evaluate predictive capability, the second-order kinetic model was compared with an ANN model for prediction of MDPs concentration (Ct) at different times of extraction. Consistent with the methodologies detailed in sections 2.7 and Table S5, the models were applied for ANN. Results in Table 4 show that accurate predictions with minimal error were achieved by the ANN. Although the kinetic model adequately captured the overall extraction trend (R² = 0.9440), the ANN model exhibited significantly superior performance (Fig. S3), achieved a higher coefficient of determination (R² = 0.9958) and obviously lower prediction errors. ANN predictions lowered the MAE, RMSE, and MAPE by 76 %, 73 %, and 69 %, respectively (Table 5), which indicates ANN model accuracy and reliability compared to the kinetic model.

Table 5.

Comparison of RSM and ANN model performance for predicting MDPs concentration (Ct) over time.

Metric	Kinetic Model Predicted	ANN Model Predicted
MAE (mg/mL)	0.0046	0.0011
RMSE (mg/mL)	0.0052	0.0014
MAPE (%)	7.56	2.36
Prediction R2	0.9440	0.9958

The excellent predictive performance of the ANN model derives from its intrinsic capacity to capture complex nonlinear relationships driven by cavitation effects, liquid–solid interactions and time dependent diffusion limitations. These phenomena are difficult to accurately represent using simplified mechanistic kinetic assumptions [26]. Importantly, the ANN accurately predicted Ct values across all extraction stages, including the rapid initial release and near equilibrium plateau regions. Overall, the second-order kinetic model provides valuable mechanistic insight into the US–DES extraction behavior of MDPs. The ANN model demonstrated superior predictive precision. The combined use of mechanistic kinetic modeling and ANN-based prediction provides a comprehensive framework for prediction and optimization of UA-DES extraction processes.

3.6. Evaluation of DES reusability

The use of DESs in extraction of polysaccharides and associated bioactive constituents is preferred due to their recyclability, low toxicity, and low vapour pressure. The recovery rate of DES evaluates both efficiency and sustainability of the extraction process. The practical reusability of DES-3.2 was confirmed through five extraction cycles (Fig. 6), under optimized extraction conditions (Table S4). A slight decline in the extraction yield of MDPs was observed from 86.44 % to 80.66 % and recovery rate of DES-3.2 from 97.53 % to 91.25 % confirms the good performance of solvents and their high stability in recycling. The observed attenuation in performance is probably a consequence of hydrogen bond degradation within the solvent upon repeated cycling. Previous studies highlight the strong recyclability of DESs in polysaccharide extraction. Tang et al. [53] reported that a DES prepared of tetracaine and lauric acid provided high recovery of Lycium barbarum polysaccharides with yields slightly decrease from 82.5 % to 80.2 % over five consecutive cycles. Similarly, Wang et al. [54] reported negligible loss in polysaccharides yield for a choline chloride and urea based DES used on Radix Bupleuri, with yields of 5.42 % and 5.81 % over five cycles [55]. Such findings highlight the reliability of DESs as reusable extraction solvent.

Fig. 6.

Evaluation of DES-3.2 recyclability over consecutive extraction cycles.

The optimized absolute yield of MDPs is 1.38 %, which is a 51.65 % improvement over the 0.91 % yield of conventional HWE, showcasing the efficiency of the UA-DES extraction. Additionally, DES-3.2 demonstrates excellent recyclability, maintaining consistent extraction performance over five cycles, which significantly reduces solvent costs and waste. While this method may not be suitable for large scale industrial production of bulk polysaccharides, it offers a viable and sustainable approach for extracting high value bioactive polysaccharides from plants, particularly when environmental impact and solvent reusability are prioritized.

