Selective Extraction of Flavonoids from Sophora flavescens Ait. by Mechanochemistry

Flavonoids from Sophora flavescens were selectively extracted by mechanochemical-promoted extraction technology (MPET) after using response surface methodology to determine the optimal extraction parameters. The highest yield of 35.17 mg/g was achieved by grinding the roots with Na2CO3 (15%) at 440 rpm/min for 17.0 min and water was used as the sole solvent with a ratio of solvent to solid material of 25 mL/g. Flavonoids prepared by MPET demonstrated relatively higher antioxidant activities in subsequent DPPH and hydroxyl radical scavenging assays. Main constituents in the extracts, including kurarinol, kushenol I/N and kurarinone, were characterized by HPLC-MS/MS, indicating good selective extraction by MPET. Physicochemical property changes of powder during mechanochemical milling were identified by scanning electron microscopy, X-ray powder diffraction, and UV-Vis diffuse-reflectance spectroscopy. Compared with traditional extraction methods, MPET possesses notable advantages of higher selectivity, lower extraction temperature, shorter extraction time, and organic solvent free properties.


Introduction
Sophora flavescens Ait. (S. flavescens) is a traditional herb medicine distributed in East Asia and some European countries. It has been widely used as a medicine and functional food ingredient for thousands of years because of its potential beneficial properties, such as improving mental health, anti-inflammatory, antiasthmatic, antihelminthic, free radical scavenging and antimicrobial activities [1,2]. Phytochemical research has revealed that the main active compounds in S. flavescens are quinolizidine alkaloids and prenylated flavonoids. Chemical and pharmacological research on the quinolizidine alkaloids has been quite thorough [3], whereas study of the flavonoids components remains relatively inadequate. Some research has suggested that flavonoids could enhance immunity, lower blood glucose levels [4], inhibit the activity of tyrosinase enzyme [5] and that they possess anticancer and anti-inflammatory activities [6]. In addition, flavonoids, as phenolic compounds, which usually have significant antioxidant activity, might find potential uses as natural antioxidants, food additives and functional foods. Therefore, the extraction of flavonoids from S. flavescens is a worthwhile objective.
Flavonones and flavanonols are the key representative components [7] in S. flavescens together with a few isoflavone and flavanone glycosides. However, flavonones and flavanonols are

Screening of Solid Reagents
Flavonoids can be potentially transformed into salt forms when ground with alkali. The minimal concentration of solid basic agent required depends on the kinetics of the mechanochemical process (rate of component transfer during mechanochemical treatment) and the content of target compound(s) in the plant tissues. In our preliminary study, several solid reagents were screened ( Figure 1a): (1) blank group: no solid reagent; (2) sodium borate; (3) diatomaceous earth; (4) basic aluminum oxide; (5) Ca 2 CO 3 ; (6) NaHCO 3 ; (7) Na 2 CO 3 ; (8) control group: conventional agitation of unmilled S. flavescens powder and unmilled Na 2 CO 3 ; (9) control group: conventional agitation of milled S. flavescens powder and unmilled Na 2 CO 3 . Unexpectedly, diatomaceous earth, basic aluminum oxide, and Ca 2 CO 3 had no positive effect on the yield of flavonoids, giving even lower yields than the blank group. It was proposed that the adsorbability of diatomaceous earth and basic aluminum oxide might give rise to these decreased yields. The pK b values of Na 2 CO 3 , sodium borate, Ca 2 CO 3 , and NaHCO 3 are 3.67, 4.76, 5.0, 7.65, respectively, which indicated that stronger alkalinity of the solid basic agents resulted in better extraction yield of the flavonoids. However, the pK b value alone cannot explain why Ca 2 CO 3 decreased the yield to give a yield that was lower than that of the blank group. One possible reason might be its poor solubility in water. In terms of food safety, sodium borate was discarded. It could be see that the milled group (7) had a better yield than the control groups (8) and (9), which indicated that milling could increase the yield, and milling with a solid reagent could increase the yield better than only milling the plant material followed by adding Na 2 CO 3 . Na 2 CO 3 and NaHCO 3 were chosen for further investigation at different concentrations of 3.0%, 6.0%, 9.0%, 12.0%, 15.0% and 18.0% respectively (wt %, mass ratio of Na 2 CO 3 to S. flavescens particles) under the same extraction conditions (grinding time: 10 min; rotational speed: 400 rpm; extraction solvent: water; extraction time: 20 min; ratio of solvent to material: 30 mL/g; acidification pH: 5.0; extraction temperature: 25˝C). By comparing the results (Figure 1b), it was clear that Na 2 CO 3 could achieve more than ten times better yield than NaHCO 3 . Therefore, Na 2 CO 3 was applied as the optimal solid reagent. ratio of solvent to material: 30 mL/g; acidification pH: 5.0; extraction temperature: 25 °C). By comparing the results (Figure 1b), it was clear that Na2CO3 could achieve more than ten times better yield than NaHCO3. Therefore, Na2CO3 was applied as the optimal solid reagent.

