Response Surface Optimization of Extraction Conditions for the Active Components with High Acetylcholinesterase Inhibitory Activity and Identification of Key Metabolites from Acer truncatum Seed Oil Residue

The State Council of China has called for the comprehensive development and utilization of Acer truncatum resources. However, research on one of its by-products, namely seed oil residue (ASR), from seed oil extraction is seriously insufficient, resulting in a waste of these precious resources. We aimed to optimize the conditions of ultrasound-assisted extraction (UAE) using a response surface methodology to obtain high acetylcholinesterase (AChE) inhibitory components from ASR and to tentatively identify the active metabolites in ASR using non-targeted metabolomics. Based on the results of the independent variables test, the interaction effects of three key extracting variables, including methanol concentration, ultrasonic time, and material-to-liquid ratio, were further investigated using the Box–Behnken design (BBD) to obtain prior active components with high AChE inhibitory activity. UPLC-QTOF-MS combined with a multivariate method was used to analyze the metabolites in ASR and investigate the causes of activity differences. Based on the current study, the optimal conditions for UAE were as follows: methanol concentration of 85.06%, ultrasonic time of 39.1 min, and material-to-liquid ratio of 1.06:10 (g/mL). Under these optimal conditions, the obtained extracts show strong inhibitions against AChE with half maximal inhibitory concentration (IC50) values ranging from 0.375 to 0.459 µg/mL according to an Ellman’s method evaluation. Furthermore, 55 metabolites were identified from the ASR extracted using methanol in different concentrations, and 9 biomarkers were subsequently identified as potential compounds responsible for the observed AChE inhibition. The active extracts have potential to be used for the development of functional foods with positive effects on Alzheimer’s disease owing to their high AChE inhibition activity. Altogether, this study provides insights into promoting the comprehensive utilization of A. truncatum resources.


Introduction
Acer truncatum, a perennial deciduous arbor from the genus Acer (Sapindaceae), is native to China and has been widely cultivated around the world for its ornamental, ecological, edible, and nutritional values [1]. A. truncatum seed oil (ATO) has been considered a high-quality woody oil [2], thanks to its abundant unsaturated fatty acids (up to 90%) and its high content of nutritional fatty acids including nervonic acid (5.52%) [3], linoleic acid (37.3%), and oleic acid (25.8%) [4]. Moreover, ATO has also been reported as a healthcare oil with various health benefits, including antioxidant effects, antitumor properties [5], antibacterial effects [6], hypolipidemic properties [7], enhanced brain nerve activity [8],

Ultrasound-Assisted Extraction Process
In a 15 mL test tube, ASR samples of varying weights were mixed with 10 mL of different concentrations of methanol solutions, followed by shaking (100 rpm) for 5 min and ultrasonic extraction (KQ3200E, Kunshan Ultrasonic Instrument, Kunshan, China) at different times. The samples were then centrifuged (H1850, Xiangyi Centrifuge Instrument, Changsha, China) at 3500 rpm for 5 min, and the extraction process was repeated twice. The combined supernatant was dried at 50 • C and stored at −20 • C until use.

In Vitro AChE Inhibitory Activity Assays
An AChE inhibitory activity assay was performed in a 96-well plate according to a modified Ellman's method [31]. Briefly, ASR was dissolved in phosphate-buffered saline (PBS) solution (0.1 M, pH 8.0) and then diluted with PBS to obtain sample solutions with concentrations of 100, 50, 25, 10, and 2.5 µg/mL, respectively. Subsequently, 100 µL of PBS solution, 20 µL of AChE solution (0.2 U/mL), and 20 µL of sample solution were added to each well and mixed thoroughly. The plates were incubated at 37 • C for 10 min and further incubated at 37 • C for 20 min after 20 µL of ATCI (2 mM, PBS as the solvent) was added. Finally, 20 µL of sodium dodecyl sulfate (0.1 M, SDS) and DTNB (2 mM) were added into each well as terminate and chromogenic reagents, respectively. Thus, the final sample concentration of each sample in the well changed to 10, 5, 2.5, 1, and 0.25 µg/mL, respectively. The absorbance value was measured at 405 nm using a microplate reader (Thermo Fisher Scientific, Waltham, MA, USA). The blank control was conducted with PBS instead of the samples. The background control used the same volume of PBS to replace where A 0 is the absorbance of the blank control group, A 1 is the absorbance of the background control group, and A 2 is the absorbance of the sample group.
The inhibitory rates were expressed as the average of three repeated experiments and the standard deviation. The half maximal inhibitory concentration (IC 50 ) value of enzyme activity was calculated with non-linear regression using SPSS Statistics version 26.0 (IBM SPSS Inc., Chicago, IL, USA).