3.7. Structural characterization of MDPs

3.7.1. SEM analysis

Structural changes induced by HWE and UA-DES extraction on surface morphology of untreated and treated MD biomass as well as the extracted MDPs was examined by using SEM. The untreated MD biomass powder exhibited a dense, clear microstructure with smooth, intact cell walls and a continuous epidermal layer (Fig. 7a), which forms a natural barrier for polysaccharide dissolution. HWE residue showed a compromised surface morphology with significant wrinkling and roughness (Fig. 7b), with retained structural cohesion. UA-DES process induced more severe roughness and a wrinkled morphology (Fig. 7c) of MD biomass, indicating a greater degree of cell wall disruption. This deformation is credited to ultrasonic acoustic cavitation [56] and ability of DES to efficiently degrade the cell wall structure [20]. Modified biomass surface promoted polysaccharide mass transfer into the solvent and increased the yield of MDPs. These findings are consistent with recent investigation, in which UAE of paclitaxel generates a residue with surface roughness, wrinkles and structural rupture [25]. While UA-DES extraction of polyphenols from Rhamnus alaternus creates a more porous structure with wrinkled and destroyed cell walls [57]. Regarding the comparison of HWE and UA-DES derived MDPs. HWE resulted in the creation of aggregated, compact particles with smooth, non-porous surfaces (Fig. 7d), indicating the minimal disruption of the cell walls. In contrast, the UA-DES extraction produced fragmented and dispersed particles with rough, porous surfaces (Fig. 7e). The porous nature of these particles is demonstrated by clearly visible uniform pores and an enhanced surface area. This is believed to boost solubility and improve functional characteristics of isolated MDPs. These findings are comparable with those of Imran et al. [20], who reported Okra polysaccharides extracted through UA-DES, exhibited significant porosity and rough surfaces, while the products obtained through HWU extraction remained compact and smooth. Collectively, these morphological changes clearly suggest that UA-DES treatment generates higher cell walls disruption and porosity than HWE, thereby permitting improved mass transfer and contributing to better extraction yield of MDPs.

Fig. 7.

Scanning electron microscopy (SEM) images showing the surface morphology of MD biomass and extracted polysaccharides under different treatments. (a) Untreated MD biomass exhibiting a dense, clear microstructure with smooth, intact cell walls and a continuous epidermal layer; (b) HWE-treated residue showing moderate wrinkling and roughness (partial cell walls disruption); (c) UA-DES-treated residue exhibiting highly rough and wrinkled morphology (cell wall rupture). (d) Polysaccharides extracted via HWE showing aggregated and compact particles with smooth, non-porous surfaces; (e) polysaccharides extracted via UA-DES exhibiting fragmented, dispersed particles with rough surfaces and pores formation. Images (a–c) were recorded at 3000× (scale bar = 10 µm), whereas (d–e) were obtained at 300×, 3000×, and 8000× (scale bars = 100, 10, and 5 µm, respectively).

3.7.2. FTIR analysis

The FTIR spectra of MDPs extracted by UA-DES and HWE are given in Fig. 8a. Both spectra exhibited comparable absorption patterns indicating that the polysaccharide backbone remained independent of the extraction procedure. A broad and intense absorption band observed at 3436.98 cm⁻¹ for the UA-DES extracted MDPs and 3433.35 cm⁻¹ for the HWE sample was assigned to O–H stretching vibrations for abundance of hydroxyl groups.

In addition, peaks located at 2934.90 cm⁻¹ (UA-DES) and 2934.80 cm⁻¹ (HWE) resembled to C–H stretching vibrations of CH₂ and CH₃ groups [58]. Absorption bands at 1742.12 and 1620.95 cm⁻¹ in the HWE sample and at 1746.47 and 1634.74 cm⁻¹ in the UA-DES extract were assigned to esterified carboxyl and carboxylate stretching vibrations, respectively, suggesting the existence of uronic acids [59]. The absorption bands at 1369.19 and 1231.34 cm⁻¹ for HWE and at 1375.72 and 1241.50 cm⁻¹ for the UA-DES extract were associated with C–H bending and O–H deformation vibrations. Which are commonly reported in rhamnose containing polysaccharides [60]. UA-DES exhibited peaks at 1101.47 and 1015.13  cm⁻¹, corresponding to C–O–H bending and glycosidic linkages, while HWE showed a single peak at 1053.58  cm⁻¹ [61]. The difference was the appearance of a 960 cm⁻¹ band solely in UA-DES extracts, indicative of polysaccharides with specific β-type or α-(1 → 6) linkages [62], this feature was absent in HWE extracts confirming the unique selectivity of the UA-DES process.

3.7.3. XRD analysis

X-ray diffraction is an essential technique used to evaluate the crystallinity of biological macromolecules. Amorphous molecules reveal multiple diffraction bands and crystalline solids produce sharp distinct peaks [63]. The XRD pattern of UA-DES and HWE extracted MDPs is illustrated in Fig. 8b. The diffused peak at 22° (2θ) possesses wide characteristics and lacks discrete crystalline peaks, demonstrating the poor crystallinity and amorphous nature of MDPs. The amorphous character seen in MDPs is consistent with the findings reported for Ipomoea batatas polysaccharides by Chenfeng et al. [64] and Ziziphus jujuba by Xiaolong et al. [65]. The XRD analysis confirms that both extraction methods maintained the low crystallinity and amorphous structure intrinsic to MDPs.