Acidification pH
As depicted in Figure 2, acidification pH values from 4-6 exhibited the maximum absorption wavelengths at about 291.00 nm and a weak shoulder absorption wavelength at about 339.00 nm, which correspond to the intense absorptions of flavonones and flavanonols. The main absorption was induced by the benzoyl system, and the weak shoulder peak was due to the B-ring of the flavonoid molecule that could not conjugate with the carbonyl group in the pyran ring. However, as the acidification pH ranged from 7 to 10, the maximum absorption wavelength was seen at 329.98 nm as well as a weak absorption band at 282.44 nm, which suggested that o-phenolic hydroxy groups were substituted [19]. Moreover, under mechanochemical treatment, in addition to the 7-or 4'-OH on flavonoid molecules that could react with Na2CO3, the 3-OH could be neutralized by Na2CO3 with a corresponding shift of the absorption wavelength from a long wavelength (free state) to a shorter wavelength (substituted state) [19]. This was identical to the curve seen at pH 7-10, even though it could not be carried out under normal conditions. Because flavonoids could be completely transformed into their precursors at pH 6, the optimal acidification pH value of 6 was selected.  (2) sodium borate; (3) diatomaceous earth; (4) basic aluminum oxide; (5) Ca 2 CO 3 ; (6) NaHCO 3 ; (7) Na 2 CO 3 ; (8) control group: conventional agitation of unmilled S. flavescens powder and unmilled Na 2 CO 3 ; (9) control group: conventional agitation of milled S. flavescens powder and unmilled Na 2 CO 3 ; Effect of Na 2 CO 3 and NaHCO 3 concentration on flavonoids yield (b). Data are presented as mean˘SD (n = 3).

Acidification pH
As depicted in Figure 2, acidification pH values from 4-6 exhibited the maximum absorption wavelengths at about 291.00 nm and a weak shoulder absorption wavelength at about 339.00 nm, which correspond to the intense absorptions of flavonones and flavanonols. The main absorption was induced by the benzoyl system, and the weak shoulder peak was due to the B-ring of the flavonoid molecule that could not conjugate with the carbonyl group in the pyran ring. However, as the acidification pH ranged from 7 to 10, the maximum absorption wavelength was seen at 329.98 nm as well as a weak absorption band at 282.44 nm, which suggested that o-phenolic hydroxy groups were substituted [19]. Moreover, under mechanochemical treatment, in addition to the 7-or 4'-OH on flavonoid molecules that could react with Na 2 CO 3 , the 3-OH could be neutralized by Na 2 CO 3 with a corresponding shift of the absorption wavelength from a long wavelength (free state) to a shorter wavelength (substituted state) [19]. This was identical to the curve seen at pH 7-10, even though it could not be carried out under normal conditions. Because flavonoids could be completely transformed into their precursors at pH 6, the optimal acidification pH value of 6 was selected. ratio of solvent to material: 30 mL/g; acidification pH: 5.0; extraction temperature: 25 °C). By comparing the results (Figure 1b), it was clear that Na2CO3 could achieve more than ten times better yield than NaHCO3. Therefore, Na2CO3 was applied as the optimal solid reagent.