Response Surface Methodology Design
RSM was used to evaluate the influences of independent variables, including methanol concentration (X 1 , %), ultrasonic time (X 2 , min), and material-to-liquid ratio (X 3 , g/mL), on the responses of IC 50 (Y, µg/mL). The Box-Behnken design (BBD) was selected in the response surface test design, and the influence of unexplained variability in the response was minimized via randomized experiments [32]. The variables were evaluated at three levels (1, 0, and −1) containing 17 runs and 5 center points. The level determinations of three variables were evaluated through single-factor analysis ( Table 1). The second-order polynomial equation for predicting the optimum parameter in RSM was as follows: Y = A 0 + A 1 X 1 + A 2 X 2 + A 3 X 3 + A 12 X 1 X 2 + A 13 X 1 X 3 + A 23 X 2 X 3 + A 11 X 1 2 + A 22 X 2 2 + A 33 X 3 2 .  22 , and A 33 are quadratic coefficient terms. Analysis of variance (ANOVA) was employed to determine the significance of the data in the model.

UPLC-QTOF-MS Analysis of ASR Compositions
The sample (10 mg) was added to 120 µL of 75% methanol-water and vortex dissolved. The supernatant was used for UPLC-QTOF-MS analysis, after being centrifuged at 17,000× g. An amount of 75% methanol-water was used as a blank sample. Quality control (QC) samples were prepared by mixing all sample solutions in equal proportion to analyze the repeatability and stability of the analytical process under the same treatment.

Identification of Compounds and Statistical Analysis
The raw data collected through UPLC-QTOF-MS were converted to .abf format using an ABF converter (https://www.reifycs.com/AbfConverter/), and the .abf -format data were imported into MS-DIAL 4.7 for data processing [33]. Data processing included data collection, peak detection, metabolite identification, adduct-ion merging, and isotope tracking. Information of on retention time, compound molecular weight, molecular formula, peak area, and identification were exported. Confirmation of the putative identifications was performed by checking compound fragments (−10 ppm ≤ mass error ≤ 10 ppm) in Peak View 1.2 (AB SCIEX, Framingham, MA, USA) and previously published reported data.
MetaboAnalyst 5.0 was used for multivariate statistical analysis, including principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), orthogonal partial least squares discriminant analysis (OPLS-DA), metabolite classification, and enrichment analysis. The differential metabolites in OPLS-DA were screened using variable importance projection (VIP) values and the results of fold change (FC) and t-test from ANOVA, with parameter settings of p value < 0.05, VIP ≥ 1, and log 2 FC ≥ 1 or ≤−1.