3.7.4. Monosaccharide composition analysis

Monosaccharide composition of MDPs extracted through UA-DES extraction and HWE was determined by chromatographic analysis (Table 6), which is consistent with previously reported monosaccharide composition profile for MDPs [7]. Both samples primarily consisted of similar monosaccharide contents (Fig. 9), although their relative distributions differed significantly in both extraction methods. The molar ratio of UA-DES extracted MDPs monosaccharide composition was dominated by galactose (39.70 %), galacturonic acid (22.87 %), glucose (11.70 %), rhamnose (9.02 %), and arabinose (7.59 %), with minor amounts of mannose (3.22 %), xylose (2.95 %), fucose (1.38 %), and glucuronic acid (1.56 %). In comparison, monosaccharide of HWE exhibited a relatively different composition, mainly comprised of galactose (26.05 %), glucose (22.56 %), rhamnose (14.64 %), arabinose (11.35 %), mannose (7.28 %), and galacturonic acid (10.40 %), with smaller proportions of fucose (2.34 %), xylose (2.66 %), and glucuronic acid (2.73 %).

Table 6.

Monosaccharide composition (%) molar ratios of MDPs extracted through HWE and UA-DES methods.

Monosaccharide	HWE	UA-DES
Fucose	2.34	1.38
Rhamnose	14.64	9.02
Arabinose	11.35	7.59
Galactose	26.05	39.70
Glucose	22.56	11.70
Xylose	2.66	2.95
Mannose	7.28	3.22
Galacturonic acid	10.40	22.87
Glucuronic acid	2.73	1.56

Fig. 9.

HPLC profile of monosaccharide composition of MDPs, extracted through HWE and UA-DES extraction. Peaks are labelled according to monosaccharide identity, based on retention time matching with the standard (STD) monosaccharides.

Notably, the UA-DES polysaccharides contained higher content of galacturonic acid compared with HWE, suggesting that the UA-DES system preferentially enriched acidic polysaccharide fractions [66]. This is likely due to the protons provided by citric acid forming a specific hydrogen-bonding network with the carboxyl groups (−COOH) of galacturonic acid residues, leading to increased solubility and stability in the DES system compared to neutral polysaccharides [67]. Higher concentrations of galacturonic acid and galactose content play a significant role in enhancing the scavenging ability of the polysaccharides against ABTS and DPPH radicals, which is aligned with earlier findings on antioxidant properties of polysaccharides [68]. The coexistence of rhamnose, arabinose, galactose, and galacturonic acid contents indicates that MDPs may contain pectic polysaccharide domains, particularly homogalacturonan (HG) and rhamnogalacturonan-I (RG-I) structures [69]. The compositional differences between UA-DES and HWE can be attributed to the synergistic effects of ultrasonic cavitation and the strong hydrogen bonding capability of DES, which facilitate disruption of the plant cell wall matrix and enhance the solubilization of tightly bound acidic polysaccharides. This selective enrichment of uronic acid containing components may contribute to the improved physicochemical and antioxidant properties observed for the UA-DES extracted MDPs.

3.7.5. Thermogravimetric analysis (TGA)

Thermogravimetric analysis (TGA) serves as a critical analytical technique for characterizing the thermal stability and compositional profile of plant derived polysaccharides. The thermal degradation behaviors of HWE, UA-DES extracted MDPs and MD biomass are presented in Fig. 10a, exhibited a typical multi-stage degradation pattern characteristic of plant polysaccharides. The first weight loss below 120 °C was 15 %, 13 % and 6 % for HWE, UA-DES and MD biomass respectively, mainly attributed to evaporation of absorbed and bound water molecules [70]. The second major degradation stage total weight loss was 60 %, 67 % and 59 % of HWE, UA-DES MDPs and MD biomass respectively, at 180 to 400 °C corresponding to cleavage of glycosidic bonds, degradation of polysaccharide chains and volatilization of low molecular weight fragments, which is consistent with previous reports of natural polysaccharides from Cordyceps militaris [71]. UA-DES extracted MDPs exhibited approximately 7 % greater weight loss as compared to.

