Optimization for Ultrasonic-and Microwave-assisted Extraction of Flavonoids from Burdock ( Arctium lappa L . ) Root by Response Surface Methodology

An efficient Ultrasonic and Microwave Assisted Extraction (UMAE) was developed to extract total flavonoids from the Arctium lappa L. boot. Response Surface Methodology (RSM) combined Box-Behnken Design (BBD) was employed to optimize extraction condition based on the yield of total flavonoids. The optimal condition was identified: ethanol concentration, 59% (v/v); liquid/ratio ratio, 50 mL/g; ultrasound time, 3000 sec and microwave power 480 W. The predicted total flavonoids extraction yield of 0.473 g RUE/100 g DW was obtained under optimized UMAE conditions.


INTRODUCTION
Arctium lappa Linn.(burdock), called Niubang in Chinese, is an edible perennial herb of Asteraceae family.The plant has been used as a traditional medicine and a dietary vegetable for a period of long time by the Chinese civilization (Morita et al., 1993).It is also popular in North America, Europe and Asia for over thousands of years due to its therapeutic qualities.Although burdock leaves, fruit and seeds all can be used, the dried burdock root is the main part used for different therapeutic intentions, such as throat pain, tonsillitis, rashes, arthritis, blood purifier and various skin diseases (Chan et al., 2011b).These pharmacological effects are positively believed to be related to the fact that burdock root is rich in biologically active substances (Han et al., 2013;Tousch et al., 2014).Of all the compounds presented in burdock root, the flavonoids components were considered to be critical to their pharmacological activities (Miyamoto et al., 1993;Tamayo et al., 2000) and have received most attention by food manufacturers.The extraction efficiency of flavonoid components is influenced by many factors including the extraction technology (Liu et al., 2010;Mao et al., 2008).However, there is little information concerning the optimization of extraction of the flavonoids in burdock boot by modern extraction techniques.
Microwave-Assisted Extraction (MAE) is the process by which microwave energy is used to heat solvents in contact with solid samples and to partition compounds of interest from the sample into the solvent (Spigno and De Faveri, 2009).In recent years, MAE has received a great attention as a potential prospective technique to replace conventional extraction methods, mainly due to considerable savings in processing time, solvent consumption and energy (Camel, 2000;Chan et al., 2011a).Ultrasonic and has been widely used as one of the most industrially technique because of its potential in improving the extraction effects (Matsumoto et al., 2014).Recently, simultaneous Ultrasonic/Microwave Assisted Extraction (UMAE) presented many benefits by combining the vantage of ultrasonic and microwave (Lou et al., 2012), However, to the best of our knowledge, there are no studies related to simultaneous UMAE of TF from burdock boot.
Therefore, in the current study the total flavonoids were extracted from burdock boot.The purpose of this study was to optimize the operational conditions, including the liquid/solid ratio, the ethanol concentration, the ultrasound time and the microwave power by applying RSM.The response variable was examined based on the total flavonoids yield under different extraction conditions.

Materials:
Burdock root was purchased in Hongqi agricultural market (Tianjin, China).The samples were washed and dried.A relatively powder was obtained by ground with a blade-mill and sieved through the 60mesh sifter.The powder was kept in 4°C until use.
Rutin were purchased from Sigma Chemical Company.The other chemicals were of reagent grade.

UMAE extraction procedure:
The process of UMAE of burdock was carried out in a microwave extraction apparatus (CW-2000, Shanghai XTrust Instruments Company, China) equipped with a 250 mL quartz vessel and a cool water circulation system.The dried sample was placed in a round bottom flask with different volume of extraction solvent according to the BBD.After the extraction procedure, the leaves extracts were filtered through Whatman filter paper.The filtrate was transferred into a volumetric flask and diluted to 100 mL for quantitative analysis.

Determination of total flavonoids:
The determination of TF in the extracts was performed according to Shao et al. (2012).Briefly, 6.0 mL extracts solution was added separately with 1 mL 5 wt% NaNO 2 .One mL 10 wt% Al 2 (NO 3 ) 3 , 10 mL 5 wt% KOH.After mixing, the absorbance of the sample was measured at 500 nm by an ultraviolet spectrophotometer (UV-5200, Shanghai Metash instruments Company, China).Rutin standards was employed to prepare the calibration curve (y = 9.7868x + 0.0263, R 2 = 0.9908), where y is the absorbance at 500 nm and x is the concentration of flavonoids (mg/mL).TF yield of the samples was calculated using the calibration standard and expressed as rutin equivalents per 100 g dry weight (g RUE/100 g DW).
Box-Behnken design: According to a BBD with 4variable at 3-level on the TF yield, optimization of the UMAE process was investigated.Based on previous single-factor experiments, four experimental parameters, including Microwave Power (MP), Solid/Liquid Ratio (SLR), Ethanol Concentration (EC) and Ultrasound Time (UT), were chosen as independent variables with ranges of 480-600 W, 30-50 mL/g, 50-70% and 1800-3000 sec, respectively.Multiple regression was used to fit the quadratic model by analyzing data from the BBD.The second-order model equation for the response variable was as follows: where, Y is the response, β 0 is a constant and β i , β ii and β ij represent the linear, quadratic and interactive regression coefficients, respectively.X i and X j were the levels of the independent coded variables.Table 1 shows the coded levels of the independent variables and the parallel parameter values.