Determination of the Range of Independent Variables
The influence of methanol concentration, ultrasonic time, and material-to-liquid ratio on AChE inhibition rate and IC 50 values were studied first. As shown in Figure 1a, the inhibition rates of AChE increased significantly when the methanol concentration was less than 80% and gradually decreased when the methanol concentration exceeded 80%, except for the concentration of 2.5 µg/mL. When the ASR concentration was 2.5 µg/mL, the inhibition rate of AChE increased with an increase in methanol concentration, but its growth rate slowed down when the methanol concentration exceeded 80%. When the ASR concentration was 10, 5, 1, or 0.25 µg/mL and methanol concentration was 80%, the AChE inhibition rate reached the maximum value, and IC 50 reached its minimum value (IC 50 = 2.1216 ± 0.0888 µg/mL, Figure 1b). It has been reported that solvents with low methanol concentrations are more effective in separating water-soluble substances such as polysaccharides, proteins, and pigments. However, these solvents may also lead to the extraction of impurities in the obtained extract [34]. In addition, a previous study also showed that increasing the methanol concentration was conducive to the dissolution of flavonoids [35]. The concentration of 80% methanol may be more suitable for extracting the AChE inhibitory component from ASR due to its polarity, resulting in relatively better activity. From the above results, the methanol concentrations ranging from 65% to 95% were selected to establish the models.
As shown in Figure 1c, the inhibition rates of AChE fluctuated and improved with the ultrasonic time from 10 min to 40 min, peaked at 40 min, and then slowly declined with increasing ultrasound time. When the ultrasonic time was 40 min, the inhibition rate of AChE reached the maximum value, and the minimum value of IC 50 was 0.5380 ± 0.0485 µg/mL (Figure 1d). Research has shown that longer ultrasound times results in higher yields of active ingredients. However, prolonged ultrasonic time can lead to an increase in the temperature of the extraction medium, reducing the solvent's permeability into the cell wall [36]. This can result in the precipitation of impurities, disruption of the structure of small molecules, and degradation or transformation of flavonoids [37]. Therefore, the ultrasound time from 30 min to 50 min was chosen to obtain the models. value (IC50 = 0.3310 ± 0.0368 µg/mL, Figure 1f). Studies have shown that the driving force of mass transfer in ultrasonic extraction is the concentration gradient between solid and solvent [38]. Generally, the rate of active substance extraction from the solid matrix to the solvent increases with the concentration gradient. However, this effect becomes weaker when the ratio of solvent to solid is too high [39]. The contact area between ASR and solvent decreases as the material-to-liquid ratio increases, which can result in some active materials not being effectively dissolved or the dissolution rate slowing down [40]. In this study, a material-to-liquid ratio range of 0.825:10 to 1.225:10 (g/mL) was selected for model building, taking into account the need for economic and efficient extraction. As depicted in Figure 1e, with the exception of ASR concentration of 0.25 µg/mL, the inhibition rates of AChE increased with the increase in the material-to-liquid ratio. However, when the material-to-liquid ratio exceeded 1.025:10 (g/mL), the growth rate slowed down. When the material-to-liquid ratio was 1.025:10 (g/mL), the IC 50 was the minimum value (IC 50 = 0.3310 ± 0.0368 µg/mL, Figure 1f). Studies have shown that the driving force of mass transfer in ultrasonic extraction is the concentration gradient between solid and solvent [38]. Generally, the rate of active substance extraction from the solid matrix to the solvent increases with the concentration gradient. However, this effect becomes weaker when the ratio of solvent to solid is too high [39]. The contact area between ASR and solvent decreases as the material-to-liquid ratio increases, which can result in some active materials not being effectively dissolved or the dissolution rate slowing down [40]. In this study, a material-to-liquid ratio range of 0.825:10 to 1.225:10 (g/mL) was selected for model building, taking into account the need for economic and efficient extraction.

Box-Behnken Design and Model Fitting
The BBD experiment design and results are presented in Table 2, and the second-order polynomial equation showing the effect of methanol concentration (X 1 , %), ultrasonic time (X 2 , min), and material-to-liquid ratio (X 3 , g/mL) on IC 50 (Y, µg/mL) is expressed as follows: To test the validity and predictability of the model, the results were analyzed using ANOVA ( Table 3). The p-value and F-value were used to estimate the statistical significance of the model. The p-value < 0.05 implies that the model is significant and the p-value < 0.01 implies that the model is extremely significant. It was observed that the model was significant (p < 0.05), and the lack of fit was insignificant (p > 0.05). These results indicated that the established model could be used to predict the IC 50 from the extraction conditions (methanol concentration, ultrasonic time, and material-to-liquid ratio). The correlation coefficient value (R 2 ) was 0.9505, indicating that 95.05% of the IC 50 value could be explained with the regression model. However, R 2 may not be accurate when there are many variables that continue to rise. Thus, the adjusted determination coefficient value (R Adj 2 ) was often used to replace R 2 and further validated the significance between independent variables and responses [41]. The predicted values obtained from the second-order polynomial equations were close to the experiment values and scattered around the theoretical line, as illustrated in Figure 2a. Figure 2b is the normal plot of the residual, and the spot approximately along the straight line represents the acceptable reproducibility of the method. These results indicated that the relationship between the test and predicted values was reliable and accurate. Moreover, it was found that in addition to linear (X 1 ) and quadratic (X 1 2 ) coefficients, the other variables did not show significant effects on the IC 50 values (Table 3). Figure 2c depicts the perturbation plot of the IC 50 ; the steep curvature with the methanol concentration demonstrated that the IC 50 is rapidly responsive to this factor, while the relatively flat line of the ultrasonic time and material-to-liquid ratio indicates their minimal effect on the IC 50 . In summary, the factors affecting IC 50 in descending order of importance were X1 followed by X3 and X2. Moreover, it was found that in addition to linear (X1) and quadratic (X1 2 ) coefficients, the other variables did not show significant effects on the IC50 values (Table 3). Figure 2c depicts the perturbation plot of the IC50; the steep curvature with the methanol concentration demonstrated that the IC50 is rapidly responsive to this factor, while the relatively flat line of the ultrasonic time and material-to-liquid ratio indicates their minimal effect on the IC50. In summary, the factors affecting IC50 in descending order of importance were X1 followed by X3 and X2.