Fig. 10.

(a) Thermogravimetric analysis (TGA) curves showing weight loss as a function of temperature. (b) HPGPC chromatograms showing molecular weight distribution of MDPs from HWE (RT = 31.25 min, Mw = 77.4 kDa) and UA-DES (RT = 27.40 min Mw = 560.1 kDa).

HWE of MDPs that might be due to its higher contents of galacturonic acid and structural modification induced by UA-DES extraction. At third stage about 450 °C, all samples showed gradual mass reduction associated with carbonization and char formation. At 800 °C total weight losses reached 74 %, 81 % and 81 % for HWE, UA-DES MDPs and MD biomass respectively, comparable to thermal degradation behavior reported for polysaccharides from Ganoderma lucidum [72]. The MD biomass displayed a broader degradation profile due to the coexistence of hemicellulose, cellulose and lignin components. TGA results demonstrated that both extraction methods successfully produced thermally stable MDPs, while UA-DES yielded structurally modified polysaccharides with distinct thermal behavior.

3.7.6. Molecular weight distribution analysis

The HPGPC chromatogram (Fig. 10b) demonstrated significant variations in the molecular weight distributions of crude MDP extracted using HWE and UA-DES. Both extracts have single peaks with retention time of 31.25 min (HWE) and 27.40 min (UA-DES) with polydispersity index (PDI) of 1.026 and 1.039, respectively. These PDI values near to 1.0 indicate that both extraction methods produced polysaccharides with uniform molecular weight distributions and no detectable low molecular weight fragments. HWE MDP exhibited Mw = 77.4 kDa, Mp = 77.7 kDa and Mn = 75.4 kDa. In contrast, UA-DES MDPs showed high values of Mw = 560.1 kDa, Mp = 563.7 kDa and Mn = 539.0 kDa. This is a 7.3 fold increase in weight average molecular weight over HWE. The increased Mw of UA-DES MDPs might be due to the milder extraction temperature (60 °C) over HWE (90 °C) with ultrasonic cavitation that disrupts cell walls without degrading the released polysaccharides. The higher Mw of UA-DES MDPs supports the increase in antioxidant activity as shown in Fig. 11a-d and Table 7. Previous studies showed that higher molecular weight polysaccharides have better radical scavenging ability due to greater exposure of hydroxyl and carboxyl groups of polysaccharides for electron donation [73]. Additionally, as shown in Table 6, UA-DES MDPs contained higher galacturonic acid (22.87 %) contents as compared to HWE (10.40 %) MDPs, which together with higher Mw provides more abundant active sites for neutralizing free radicals. Overall, between two extraction methods for MDPs yield, UA-DES extraction is better for preserving high molecular weight MDPs with enhanced intrinsic bioactivity.

Fig. 11.

Comparative evaluation of the antioxidant properties of MDPs isolated by UA-DES and HWE: (a) DPPH radical scavenging activity, (b) Hydroxyl radical scavenging capacity, (c) Lipid peroxidation inhibition, and (d) Reducing power. Statistical significance was determined by one-way ANOVA followed by Tukey's post-hoc test at each concentration. ***p < 0.001, **p < 0.01, *p < 0.05 between indicated groups and ns = not significant.

Table 7.

Half maximal inhibitory concentration (IC₅₀) values of HWE and UA-DES extracted MDPs and ascorbic acid (Vc) in different antioxidant assays.

Antioxidant Activity	HWE (IC50)	UA-DES (IC50)	Vc (IC50)
DPPH Radical Scavenging	42.7 ± 2.5	39.1 ± 2.1	33.9 ± 1.8
Hydroxyl Radical Scavenging	4.86 ± 0.21	4.55 ± 0.18	0.19 ± 0.21
Lipid Peroxidation Inhibition	3.00 ± 0.12	2.89 ± 0.13	0.91 ± 0.04

3.7.7. Analysis of protein, uronic acid and total phenolic content

After deproteinization through Sevag protocol, the protein concentration in the MDPs solutions was assessed by using bicinchoninic acid (BCA) assay. The MDPs contained 0.020 mg/mg of protein content in the UA‑DES extract and 0.032 mg/mg in the HWE extract.