RESULTS AND DISCUSSION
Fitting the model: In this study, four effectual variables (MP, EC, LSR, UT) that affect UMAE of burdock boot on yield of TF were optimized using BBD.All data received from 29 experimental runs (Table 2).The highest yield of TF was obtained using treatments 22 with the values of 0.460 g RUE/100 g DW and the lowest yield of TF were obtained using treatments 5 with the values of 0.344 g RUE/100 g DW.
After the regression analysis of the data shown in Table 3, the second-order polynomial equation developed for TF in terms of coded units was as follows:   3, the model of the TF extract yield were statistically significant (F p<0.0001).In addition, the model with R considered appropriate (Yang et al., 2010) small value of CV indicated that the variation of the mean value is low and can preferably develo   3, the model of the TF extraction yield were statistically significant (F-value 20.13, p<0.0001).In addition, the model with R 2 >0.75 was ., 2010) and the small value of CV indicated that the variation of the mean value is low and can preferably develop an adequate response model (Liyana Shahidi, 2005), in this study the correlation coefficient (R 2 = 0.7789) and the coefficient of variation (4.41%) demonstrated that the response predicted model was suitable for the actual situation.It was that the linear term of LSR and quadratic term of EC have large significant effect on the yield of TF because of the high F-value of 18.33 and 23.88, respectively.The results indicated that the influence variables did not have a simple linear relationship.According to the Significant Not significant (Liyana-Pathirana and , in this study the correlation coefficient = 0.7789) and the coefficient of variation (4.41%) demonstrated that the response predicted model was suitable for the actual situation.It was also observed that the linear term of LSR and quadratic term of EC have large significant effect on the yield of TF because value of 18.33 and 23.88, respectively.The results indicated that the influence variables did not ar relationship.According to the analysis, the studied variables influence the response variable in the following order: liquid/solid ratio > microwave power > ultrasound time > ethanol concentration.
Figure 1 depicted the plot of actual values versus predicted values for the estimated model, the relationship between the actual and predicted values shows that the actual points cluster around the diagonal line, which reveals the experimental values was in good agreement with the regression model.

Analysis of response surface:
The mutual interaction of the independent variables on the extraction yield of TF can be seen on 3D response surface plots shown in Fig. 2a to f.The curves were generated by plotting the values of response variable while keeping the other two independent variables at their zero level.The steeper plots illustrated the sensitivity of the response towards the change in the extraction conditions.Otherwise, the observed effects were slight (Bezerra et al., 2008).
Figure 1 shows that the increase in LSR from 30 to 50 mL/g with EC from 50 to 70% increased the yield of TF.While with increase of EC over 60%, there was a gradual decline in the response.These results suggest that the LSR and EC had a quadratic effect on the response, but the mutual interactions between the two variables were not significant.Figure 2b depicts that increase of MP from 450 to 600 W at EC of 60% induce the slight decline of the yield.The yield increased when the EC changed from 50 to 60%, but decreased thereafter.Figure 2c shows that the yield of TF increased rapidly with the increase of LSR at a fixed microwave power, while the influence of the MP on the yield was less significant than the LSR. Figure 2d reveals that no obvious effect was obtained on TF yield with the increase of UT from 1800 to 3000 (s).By contrast, when the EC was lower than 60%, the TF yield increased with the enhancement of the EC, but the yield decreased when the EC exceeded 60%.The effects of the UT and LSR are shown in Fig. 2e.As LSR rose from 30 to 50 mL/g, the yield was increased.The influences of UT with range from 1800 to 3000 sec were not obvious.The highest of TF yield was obtained at 3000 sec UT with 50 mL/g LSR. Figure 2f shows that the TF yield was definitely correlated to the increase of UT, whereas no obvious change in yield was observed when MP increased from 480 to 600 W. The lowest yield of TF could be observed when extraction was performed at 600W MP with 3000 sec UT.
Validation of the model: (0.465 g) RUE/100 g DW of TF yield was obtained under the optimized operating condition (EC of 59%, LSR of 50 mL/g, UT of 3000 sec, MP of 480 W).The experimental yield of TF was in agreement with the predicted value (0.473 g RUE/100 g DW).

CONCLUSION
In this study, RSM and BBD was successfully used to optimize the UMAE process.The second order polynomial model can be applied to optimize the parameters of burdock boot extraction to obtain an extract with high TF yield (0.473 g RUE/100 g DW).
The yield was significantly increased under the optimized conditions.Through this study, we managed to obtain more TF form burdock boot, dramatically increased extraction efficiency, making it possible to guide the industrial production of TF from burdock boot.

Fig. 1 :
Fig. 1: Comparison between predicted and actual values of TF extraction yield ANOVA offers the availability of the quadratic model and could evaluate the goodness of fit.It is said that a regression model would be well fitted to the experimental data if the model has a significant regression.In Table3, the model of the TF extract yield were statistically significant (F p<0.0001).In addition, the model with R considered appropriate(Yang et al., 2010) small value of CV indicated that the variation of the mean value is low and can preferably develo

Fig. 1 :
Fig. 1: Comparison between predicted and actual values of TF extraction yield ANOVA offers the availability of the quadratic model and could evaluate the goodness of fit.It is said that a regression model would be well fitted to the experimental data if the model has a significant regression.In Table3, the model of the TF extraction yield were statistically significant (F-value 20.13, p<0.0001).In addition, the model with R 2 >0.75 was ., 2010) and the small value of CV indicated that the variation of the mean value is low and can preferably develop an

Table 1 :
Box-Behnken design factors and levels of encoded values

Table 3 :
Analysis of Variance (ANOVA) for response surface quadratic model for the extracted protein S.: Sum of square; M.S.: Mean square

Table 3 :
Analysis of Variance (ANOVA) for response surface quadratic model for the extracted protein M.S.