Analysis of the Variable Interaction
The three-dimensional (3D) response surface and contour plots were generated to facilitate the visualization of the significant variables and explore the interaction of each factor, as illustrated in Figure 3. The IC50 decreased as the interaction between methanol concentration and ultrasonic time increased up to an optimum point, after which it slowly increased (Figure 3a,b). Notably, the negative model term (−0.0111 X1X2) reveals a

Analysis of the Variable Interaction
The three-dimensional (3D) response surface and contour plots were generated to facilitate the visualization of the significant variables and explore the interaction of each factor, as illustrated in Figure 3. The IC 50 decreased as the interaction between methanol concentration and ultrasonic time increased up to an optimum point, after which it slowly increased (Figure 3a,b). Notably, the negative model term (−0.0111 X 1 X 2 ) reveals a confrontational behavior between the variables, indicating that the IC 50 may decrease with higher methanol concentrations and longer ultrasonic times. concentration and ultrasonic time, this interaction was more prominent. Interestingly, the positive model term (+ 0.0435 X1X3) indicates a synergistic effect, confirming that decreasing both factors to their minimum values leads to a reduction in the IC50. Figure 3e,f elucidate the response surface and contour plots of the interaction of ultrasonic time and material-to-liquid ratio. The positive model term (+0.0385 X2X3) promulgates a synergistic effect between the variables. It was noticed that the interaction between various variables was not significant (p > 0.05, Table 3).   50 . Compared to the interaction between methanol concentration and ultrasonic time, this interaction was more prominent. Interestingly, the positive model term (+0.0435 X 1 X 3 ) indicates a synergistic effect, confirming that decreasing both factors to their minimum values leads to a reduction in the IC 50 . Figure 3e,f elucidate the response surface and contour plots of the interaction of ultrasonic time and material-to-liquid ratio. The positive model term (+0.0385 X 2 X 3 ) promulgates a synergistic effect between the variables. It was noticed that the interaction between various variables was not significant (p > 0.05, Table 3).

Optimization and Validation of the Model
According to the Design Expert 11.1.0 software (Stat-Ease Inc., Minneapolis, MN, USA), the optimal extraction conditions of the active components, in terms of AChE inhibition in ASR, were as follows: methanol concentration of 85.06%, ultrasonic time of 39.1 min, and material-to-liquid ratio of 1.06:10 (g/mL), whereby the predicted IC 50 value (Y) was 0.292 µg/mL. Under these optimum conditions, the observed value in the verification test ranged from 0.375 to 0.459 µg/mL, which was slightly different compared to the predicted one, suggesting that the established optimized conditions were reliable and that the regression model was suitable for extracting the active ingredients from ASR. The IC 50 measured value was different from that of our previous study, which may be due to the data deviation caused by various factors, such as different years of material harvesting, different test operators, and environment conditions. Additionally, the activity of AChE can be easily influenced by environmental factors. However, the deviation is small and within acceptable limits. Overall, ASR shows a strong AChE inhibitory activity.