The uronic acid content of polysaccharides extracted through UA-DES (35.62 %) was significantly greater than that of those obtained through conventional HWE (24.87 %). This result suggests that the UA-DES method preferentially enriched the acidic polysaccharide fractions. Similar trends have been reported for UA-DES extraction of okra polysaccharides, where DES-derived extracts exhibited higher uronic acid contents (37.87 %) compared with hot water ultrasonic extraction (28.80 %) and demonstrated enhanced antioxidant activity due to differential solubilisation of acid rich components [20]. The higher uronic acid content in UA-DES extracts may contribute to improve the functional properties. Uronic acid contents are known to play an important role in increasing the antioxidant activity of plant polysaccharides [74]. These results support the idea that UA-DES extraction not only increases the yield of polysaccharide but also considerably alters the chemical composition that may influence the biological activity.

To eliminate any possible interference from non-polysaccharide constituents the total phenolic content (TPC) of the MDPs was assessed through Folin–Ciocalteu method [33]. Gallic acid was used as the standard and results were reported as milligrams of gallic acid equivalents per gram of sample (mg GAE/g). The TPC for HWE MDPs was 0.143 mg GAE/g, whereas UA-DES was 0.147 mg GAE/g. These relatively low values of TPC indicate that phenolic contamination is minimal and antioxidant activities attributed to MDPs rather than to non-polysaccharide components.

3.8. Antioxidant activity of UA-DES and HWE of MDPs

3.8.1. DPPH radical scavenging assay

The DPPH radical scavenging assay is a well-known method to determine the antioxidant activity of natural compounds. DPPH radical freshly prepared solution in ethanol shows a deep purple colour and exhibits maximum absorbance at 517 nm. MDPs extracted using UA-DES, HWE, and Vc as a reference were tested for their DPPH radical scavenging capabilities at the concentrations ranging from 10 to 100 µg/mL (Fig. 11a). A dose-dependent response was observed for all samples. The HWE derived MDPs exhibited scavenging activities from 27.2 ± 0.6 % (10 µg/mL) to 86.3 ± 0.5 % (100 µg/mL) and IC₅₀ of 42.7 µg/mL. In contrast, MDPs extracted through UA-DES displayed considerably higher antioxidant capacity (p < 0.05) with scavenging efficiencies varying from 39.0 ± 0.7 % to 97.6 ± 0.5 % across the same concentration of 10 and 100 µg/mL with an enhanced IC₅₀ value of 39.1 µg/mL. Vc possessed strongest radical scavenging activity from 32.1 ± 0.5 % to 91.8 ± 0.6 % and achieved IC₅₀ value of 33.9 µg/mL. The UA-DES technique significantly improves the bioactive characteristics of MDPs compared to HWE, as evidenced by a 9.3 % increase in antioxidant activity (IC₅₀ value). It might be due to DESs stronger solvation power and ultrasounds capacity to rupture cell walls [20]. These results are in line with previous studies.

on UA-DES based extraction of plant polysaccharides with notable DPPH radical scavenging abilities. UA-DES polysaccharides from Cercis chinensis bark executed 1.7 times better DPPH radical scavenging activity than HWE polysaccharides [52] and okra polysaccharides exhibited approximately 29.79 % increase in DPPH radical scavenging activity of UA-DES polysaccharides as compared to HWE polysaccharides [20]. These results highlight UA-DES as a better method for polysaccharides extraction from plants with enhanced DPPH free radical scavenging potential. Relatively small difference between IC₅₀ value of Vc (Table 7) and UA-DES extracted MDPs, making Vc approximately 1.15 times more potent. These results suggest that the complex macromolecular mixtures of UA-DES extracted MDPs possess remarkable radical scavenging abilities. That might be due to interaction of multiple antioxidant mechanisms. The good efficiency of UA-DES highlights, MDPs extract as potential natural antioxidants worthy of further research and development.

Data are presented as mean ± SD (n = 3). IC₅₀ values for DPPH assay are expressed in µg/mL; hydroxyl radical scavenging and lipid peroxidation inhibition are in mg/mL.