Acidification pH
As depicted in Figure 2, acidification pH values from 4-6 exhibited the maximum absorption wavelengths at about 291.00 nm and a weak shoulder absorption wavelength at about 339.00 nm, which correspond to the intense absorptions of flavonones and flavanonols. The main absorption was induced by the benzoyl system, and the weak shoulder peak was due to the B-ring of the flavonoid molecule that could not conjugate with the carbonyl group in the pyran ring. However, as the acidification pH ranged from 7 to 10, the maximum absorption wavelength was seen at 329.98 nm as well as a weak absorption band at 282.44 nm, which suggested that o-phenolic hydroxy groups were substituted [19]. Moreover, under mechanochemical treatment, in addition to the 7-or 4'-OH on flavonoid molecules that could react with Na2CO3, the 3-OH could be neutralized by Na2CO3 with a corresponding shift of the absorption wavelength from a long wavelength (free state) to a shorter wavelength (substituted state) [19]. This was identical to the curve seen at pH 7-10, even though it could not be carried out under normal conditions. Because flavonoids could be completely transformed into their precursors at pH 6, the optimal acidification pH value of 6 was selected.

Optimization of the Operating Parameters
Preliminary experiments to determine the main factors and the appropriate ranges of the CCD were performed. The range of rotational speed, milling time, solid reagent amount and solvent to material ratio were determined based on preliminary single factor experiments.

Model Fitting
The values of responses (yield of flavonoids) under different experimental combinations are given in Table 1. The significance of each coefficient was determined using the p-value ( Table 2). The corresponding variables would be more significant if the p-value becomes smaller. It was found that the variables rotational speed (X 1 ), grinding time (X 2 ), solvent to material ratio (X 3 ), amount of solid reagents (X 4 ) and their quadratic parameters were highly significant at the level of p < 0.01. The interactions of X 1 X 2 and X 3 X 4 were highly significant (p < 0.01), and X 2 X 3 were significant (p < 0.05). The regression model can be described by the following quadratic polynomial: (1) Note: * p < 0.05 significant, ** p < 0.01 highly significant.

Analysis of Variance
The coefficients of the above Equation (1) were calculated, and the linearity and quadratic effect of the treatment variables, their interactions and coefficients on the response variables were obtained by analysis of variance (ANOVA, Table 2). For each term in the model, a small p-value (p < 0.05) and a large F-value would imply a more significant effect on the extraction yield [20]. The linear coefficients (X 1 , X 2 , X 3 and X 4 ), quadratic term coefficients (X 1 X 1 , X 2 X 2 , X 3 X 3 and X 4 X 4 ) and interaction coefficients (X 1 X 2 , X 2 X 3 , and X 3 X 4 ) were significant, with very small p-values (p < 0.05). The determination coefficient (R 2 = 0.9962) of the quadratic regression model indicated that only 0.38% of the total variations were not explained by the model. The value of the adjusted determination coefficient (Adj-R 2 = 0.9926) also confirmed that the model was highly significant, which indicated good agreement between the experimental and predicted values of flavonoid yield. The analysis of error results indicated that the lack of fit test (0.1050) was insignificant at the 95% confidence level, confirming the validity of the model. Moreover, the model p-value (Prob > F) was very low (<0.00001), indicating that the model terms were significant.

Analysis of Response Surfaces and Optimal Processing Conditions
Through the 3D plots and their respective contour plots, it was very easy to understand the interactions between two variables and to determine their optimum levels ( Figure 3). Figure 3a,f shows the 3D graphic surface and contour plot of the combined effects of rotational speed and grinding time (X 1 X 2 ), and ratio of solvent to material and amount of solid reagent (X 3 X 4 ). The tortuous surface and oval contour plot show a very strong interaction between the factors (X 1 and X 2 , X 3 and X 4 , p < 0.01). The effects of grinding time and ratio of solvent to material on extraction yield are shown in Figure 3d, where interaction between the two factors was also strong (p < 0.05). As shown in Figure 3b,c,e, the interactions between rotational speed and ratio of solvent to material (X 1 X 3 ), rotational speed and amount of solid reagent (X 1 X 4 ), grinding time and amount of solid reagent (X 2 X 4 ) were not significant. On the basis of the response surfaces, the optimal conditions were determined as follows: rotational speed of 439.74 rpm, grinding time of 17.22 min, ratio of solvent to material of 25:1, amount of solid reagent of 14.97%. Under the optimal conditions, the model gave a maximum predicted value of 35.36 mg/g. For operational convenience, the optimal extraction parameters were set as a rotational speed of 440 rpm, grinding time of 17 min, ratio of solvent to material of 25:1 and amount of solid reagent of 15%. Triplicate experiments were performed under the determined conditions and they yielded 35.17 mg/g, in agreement with the predicted value, indicating that the model was adequate for the extraction process.

The Selectivity Analysis by HPLC-MS/MS
In Figure 4 lines a-b correspond to the conventional heating extraction (CHE) and MPET extracts, respectively. The peaks appearing in b were essentially identical to those in a, suggesting that MPET could extract the same kinds of flavonoids as CHE. However, some differences were observed. The peak-area gaps between 4, 5, 6 and 1, 2, 3 in trace b were bigger than that in a. To investigate the mechanism that caused this peak-area gap, peaks 4, 5, 6 were identified by comparing their UV absorbance and MS data based on some references [21,22]. The UV λmax of peaks 4, 5, 6 had the

The Selectivity Analysis by HPLC-MS/MS
In Figure 4 lines a-b correspond to the conventional heating extraction (CHE) and MPET extracts, respectively. The peaks appearing in b were essentially identical to those in a, suggesting that MPET could extract the same kinds of flavonoids as CHE. However, some differences were observed. The peak-area gaps between 4, 5, 6 and 1, 2, 3 in trace b were bigger than that in a. To investigate the mechanism that caused this peak-area gap, peaks 4, 5, 6 were identified by comparing their UV absorbance and MS data based on some references [21,22]. (See Supplementary Materials), which allowed their identification as kurarinol, kushenol I/N, and kurarinone, respectively, while peaks 1, 2, 3 had relatively low UV absorptions at 220 nm, 209 nm and 223 nm, and were determined them to be alkaloids according to the typical UV and MS data of alkaloids [23]. This indicated that MPET could selectively extract the flavonoid components rather than alkaloids.  Supplementary Materials), which allowed their identification as kurarinol, kushenol I/N, and kurarinone, respectively, while peaks 1, 2, 3 had relatively low UV absorptions at 220 nm, 209 nm and 223 nm, and were determined them to be alkaloids according to the typical UV and MS data of alkaloids [23]. This indicated that MPET could selectively extract the flavonoid components rather than alkaloids.

The DPPH Radical Scavenging Activity
Scavenging of DPPH radicals is a common antioxidant assay used to determine the antioxidant activities of compounds. As shown in Figure 5a, the flavonoids prepared by MPET and CHE showed concentration dependent radical scavenging effects, although they were weaker than those of vitamin C (Vc) in the same concentration. Apparently, flavonoids obtained by MPET displayed relatively higher antioxidant activity than those obtained by CHE. It was supposed that MPET had better selectivity for the extraction of flavonoids, resulting in higher purity, which is better for the antioxidant activities.

The DPPH Radical Scavenging Activity
Scavenging of DPPH radicals is a common antioxidant assay used to determine the antioxidant activities of compounds. As shown in Figure 5a, the flavonoids prepared by MPET and CHE showed concentration dependent radical scavenging effects, although they were weaker than those of vitamin C (Vc) in the same concentration. Apparently, flavonoids obtained by MPET displayed relatively higher antioxidant activity than those obtained by CHE. It was supposed that MPET had better selectivity for the extraction of flavonoids, resulting in higher purity, which is better for the antioxidant activities.  Supplementary Materials), which allowed their identification as kurarinol, kushenol I/N, and kurarinone, respectively, while peaks 1, 2, 3 had relatively low UV absorptions at 220 nm, 209 nm and 223 nm, and were determined them to be alkaloids according to the typical UV and MS data of alkaloids [23]. This indicated that MPET could selectively extract the flavonoid components rather than alkaloids.

The DPPH Radical Scavenging Activity
Scavenging of DPPH radicals is a common antioxidant assay used to determine the antioxidant activities of compounds. As shown in Figure 5a, the flavonoids prepared by MPET and CHE showed concentration dependent radical scavenging effects, although they were weaker than those of vitamin C (Vc) in the same concentration. Apparently, flavonoids obtained by MPET displayed relatively higher antioxidant activity than those obtained by CHE. It was supposed that MPET had better selectivity for the extraction of flavonoids, resulting in higher purity, which is better for the antioxidant activities.

Hydroxyl Radical Scavenging Activity
Hydroxyl radical, which is well known as one of the most reactive free radicals, can react with almost all biomacromolecules in living cells and induce severe damages. As shown in Figure 5b, flavonoids obtained by MPET displayed relatively higher antioxidant activity than that of CHE. At the concentration of 0.6 mg/mL, the scavenging ability of MPET was 97.01%, which was slightly better than Vc (95.7%), so the flavonoids from S. flavescens might be good antioxidants for the functional food industries.

Morphology Analysis
Morphology changes were observed by scanning electron microscopy (SEM). It could be clearly seen that the shattered powder (Figure 6a) had some intact tissues and closed cells. However, powders after grinding treatment (Figure 6b) had a relatively homogeneous small particle size. Integration with cracks on the particle surface and agglomeration phenomena occurred. Mechanical force could not only result in a diminishing of particle sizes, and breaking of cell walls, but also lead to some physicochemical changes, such as electric charge changes on the particle surface by the strong squeezing and shearing forces, which facilitated the release of substances within the cell.

Hydroxyl Radical Scavenging Activity
Hydroxyl radical, which is well known as one of the most reactive free radicals, can react with almost all biomacromolecules in living cells and induce severe damages. As shown in Figure 5b, flavonoids obtained by MPET displayed relatively higher antioxidant activity than that of CHE. At the concentration of 0.6 mg/mL, the scavenging ability of MPET was 97.01%, which was slightly better than Vc (95.7%), so the flavonoids from S. flavescens might be good antioxidants for the functional food industries.

Morphology Analysis
Morphology changes were observed by scanning electron microscopy (SEM). It could be clearly seen that the shattered powder (Figure 6a) had some intact tissues and closed cells. However, powders after grinding treatment (Figure 6b) had a relatively homogeneous small particle size. Integration with cracks on the particle surface and agglomeration phenomena occurred. Mechanical force could not only result in a diminishing of particle sizes, and breaking of cell walls, but also lead to some physicochemical changes, such as electric charge changes on the particle surface by the strong squeezing and shearing forces, which facilitated the release of substances within the cell.

XRD Analysis
The XRD diagrams of a physical mixture and a mechanochemical mixture of S. flavescens powder with Na2CO3 are shown in Figure 7a

XRD Analysis
The XRD diagrams of a physical mixture and a mechanochemical mixture of S. flavescens powder with Na 2 CO 3 are shown in Figure 7a . X-ray diffraction spectra of (a) physical mixture; (b) mechanochemical mixture.

UV-Vis Diffuse-Reflectance Analysis
As shown in Figure 8, lines a, b show the UV-Vis diffuse-reflectance spectra of the physical mixture and mechanochemical mixture, respectively. Compared to a, curve b had more absorption and a slight red shift, suggesting that neutralization reactions in the solid state might occur, which was consistent with Figure 2. It is well known that mechanical treatment could crack the plant cell walls, distort its tissues, and lead to the reactions between solid reagents and active compounds, which would change the surface properties of the plant powders and give rise to the different reflectivity.  Table 3 summarizes an overall comparison of the characteristics of different extraction methods. MPET gave the highest flavonoids content of 4.76% from S. flavescens. However, poor selectivity can be determined from the lower contents of 1.98%, 1.73%, and 2.01% seen in the traditional extraction process, no matter whether ultrasonic or microwave techniques, or conventional heating were applied. MPET could extract flavonoids using water at 25 °C, while conventional heating, ultrasonic or microwave techniques needed to use different ethanol/water cosolvents to extract flavonoids at higher temperature. Furthermore, MPET also displayed obviously advantages over ultrasonic treatment and conventional heating method in both extraction time and flavonoid yields. With the extraction time of 29.5 min, the ultrasonic technique could obtain a yield of 33.56 mg/g, and CHE could only

UV-Vis Diffuse-Reflectance Analysis
As shown in Figure 8, lines a, b show the UV-Vis diffuse-reflectance spectra of the physical mixture and mechanochemical mixture, respectively. Compared to a, curve b had more absorption and a slight red shift, suggesting that neutralization reactions in the solid state might occur, which was consistent with Figure 2. It is well known that mechanical treatment could crack the plant cell walls, distort its tissues, and lead to the reactions between solid reagents and active compounds, which would change the surface properties of the plant powders and give rise to the different reflectivity. substances' status, and might promote the neutralization reaction between flavonoids and the solid basic reagents.

UV-Vis Diffuse-Reflectance Analysis
As shown in Figure 8, lines a, b show the UV-Vis diffuse-reflectance spectra of the physical mixture and mechanochemical mixture, respectively. Compared to a, curve b had more absorption and a slight red shift, suggesting that neutralization reactions in the solid state might occur, which was consistent with Figure 2. It is well known that mechanical treatment could crack the plant cell walls, distort its tissues, and lead to the reactions between solid reagents and active compounds, which would change the surface properties of the plant powders and give rise to the different reflectivity.  Table 3 summarizes an overall comparison of the characteristics of different extraction methods. MPET gave the highest flavonoids content of 4.76% from S. flavescens. However, poor selectivity can be determined from the lower contents of 1.98%, 1.73%, and 2.01% seen in the traditional extraction process, no matter whether ultrasonic or microwave techniques, or conventional heating were applied. MPET could extract flavonoids using water at 25 °C, while conventional heating, ultrasonic or microwave techniques needed to use different ethanol/water cosolvents to extract flavonoids at higher temperature. Furthermore, MPET also displayed obviously advantages over ultrasonic treatment and conventional heating method in both extraction time and flavonoid yields. With the extraction time of 29.5 min, the ultrasonic technique could obtain a yield of 33.56 mg/g, and CHE could only  Table 3 summarizes an overall comparison of the characteristics of different extraction methods. MPET gave the highest flavonoids content of 4.76% from S. flavescens. However, poor selectivity can be determined from the lower contents of 1.98%, 1.73%, and 2.01% seen in the traditional extraction process, no matter whether ultrasonic or microwave techniques, or conventional heating were applied. MPET could extract flavonoids using water at 25˝C, while conventional heating, ultrasonic or microwave techniques needed to use different ethanol/water cosolvents to extract flavonoids at higher temperature. Furthermore, MPET also displayed obviously advantages over ultrasonic treatment and conventional heating method in both extraction time and flavonoid yields. With the extraction time of 29.5 min, the ultrasonic technique could obtain a yield of 33.56 mg/g, and CHE could only get a yield of 33.87 mg/g with an extended extraction time of 120 min, which was time consuming. Although a slightly higher yield was achieved by the microwave method, 60% ethanol was required as extraction solvent at an elevated microwave power setting power (420 W), which gave rise to higher cost and energy consumption. It was thus proposed that the shear force and instant high pressure during mechanochemical treatment that caused the breakage of cell walls, lowered diffusion hindrances and finally initiated mechanochemical reactions to afford soluble salts of the target compounds, efficiently facilitated the extraction process and increased the extraction efficiency and yield. Data are presented as mean˘SD (n = 3). a microwave power 420 W.

Materials and Reagents
S. flavescens roots were purchased from Zhejiang CONBA Pharmaceutical Co., Ltd. (Hangzhou, China) and shattered by a HC-500T2 pulverizer (Song Qing Hardware Factory, Yongkang, China) to an average particle size of 0.5 mm. The powder was then stored in a dry place at room temperature. Standard rutin, analytical grade reagents sodium carbonate and citric acid were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Analytical-grade reagents were purchased from Tianjin Yongda Chemical Reagent Development Centre (Tianjin, China) and HPLC grade solvents were purchased from Tedia Company Inc. (Fairfield, CA, USA).

Mechanochemical-Promoted Extraction Technology (MPET)
In z preliminary study, various kinds of solid reagents such as sodium borate, diatomaceous earth, basic aluminum oxide, Ca 2 CO 3 , Na 2 CO 3 and NaHCO 3 were screened at an excess dosage of 25% (wt %, mass ratio of solid reagents to S. flavescens particles). The extraction procedure was as follows: S. flavescens roots (10.0 g), solid reagents, and 72 g of stainless steel balls with 12 mm diameter were added into a 50 mL vial (PM-200 planetary mill, Retsch, Haan, Germany). After co-grinding at 400 rpm for 10 min, the powders were extracted with water for 20 min and then centrifuged at 3077 g for 10 min. The solution pH was adjusted to 4-5 with citric acid. The solution was condensed, centrifuged at 9391 g, the supernatant was discarded and the precipitants were analyzed by ultraviolet spectrophotometry (UV-2550 PC, Shimadzu, Kyoto, Japan) and HPLC/MS (Agilent 1100 HPLC system, Santa Clara, CA, USA) consisting of a Surveyor autosampler, pumps and photodiode array detector connected to an LCQ-Advantage ion trap mass spectrometer (Thermo Finnigan, San Francisco, CA, USA) fitted with an ESI source. Acidification pH value was optimized by acidifying the extracted solution to the pH values of 10 to 4, and then analyzing the extracts.

Conventional Heating Extraction (CHE)
According to the optimal conditions, S. flavescens roots (10.0 g) were refluxed with ethanol-water (200 mL, 80:20, v/v) solution at 85˝C for 1 h, and then the mixture was centrifuged for 10 min at 3077 g.
The process was repeated two times, the supernatants were combined, condensed and analyzed by ultraviolet spectrophotometry and HPLC.

Experimental Design
RSM was used to determine the optimal MPET conditions for the extraction of flavonoids from S. flavescens roots. To explore the effect of independent variables on the response within the range of investigation, a central composite rotate design with four independent variables (X 1 , rotational speed; X 2 , grinding time; X 3 , solvent to material ratio and X 4 , amount of solid reagents) at five levels [24] was performed with some modifications, as shown in Table 4. The variables were coded according to Equation (1): x = (X i -X 0 )/∆X i , where Xi is the coded value of an independent variable, X i is the real value of the independent variable, X 0 is the real value of an independent variable at the centre point, and ∆X i is the step change value. The yield of flavonoids was considered as the dependent variable or response. For a central composite rotate design with four independent variables at five levels, 30 experimental runs are required. The actual design of experiments is given in Table 2. The experimental results were fitted to a second-order polynomial model, and the regression coefficients were determined. The quadratic model for predicting the optimal point was expressed according to Equation β ij X i X j , where β 0 , β i , β ii and β ij are constant regression coefficients of the model, while X i , X j are the independent variables.

Total Flavonoids Content
The total flavonoid content from S. flavescens in extracts was determined according to the NaNO 2 -Al(NO 3 ) 3 -NaOH colorimetry with some modifications [25]. The reaction mixture contained 2.0 mL of extract, 6 mL of 60% ethanol, and 1 mL of 5% sodium nitrite. Six minutes later, 1 mL of 10% aluminum nitrite was added. In the next six minutes, 10 mL of 1 M sodium hydroxide solution were added and the volume was adjusted to 25 mL by adding 60% ethanol. Immediately, the reaction mixture absorbance was measured by UV spectrophotometry at 510 nm against a blank (control) and used to calculate yields using rutin as a standard. Measurements were calibrated to a standard curve y = 3.239x´0.020, (R 2 = 0.999). Flavonoid content is calculated as follows: Yield " flavonoid content of extracts pmgq{weight of S. flavescens powder pgqˆ100%

HPLC-MS Analysis
The chromatographic separation was performed on XB-C 18 column (4.6 mmˆ250 mm, 5 µm, Welch, Shanghai, China) at 30˝C. Solvent A (formic acid) and solvent B (acetonitrile) were selected as the mobile phases at the flow rate of 1.0 mL/min. The linear gradient elution started with 0 min, 90% A; 20 min, A 60%; 40 min, 40% A; 60 min, 5% A. The signal was monitored at 280 nm using the diode array detector. MS analysis was carried out in the positive ion mode recorded over a mass range of 120-500 m/z. Nitrogen (N 2 ) was used as the sheath and auxiliary gas, and helium (He) was used as the collision gas. Capillary voltage was 24 V, capillary temperature was 300˝C, and sheath gas flow rate was 40 mL/min.

Assay of DPPH Radical Scavenging Activity
The scavenging effect of the flavonoids extracted by MPET and CHE on DPPH radical was compared basing on Brand-Williams et al. [26] with some modifications. Briefly, 2.0 mL of 0.2 mM DPPH in anhydrous ethanol was added to 2.0 mL of sample (at different concentrations). The mixture was shaken and incubated for 30 min at room temperature in dark, and the absorbance of the resulting solution was measured at 517 nm. Ascorbic acid was used as control substance. The scavenging percentage was calculated according to the following equation: Scavenging percentage % " pA 0`A1´A2 q{A 0ˆ1 00% where A 2 was the absorbance of the test sample; A 0 was the absorbance of the control group; and A 1 was the absorbance of 2.0 mL sample in 2.0 mL anhydrous ethanol.

Assay of Hydroxyl Radical Scavenging Activity
Hydroxyl radical scavenging activity was determined according to the literature with some modifications [27]. One mL sample solutions of different concentration, 0.5 mL of salicylic acid-ethanol solution (9.0 mM), 0.5 mL of FeSO 4 solution (9.0 mM) and 3.0 mL of distilled water were successively mixed in a tube. The reaction was initiated by the addition of 2.0 mL H 2 O 2 (8.8 mM) to the mixture above, and the absorbance at 510 nm was recorded. The hydroxyl radical scavenging activity was calculated as follows: Scavenging percentage % " pA 0`A1´A2 q{A 0ˆ1 00% where A 2 was the absorbance of the test sample; A 0 was the absorbance of the control group (deionized water instead of sample); and A 1 was the absorbance of water instead of H 2 O 2 .

Physicochemical Property Analysis
The morphology of the samples with different pretreatments was observed using scanning electron microscopy (S-4700, Hitachi, Tokyo, Japan). The X-ray powder diffraction (XRD) analysis was carried out on a ARL X'TRA X-ray diffractometer (Thermo, Waltham, MA, USA). Diffuse-reflectance UV-Vis spectra were acquired with the Shimadzu UV-2600 spectrometer (Shimadzu).

Statistical Analysis
The results were expressed as means of yield˘SD (standard deviation). Design-Expert 8.0.6 (Stat-Ease Inc., Minneapolis, MN, USA) was used to calculate the coefficients of the quadratic polynomial model and the optimization. p values of less than 0.05 were considered to be statistically significant.

Conclusions
The present study showed that MPET is a high efficiency and green method to selectively extract flavonoids from S. flavescens. The extraction variables were optimized through RSM, whereby a rotational speed of 440 rpm, grinding time of 17 min, solvent to material ratio of 25 mL/g and amount of solid reagents of 15% were proved to be the best conditions to maximize the total yield of flavonoids. With regard to antioxidant activity results, a higher radical scavenging activity was obtained for