Multivariate Statistical Analysis
MS-DIAL software and public database (MS/MS-Public-Pos/Neg 17) were used for data processing, resulting in the characterization of 3486 mass features. Among these, 1598 were found using positive ion mode and 1888 using negative ion mode. These features were exported in the format of .csv for the first feature screening to eliminate uncertain and duplicate metabolite information. MetaboAnalyst omics online platform was employed for multivariate statistical analysis based on these 3486 features from ASR extraction using different methanol concentrations ( Figure 5). The results show that the QC group overlapped together in PCA, which indicates that the systems were stable during data acquisition. In addition, the PLS-DA model was presented with a satisfactory discriminating ability to divide the five groups in positive ion mode (R 2 Y = 0.9427, Q 2 = 0.6907), but the difference in negative ion mode is not very obvious (R 2 Y = 0.9919, Q 2 = 0.7146), especially regarding M4 and M5. Furthermore, the OPLS-DA method was used to further explore the difference between the most active group (M4) and the least active group (M1). were found using positive ion mode and 1888 using negative ion mode. These features were exported in the format of .csv for the first feature screening to eliminate uncertain and duplicate metabolite information. MetaboAnalyst omics online platform was employed for multivariate statistical analysis based on these 3486 features from ASR extraction using different methanol concentrations ( Figure 5). The results show that the QC group overlapped together in PCA, which indicates that the systems were stable during data acquisition. In addition, the PLS-DA model was presented with a satisfactory discriminating ability to divide the five groups in positive ion mode (R 2 Y = 0.9427, Q 2 = 0.6907), but the difference in negative ion mode is not very obvious (R 2 Y = 0.9919, Q 2 = 0.7146), especially regarding M4 and M5. Furthermore, the OPLS-DA method was used to further explore the difference between the most active group (M4) and the least active group (M1).

Screening of Differential Metabolites
We employed OPLS-DA to evaluate the difference between the 55 metabolites in extracts of ASR with different methanol concentrations, and the results indicated that M4 (the highly active extract) and other groups (the weakly active extract) can be separated ( Figure 6). Next, we performed VIP and FC analyses to find the differential metabolites between groups, and a permutation test was applied to validate the OPLS-DA model (Figure 6). R 2 Y and Q 2 were close to 1, and Q 2 was higher than 0.5, indicating that the model was stable and reliable. Through the above screening methods, 9 differential metabolites between the two methanol extracts of ASR were tentatively identified (Table 5). Figure 7 shows the normalized peak intensity box plot of the different compounds. Each point on the graph represents a sample, and the content difference of the compounds among the groups can be found distinctly.

Screening of Differential Metabolites
We employed OPLS-DA to evaluate the difference between the 55 metabolites in extracts of ASR with different methanol concentrations, and the results indicated that M4 (the highly active extract) and other groups (the weakly active extract) can be separated ( Figure 6). Next, we performed VIP and FC analyses to find the differential metabolites between groups, and a permutation test was applied to validate the OPLS-DA model ( Figure 6). R 2 Y and Q 2 were close to 1, and Q 2 was higher than 0.5, indicating that the model was stable and reliable. Through the above screening methods, 9 differential metabolites between the two methanol extracts of ASR were tentatively identified (Table 5). Figure 7 shows the normalized peak intensity box plot of the different compounds. Each point on the graph represents a sample, and the content difference of the compounds among the groups can be found distinctly.

Conclusions
The experimental results indicated that the optimized conditions for UAE were effective in extracting AChE inhibitory components from ASR. The extracts with the IC50 values ranging from 0.375 to 0.459 µg/mL were obtained using a methanol concentration of 85.06%, ultrasonic time of 39.1 min, and material-to-liquid ratio of 1.06:10 (g/mL). Furthermore, 55 metabolites were identified from the ASR extracted using different methanol concentrations. Among them, resveratrol, riodictyol, scopoletin, ferulic acid, cinnamic acid, N-acetyltryptophan, L-Epicatechin, D-Tryptophan, and (2S,3S)-3,5,7-trihydroxy-2-  . At a concentration of 50 µg/mL, N-acetyltryptophan exhibited a significant AChE inhibition rate of 64.90 ± 1.61%. Therefore, it is considered a promising compound for the treatment of AD [56]. Resveratrol has been proven to protect neuronal cells with its antioxidant activity, improve the memory function of patients with dementia, and reverse AChE activity [57]. The study confirmed that when the concentration of L-Epicatechin was 5 mg/mL, the inhibition rate of AChE was 13.48% [58]. Cinnamic acid derivatives showed a good inhibition effect of Aβ (1-42) aggregation and good neuroprotection on PC12 cells against amyloid-induced cell toxicity, indicating that they were promising for further development as lead compounds in the treatment of AD [59]. Eriodictyol can alleviate LPS-induced neuroinflammation, amyloidogenesis, and memory impairment, and has been fully proven to possess excellent anti-inflammatory, antioxidant, and anticancer biological activities [60]. Therefore, we proposed that the differential metabolites screened by OPLS-DA could be the active compounds in the ASR extracts with high AChE inhibitory activity.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.