3.8.2. Hydroxyl radical scavenging activity

The hydroxyl radical can easily penetrate through the cell membrane and induce severe oxidative stress, leading to cell damage or even death. Scavenging hydroxyl radicals is therefore crucial for preventing oxidative damage [75]. The hydroxyl radical scavenging potential of MDPs obtained through UA-DES and HWE was evaluated by using ascorbic acid (Vc) as a positive control. Both MDPs extract exhibited a dose-dependent increase in scavenging efficacy across the tested concentration of 0.125 to 6.0 mg/mL. Notably, UA-DES extract showed significantly higher inhibition than HWE extract (p < 0.05). At 6.0 mg/mL, the UA-DES extract achieved 58.40 ± 0.59 % scavenging of hydroxyl radicals compared with 52.26 ± 0.59 % for the HWE extract (Fig. 11b). Vc revealed prominently stronger hydroxyl radical scavenging (94 %) just at 0.5 mg/mL, may be due to low molecular weight and strong electron-donating capacity. UA-DES extract exhibited a significantly lower IC₅₀ value (4.55 ± 0.18 mg/mL) than the HWE (4.86 ± 0.21 mg/mL) (p < 0.05). The enhanced hydroxyl radical scavenging activity of MDPs extracted through UA-DES is likely due to the collective effects of DESs and ultrasonic cavitation [17]. Ultrasonic treatment promotes rapid cell wall disruption, improves the mass transfer, and facilitates the extraction of polysaccharide fractions with increased exposure of hydroxyl and carboxyl groups, which play a critical role in neutralization of free radical. Additionally, ultrasound induced mechanical effects may partially reduce the molecular weight or alter the polysaccharide conformation [73], [76] and enhance the antioxidant activity. In contrast, prolonged thermal exposure during HWE may lead to structural damage, rearrangement or aggregation of polysaccharide chains, resulting in moderately lower the scavenging efficiency for hydroxyl radicals [77]. Overall, UA-DES derived MDPs showed significantly improved antioxidant performance as compared with HWE derived MDPs, highlighting the effectiveness of UA-DES extraction for enhancing the hydroxyl radical scavenging activity of plant polysaccharides.

3.8.3. Lipid peroxidation inhibition ability of MDPs

Lipid peroxidation inhibition ability of MDPs extracted via UA-DES and HWE methods, along with Vc as a reference standard, was evaluated by using a lipid peroxidation inhibition assay (Fig. 11c). All samples demonstrated concentration dependent activity across 0.125–6 mg/mL. HWE MDPs exhibited inhibition ranging from 7.34 ± 0.79 % at 0.125 mg/mL to 78.02 ± 0.87 % at 6 mg/mL. UA-DES extracted MDPs showed enhanced activity with inhibition values ranging from 7.95 ± 0.44 % to 79.81 ± 1.04 %. Vc displayed the most potent inhibition ability, from 11.34 ± 1.16 % to 99.27 ± 0.43 % over 0.125–6 mg/mL. The study showed the superior bioactivity of MDPs extracted via UA-DES compared to HWE. The IC₅₀ value for UA-DES extracted MDPs was 2.89 ± 0.13 mg/mL, indicating a 3.8 % increase in efficacy over MDPs from HWE (IC₅₀ = 3.00 ± 0.12 mg/mL). These results validate the effectiveness of UA-DES in preserving the lipid peroxidation inhibition ability of MDPs. Vc as standard demonstrated the highest inhibitory potency with IC₅₀ value of 0.91 ± 0.04 mg/mL, making it approximately 3.2 times more potent than UA-DES extracted MDPs.

3.8.4. Reducing capacity

The reducing capacity of polysaccharides reflects their electron donating potential and provides critical insight into antioxidant potential. The reducing capacity of the extracted MDPs was evaluated using the K₃Fe(CN)₆ reduction assay. The absorbance values at 700 nm for HWE, UA-DES, and Vc samples increased with concentration of MDPs, which ultimately increased the reducing power (Fig. 11d). UA-DES extracted MDPs demonstrated higher reducing capacity as compared to HWE across the entire tested range of concentrations. At the highest tested concentration (6 mg/mL), UA-DES extract exhibited an absorbance of 0.0747. That is significantly higher than HWE absorbance (0.0647) but lower than the absorbance of Vc (0.2467). These results exhibits that UA-DES extraction enhances the electron donating ability of the MDPs. This development might due to more efficient extraction of polysaccharide with greater availability of hydroxyl groups that can act as electron donors. Our findings are consistent with previous reports on polysaccharides from lychee nuclear [37]. The observed concentration dependent trend indicates that the reducing capacity is substantially connected to the accessibility of hydroxyl groups and other redox-active moieties inside polysaccharide chains. The differences between HWE and UA-DES extracts were negligible at lower concentration of ≤ 0.5 mg/mL and at higher concentration (≥2 mg/mL) the difference became more apparent. This suggests a cumulative influence of polysaccharides composition on antioxidant capacity. Interestingly, the linear increase in absorbance without rapid saturation suggests that MDPs have a considerable reducing potential even at comparatively high concentrations that might improve their ability to scavenge free radicals in biological systems. Overall, these results show that UA-DES extraction has two benefits. It increases the yield of MDPs and improves the functional quality of MDPs by enhancing their capacity to donate electrons. The enhanced reducing capacity of UA-DES polysaccharides demonstrates their potential as natural antioxidants in functional foods and pharmaceutical applications.

3.8.5. In vitro cellular antioxidant and anti-inflammatory activities

To evaluate the biological efficacy of the extracted polysaccharides and exclude potential cytotoxicity, a CCK-8 assay was first performed on RAW264.7 macrophages. HWE and UA-DES MDPs exhibited no significant cytotoxicity across a broad concentration range from 25 to 800 μg/mL indicating excellent biocompatibility (Fig. 12a, b). Therefore, safe concentrations low (L), medium (M) and high (H) (50, 100 and 200 μg/mL) were selected for subsequent biological assays.

Fig. 12.

Cellular antioxidant and anti-inflammatory activities of HWE and UA-DES extracted MDPs in LPS-stimulated RAW264.7 macrophages. (a, b) CCK-8 assays showing no cytotoxicity up to 800 μg/mL. (c) Inhibition of LPS-induced NO production. (d) DCFH-DA fluorescence staining showing intracellular ROS scavenging (images and quantification). UA-DES-MDPs dose-dependently reduced ROS and NO, outperforming HWE-MDPs. Data are mean ± SD (n = 3).

Inflammation and oxidative stress are tightly coupled in macrophage activation. We evaluated the inhibitory effects of the polysaccharides on LPS induced NO production. LPS stimulation drastically increased NO levels in the culture supernatant compared to the normal control (Fig. 12c). Pretreatment with UA-DES MDPs significantly and dose-dependently suppressed NO production across all tested concentrations. Notably, the high dose group (200 μg/mL) achieved the maximal inhibitory efficacy among the polysaccharide treatments, significantly outperforming HWE MDPs at all corresponding doses, although it remained slightly inferior to the positive control dexamethasone (DEX). Furthermore, the intracellular ROS scavenging ability was visualized and quantified using the DCFH-DA probe (Fig. 12d) and fluorescence images). LPS exposure triggered a massive accumulation of intracellular ROS, exhibiting intense green fluorescence. Consistent with the NO inhibition trend, UA-DES-MDPs effectively mitigated ROS overproduction in a dose-dependent manner. Remarkably, both the medium (100 μg/mL) and high-dose (200 μg/mL) UA-DES-MDPs groups quenched the fluorescence intensity more profoundly than the positive control, demonstrating an exceptionally strong intracellular antioxidant capacity. In contrast, the HWE-MDPs exhibited significantly weaker ROS clearance abilities, highlighting the superior biological potential of the polysaccharides extracted via the UA-DES method.

Although our in vitro chemical assays (e.g., DPPH and hydroxyl radical scavenging) confirmed the antioxidant potential of the extracted polysaccharides, chemical models lack the physiological complexity of living cells. Therefore, a macrophage-based LPS stimulation model was utilized to further validate their biological relevance. The massive production of ROS in LPS-activated macrophages is intrinsically linked to oxidative damage and inflammatory responses. Our cellular results explicitly demonstrate that UA-DES-MDPs possess profound intracellular ROS scavenging and NO inhibiting capacities, significantly outperforming HWE-MDPs. This better bioactivity of UA-DES-MDPs can be attributed to the specific advantages of the deep eutectic solvent extraction. As noted earlier, the UA-DES method effectively enriched specific uronic acids (e.g., galacturonic acid). The highly functionalized carboxyl and hydroxyl groups within these enriched uronic acid units provide abundant hydrogen-donating sites, enabling more efficient neutralization of intracellular free radicals. Furthermore, compared to the harsh thermal degradation often caused by conventional prolonged hot water extraction, the gentle acoustic cavitation combined with the hydrogen-bond network of the DES solvent likely preserved the structural integrity and optimal spatial conformation of the polysaccharides. These findings solidly bridge the gap between structural features and cellular bioactivities, confirming that the proposed green UA-DES methodology is highly effective in producing bioactive polysaccharides for potential functional food and therapeutic applications.

4. Conclusion

This work successfully established the UA-DES approach as an efficient strategy for extraction of MDPs. Twenty-five DESs were prepared and analysed for the extraction of MDPs; among them, DES-3.2 (choline chloride: citric acid, 1:2) was found to be remarkably efficient in combination with UAE in yield improvement of MDPs. RSM successfully optimized the interaction of parameters liquid–solid ratio, ultrasonic temperature, ultrasonic power, and ultrasonic time. ANN outperformed the RSM in yield prediction accuracy by efficiently captured the complicated nonlinear interactions intrinsic to extraction process. MDPs yield of 1.38 % was achieved using optimized UA-DES conditions (liquid–solid ratio of 30 mL/g, ultrasonic temperature of 60 °C, ultrasonic power of 360 W and ultrasonic time of 90 min), which was increase of 51.65 % over the yield achieved with HWE (0.91 %). Kinetic studies showed that a second-order diffusion controlled process prevailed during the rapid initial release and equilibrium phases of polysaccharide recovery. ANN demonstrated better accuracy over the second-order kinetic model for predicting time-dependent polysaccharide concentration. Comprehensive structural characterization was performed to support structure–activity relationships. SEM analysis revealed that UA-DES significantly ruptured the cell walls of the MD biomass and that the MDPs exhibited fragmented, dispersed and porous surfaces compared to the aggregated compact particles of the HWE MDPs. FT-IR and XRD confirmed the preservation of functional groups and amorphous structure. HPGPC analysis exposed that UA-DES extracted MDPs possess a significantly higher molecular weight (560.1 kDa) compared to HWE (77.4 kDa) MDPs. TGA confirmed the improved thermal stability of MDPs. Galactose, galacturonic acid, and glucose were found to be most predominant monosaccharides in the MDPs with UA-DES extraction significantly enriched galacturonic acid content by 22.87 % compared to HWE (10.40 %). Uronic acid content was 35.62 % for UA-DES and 24.87 % for HWE and total phenolic content was low and comparable between extracts confirming that bioactivity is attributed to polysaccharides rather than phenolic interference. Regarding bioactivity, UA-DES extracted MDPs exhibited higher antioxidant activity, including radical scavenging and reducing power than HWE at same tested concentration. Importantly, cellular assays using RAW264.7 macrophages demonstrated that UA-DES extracted MDPs significantly reduced the LPS-induced intracellular ROS and NO production in a dose-dependent manner with no cytotoxicity up to 800 μg/mL. This confirms their potential as natural anti-inflammatory and antioxidant agents. The DES recycling study further validates its environmental and economic advantages as extraction solvent. Although 1.38 % yield is relatively modest, DESs exhibited excellent extraction performance over five recycling runs. This efficiency provides a significant economic benefit by decreasing solvent costs and waste generation. Additionally, enhancement in 51.65 % yield over HWE underscores the practical feasibility of the UA-DES approach for recovering high value bioactive MDPs. It may not be practical for industrial preparation of bulk polysaccharides on large scale, but certainly provides a green and sustainable strategy for extraction of high value bioactive polysaccharides where environment concern and solvent recyclability is prioritized. Overall, this study established UA-DES extraction as an effective and high-performance approach that addresses key limitations of conventional HWE by preserving polysaccharide integrity and enhancing bioactivity, offering promising applications in functional foods, pharmaceuticals and nutraceuticals.

CRediT authorship contribution statement

Waqas Haider: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Zhixi Lin: Investigation, Data curation. Kuixu Gao: Formal analysis, Data curation. Jingchao Li: Visualization, Validation, Data curation. Yafei Li: Investigation, Data curation. Jinnan Zhong: Formal analysis, Data curation. Liuwei Yuan: Formal analysis, Data curation. Zhouxiang Chen: Formal analysis, Data curation. Yuchi Chen: Formal analysis, Data curation. Zhishan Ding: Writing – review & editing, Visualization, Validation, Supervision, Resources, Project administration, Funding acquisition, Conceptualization. Fangmei Zhou: Visualization, Validation, Resources, Data curation. Mingyuan Zhou: Visualization, Validation, Resources, Funding acquisition, Data curation.

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.

Appendix A. Supplementary data

The following are the Supplementary data to this article: