Optimization of the extraction process for the seven bioactive compounds in Yukmijihwangtang, an herbal formula, using response surface methodology
JungHoon Kim, HyeunKyoo Shin, ChangSeob Seo
Herbal Medicine Formulation Research Group, Herbal Medicine Research Division, Korea Institute of Oriental Medicine, Daejeon 305811, Korea
Date of Submission  27Nov2013 
Date of Acceptance  14Jan2014 
Date of Web Publication  30Aug2014 
Correspondence Address: ChangSeob Seo Herbal Medicine Formulation Research Group, Herbal Medicine Research Division, Korea Institute of Oriental Medicine, Daejeon 305811 Korea
Source of Support: This study was supported by a grant from the Korea
Institute of Oriental Medicine (No. K13030).,, Conflict of Interest: None  Check 
DOI: 10.4103/09731296.139798
Abstract   
Background: Yukmijihwangtang (YJT) contains multiple bioactive compounds. Heatreflux extraction was employed and optimized for the extraction of the bioactive compounds in YJT. Objective: The determination of optimal conditions with maximum yields of bioactive compounds, gallic acid, 5hydroxymethylfurfural, morroniside, loganin, paeoniflorin, benzoic acid and paeonol, in YJT. Materials and Methods: The extraction ratio (ratio of water to herbal formula), extraction time and extraction number were set as individual values and the yields of the seven compounds were the response values that were optimized with a BoxBehnken design. Results: The optimal conditions obtained from response surface methodology (RSM) were 1:11.99 for the extraction ratio, 94.53 min for the extraction time and 2.21 for the extraction number. Under the optimal conditions, the response value of the experiment closely agreed with the predicted response value. Conclusions: The result suggests that RSM is successfully applied for optimizing the extraction of the marker compounds in YJT. Keywords: Bioactive compounds, heatreflux extraction, optimal condition, response surface methodology, Yukmijihwangtang
How to cite this article: Kim JH, Shin HK, Seo CS. Optimization of the extraction process for the seven bioactive compounds in Yukmijihwangtang, an herbal formula, using response surface methodology. Phcog Mag 2014;10, Suppl S3:60613 
How to cite this URL: Kim JH, Shin HK, Seo CS. Optimization of the extraction process for the seven bioactive compounds in Yukmijihwangtang, an herbal formula, using response surface methodology. Phcog Mag [serial online] 2014 [cited 2021 Aug 1];10, Suppl S3:60613. Available from: http://www.phcog.com/text.asp?2014/10/39/606/139798 
Introduction   
A herbal formula is prepared by boiling the herbal mixture with water before it is administered to the patients, and most in vivo and in vitro experimental models using a herbal formula as a treating agent have dealt with the water extract produced in the laboratory. ^{[1],[2]} The therapeutic effect of a herbal formula is attributed to the synergistic property that results from the combination and interaction of bioactive constituents from herbal medicines. ^{[3]} Thus, the extraction method must be designed to produce efficiently the bioactive compounds from the herbal formula, so that those compounds can contribute to exert the curative effect.
Heatreflux extraction (HRE) is a conventionally and widely used extraction method for the preparation of herbal medicine, ^{[4],[5],[6]} and it is close to the traditional extraction method of an herbal formula. There are many parameters determining the adequate conditions of an herbal extract, including extraction time, the number of extractions, and ratio of solvent to raw material, extraction temperature and pressure. ^{[7],[8]} In the HRE process, water is boiled at 100°C and the evaporated vapor turns to water droplets in the attached condenser on the flask; hence, the temperature and pressure are not variables to be chosen as extraction parameters.
Yukmijihwangtang (YJT, Liuweidihuangtang in Chinese) is a widely used herbal formula in Korea and China. YJT consists of six herbal medicines including Rehmannia glutinosa Libosch. ex Steudel, Dioscorea batatas Decne., Cornus officinalis Sieb. et Zucc., Paeonia suffruticosa Andrews, Poria cocos F.A. Wolf, and Alisma orientale Juzep. Several pharmacological properties of YJT have been reported, such as renal protection, ^{[9],[10]} regulation against autoimmune encephalomyelitis, ^{[11]} improving learning and memory, ^{[12]} protection against βamyloidinduced paralysis and myelosuppression, ^{[13],[14]} antiobesity ^{[15]} and antioxidant activity. ^{[16]} The main bioactive compounds of YJT are gallic acid, 5hydroxymethylfurfural (5HMF), morroniside, loganin, paeoniflorin, benzoic acid and paeonol which are analyzed using high performance liquid chromatography (HPLC)ultravioletmass spectrometry, HPLCdiode array detector (DAD) or micellar electrokinetic chromatography. ^{[17],[18],[19],[20]}
Response surface methodology (RSM) is a statistical technique to determine the optimum values of the independent variables to achieve the maximum response, and enables the user to investigate the interaction of the individual variables, which is considered more efficient than the traditional single parameter optimization because of the saving in time, space, and raw materials. ^{[21]} For those reasons, RSM has been employed in the extraction of chemical compounds from herbal medicines. ^{[22],[23],[24]}
The aim of this study was to optimize the extraction process for the seven bioactive compounds from YJT using RSM. The extraction factors, ratio of water to herbal formula, extraction time and extraction number, were chosen as the independent variables for the extraction and their influence on the yields of the compounds was studied through a BoxBehnken design (BBD). The content of the bioactive compounds was determined using HPLCDAD analysis with a validated method. To the best of our knowledge, this is the first study on the optimization of chemical components from an herbal formula using RSM.
Materials and Methods   
Chemicals and reagents
High performance liquid chromatographygrade methanol, acetonitrile, and water were purchased from J.T. Baker Inc. (Phillipsburg, NJ, USA). Gallic acid (1), 5HMF (2) and benzoic acid (6) were purchased from SigmaAldrich (St. Louis, MO, USA). Morroniside (3), loganin (4), paeoniflorin (5) and paeonol (7) were obtained from Wako Pure Chemical Industries Ltd (Osaka, Japan). All of the compounds represented a purity of more than 98%. The chemical structures of the standard compounds are shown in [Figure 1].  Figure 1: Chemical structures of standard compounds in Yukmijihwangtang; gallic acid (1), 5hydroxymethylfurfural (2), morroniside (3), loganin (4), paeoniflorin (5), benzoic acid (6) and paeonol (7)
Click here to view 
Compositional herbal medicines were purchased from the herbal medicine company, Kwangmyungdang Medicinal Herbs (Ulsan, Korea) [Table 1]. Herbal medicines were identified by Professor JeHyun Lee (Department of Herbology, Dongguk University, Korea) and Young Bae Seo (Department of Herbology, Daejeon University, Korea). A voucher specimen (2013KE0716) has been deposited in the Herbal Medicine Formulation Research Group of the Korea Institute of Oriental Medicine.
Extraction procedure of Yukmijihwangtang
The herbal medicine mixture consisting of YJT was extracted with a 10fold volume of distilled water (w/v) by boiling using reflux extractor. The extracted decoction was centrifuged at 3000 rpm for 10 min and the supernatant was lyophilized to create powder.
Accurately weighed powders of YJT water extract (10 mg) were dissolved in 1 mL of HPLC gradewater and the solutions were filtered through a 0.2 μm syringe filter (SmartPor ^{®} , Woongki Science, Seoul, Korea) prior to HPLC analysis.
Chromatographic conditions
The analysis was carried out using a Hitachi HPLCDAD system equipped with a solvent delivery unit, autosampler, column oven, and diodearray detector. The acquired data were processed using EZChrome Elite for Hitachi. Separation was performed on a Gemini C _{18} column (4.6 mm × 250 mm, 5 μm; Phenomenex, Torrance, CA, USA) at 35°C. The mobile phase, consisting of solvent A (1% aqueous acetic acid, v/v) and solvent B (acetonitrile with 1% acetic acid, v/v), was eluted using the gradient procedure, which was as follows: 540% (B) over 030 min, 40100% (B) over 3040 min, held for 5 min, and then reequilibrated to 5% for 15 min. The flow rate was 1.0 mL/min and the injection volume was set to 10 μL. The optimized detection wavelengths for standard compounds were set at 230, 272, and 280 nm.
Method validation
Accurately weighed standard compounds were dissolved in methanol at concentrations of 1000 μg/mL to produce a stock solution containing the seven standard compounds. The stock solution was diluted at five levels to make working solutions that were used to construct calibration curves in which the xaxis was the concentration of marker compound and the yaxis was the area of the marker compound. Linear regression and the coefficient of determination (r^{2} ) of the compounds were calculated based on the calibration curves. The values of limits of determination (LOD) and limits of quantification (LOQ) were evaluated from the concentrations of each compound at signaltonoise ratios of 3 and 10, respectively.
The precisions were measured by analyzing sample extracts at two concentrations of standard compounds of low and high levels on same day (intraday) and three successive days (interday), which is represented by the values of the RSD. The recovery test that was used to evaluate the accuracy of the method was determined by assessing two different concentration levels of spiked compounds (low and high) for the samples. The recovery was calculated as follows:
Recovery (%) = ([Detected concentration  initial concentration]/ Spiked concentration) × 100
Experimental design and statistical analysis
To determine the optimum condition for extraction of YJT, the preliminary range of the extraction variables, extraction ratio (ratio of water to the herbal formula), extraction time and the number of extractions, were investigated using a singlefactor test. A threelevelthreefactor BBD was employed to determine the optimal conditions for the extraction of the seven bioactive compounds in YJT.
Experimental data obtained from the BBD were fitted to a secondorder polynomial model and the regression coefficients were obtained. The equation is as follows:
Where Y is the estimated response, β_{0} , β_{j} , β_{jj} and β_{ij} are the regression coefficients for intercept, linearity, square and interaction terms, respectively. X _{i} and X _{j} are the independent variables, which were coded.
The fitness of the secondorder polynomial model was expressed by the lack of fit and coefficient of determination (r^{2} ). Ftest and P values resulting from the analysis of variance (ANOVA) were calculated to confirm the significance of the regression coefficients, which was determined at P < 0.05 or 0.01. The interaction and influence of the three variables on the yield of the bioactive compound was represented as threedimensional response surface plots and contour plots, on which the optimal extraction condition was observed. The opensource software R (ver. 2.15.1; The R Foundation for Statistical Computing) was used to generate the experimental design, statistical analysis and regression model.
Results and Discussion   
Method validation
Using the developed HPLC methods, all of the bioactive compounds were welldetected and selective without any interference from endogenous constituents on chromatograms at their maximum absorption wavelengths [Figure 2]. On the basis of the calibration curves, the coefficient of determination (r^{2} ) ranged from 0.9992 to 1.0000 for all analytes, which means good linearity. The ranges of LODs and LOQs were 0.010.09 μg/mL and 0.040.30 μg/mL, respectively [Table 2]. The precisions of the seven bioactive compounds represented as RSD values were 0.050.51% for intraday precision and 0.011.10% for interday precision at two levels of concentrations [Table 3]. The recoveries of the seven marker compounds were in the range of 91.25107.91%, with RSD values <4.1% over the concentration ranges [Table 4].  Figure 2: Chromatogram of standard compounds (a) and Yukmijihwangtang water extract (b) at their optimum wavelength; gallic acid (1), 5hydroxymethylfurfural (2), morroniside (3), loganin (4), paeoniflorin (5), benzoic acid (6), and paeonol (7)
Click here to view 
 Table 2: Linear equations, coefficients of determination (r^{2}), LOD and LOQ for the bioactive compounds
Click here to view 
Model fitting
Preliminary experiments using singlefactor tests determined the required range of ratio of water to herbal formula (X _{1} , 1:81:16), extraction time (X _{2} , 60120 min) and the number of extractions (X _{3} , 13 repeats). A threelevelthreefactor BBD comprising the 15 experiments listed in [Table 5] was employed in this study, in which three replicates (runs 7, 9 and 11) were used to measure the pure error sum of squares. The sum of the yields of the seven marker compounds was treated as the response. The three factors used in this study were represented as three coded levels (1, 0, 1) for each factor.  Table 5: Box  Behnken design and the response values for yields of compounds
Click here to view 
With the help of multiple regression analysis on the experimental data, the predicted response value was expressed by the following secondorder polynomial equation using coded variables:
Where Y is the yield of the seven compounds (mg/g), and the coded variables X_{1}, X_{2} and X_{3} represent the ratio of water to herbal formula, extraction time and extraction number, respectively.
An adequately fitted model can help the exploration and optimization of a fitted response surface, provide an adequate approximation to the true system and verify that none of the least squares regression assumptions are violated. ^{[25]} ANOVA was performed for the fitted quadratic polynomial model for extraction of the seven bioactive compounds [Table 6]. The coefficient of determination (r^{2} ) was 0.9278 with no significant lack of fit at P > 0.05, indicating that the predicted model could explain 92.78% of the results and only 7.22% of the total variance was not explained by the model.  Table 6: ANOVA for the fitted quadratic polynomial model for the extraction of compounds
Click here to view 
The significance of the model was evaluated using the Fvalue and P value, where the corresponding variables are more significant for larger Fvalues and smaller P values. ^{[26]} The Fvalue of 7.1410 and P value of 0.02169 imply that the model used to fit the response was significant and adequately represented the predicted results between the independent variables and the response. ^{[27]}
The regression coefficients of the predicted quadratic polynomial model were obtained for the coded variables and the significance of each coefficient was determined using Student's ttest and the P value, in which a larger tvalue and smaller P value show the significance of the corresponding coefficient. ^{[25]} It was observed that the extraction time was significant in both linear (P < 0.05) and quadratic terms (P < 0.01), whereas the ratio of water to herbal formula was verified to be significant only for the quadratic term (P < 0.05). The other term coefficients (X_{3} , X_{1} :X_{2}, X_{1} :X_{3} , X_{2} :X _{3}, and X_{3} :X_{3} ) were not significantly influential on the model (P > 0.05) [Table 7].  Table 7: Regression coefficients of the predicted quadratic polynomial model
Click here to view 
Analysis of response surface
The polynomial equation obtained from regression analysis was graphically visualized by a threedimensional response plot and twodimensional contour plots, where the interaction between variables and the effect of variables on the response can be observed. RSM plays a key role in determining the optimum values of the independent valuables that produce the maximum response.^{ [21]} The threedimensional response plots and contour plots were obtained using two independent variables, while keeping the other variable set at the zero level. The interactions between the variables were determined through the shape of the contour plots. An elliptical contour plot indicates that the interaction between the variables is significant, while a circular contour plot means negligible interaction. ^{[28]}
As shown in [Figure 3], the interaction between the ratio of water to herbal formula (X_{1} ) and extraction time (X_{2} ) is shown with the extraction number (X_{3} ) set at the zero level in the response plot and contour plot. The yield of marker compounds increased with increasing ratio of water to herbal formula from 1:8 (200 mL) to 1:12 (300 mL) and increasing time of extraction from 60 min to 100 min. However, it was observed that the effect of the ratio had less influential on the yield than that of extraction time in the contour plot. The yield reached the maximum value of 13.6 mg/g when the ratio and extraction time were 1:11.9 and 94 min, respectively; however, there was a gradual decline in the response beyond those levels of the variables.  Figure 3: Response surface plot and contour plot of the ratio of water to herbal formula (mL, X1) and extraction time (min, X2)
Click here to view 
The response plot and contour in [Figure 4] shows the interaction between ratio (X _{1} ) and extraction number (X_{3} ) with the extraction time (X_{2} ) set at the zero level. It was found that increasing the ratio from 1:8 to 1:12 and increasing the extraction number from 1 to 2.5 increased the yield of the compounds, and the maximum value of the yield was observed within those levels.  Figure 4: Response surface plot and contour plot of extraction volume (mL, X1) and extraction number (X3)
Click here to view 
[Figure 5] describes the effect of extraction time (X_{2} ) and extraction number (X_{3} ) on the yield of compounds and the interaction between the two variables when the other variable (X_{1} ) was kept at the zero level. The yield of compounds increased as the extraction time and extraction number increased, and the extraction time contributed to the increase in yield more than the extraction time in the contour plot. The highest level of yield was obtained at an extraction time of 95 min and extraction number of 2.5. The interactive effect of extraction time and extraction number on the yield of the compounds was not shown to be very weak (P = 0.07234) [Table 7].  Figure 5: Response surface plot and contour plot of extraction time (time, X2) and extraction number (X3)
Click here to view 
As shown in [Figure 3] and [Figure 5], and [Table 7], the extraction time obviously affected the yield of compounds (P = 0.02940), but rather excessive extraction time could decrease the yield, which can be explained by the increasing extraction time accelerating chemical decomposition of marker compounds during the extraction process, resulting in reduced extraction yield. ^{[29]}
Optimization and verification of extraction by Response surface methodology
The aim of this study was to determine the optimal conditions producing the maximum extraction yield of chemical compounds from YJT. The conditions producing the maximum extraction of the compounds in YJT were determined based on a polynomial equation. The optimal condition of ratio of water to herbal formula, extraction time and extraction number was 1:11.99 (299.69 mL), 94.53 min and 2.21 repeats, respectively. The optimized extraction yield of total compounds was predicted to be 13.66 mg/g, which is very close to the actual value of 13.68 (mg/g) determined from the modified conditions [Table 8]. These results confirm that the model for the extraction of compounds from YJT was able to predict the experimental conditions.  Table 8: Optimum conditions and the predicted and experimental values of the response at the optimum conditions
Click here to view 
Conclusions   
In this study, RSM was employed to optimize the extraction conditions for the active compounds from YJT using the HRE method. Using the contour and surface plots from RSM, the optimum values for the ratio of water to herbal formula, extraction time and extraction number were determined. Under these conditions, the optimal extraction conditions for the seven bioactive compounds were 1:11.99 (299.69 mL), 94.53 min and 2.21 repeats for ratio of water to herbal medicine, extraction time and extraction number, respectively, and the obtained response was 13.66 mg/g, which closely agreed with the predicted value.
Acknowledgments   
This study was supported by a grant from the Korea Institute of Oriental Medicine (no. K13030).
References   
1.  Liu DS, Gao W, Lin WW, Hao YY, Zhong LH, Li W, et al. Effects of the Chinese YiQiBuShen recipe extract on brainstem auditory evoked potential in rats with diabetes. J Ethnopharmacol 2011;137:41420. 
2.  Hsu YL, Kuo PL, Tzeng TF, Sung SC, Yen MH, Lin LT, et al. Huanglianjiedutang, a traditional Chinese medicine prescription, induces cellcycle arrest and apoptosis in human liver cancer cells in vitro and in vivo. J Gastroenterol Hepatol 2008;23:e2909. 
3.  Zhang A, Sun H, Wang X, Jiao G, Yuan Y, Sun W. Simultaneous in vivo RPHPLCDAD quantification of multiplecomponent and drugdrug interaction by pharmacokinetics, using 6,7dimethylesculetin, geniposide and rhein as examples. Biomed Chromatogr 2012;26:84450. 
4.  Li M, Xu Y, Yang W, Li J, Xu X, Zhang X, et al. In vitro synergistic antioxidant activities of solventextracted fractions from Astragalus membranaceus and Glycyrrhiza uralensis. WT Food Sci Technol 2011;44:174551. 
5.  Shi X, Li X, Liu J, Zhou H, Zhang H, Jin Y. Lignan extraction from the roots of Sinopodophyllum emodi wall by matrix solidphase dispersion. Chromatographia 2010;72:7137. 
6.  Chen CY, Li H, Yuan YN, Dai HQ, Yang B. Antioxidant activity and components of a traditional Chinese medicine formula consisting of Crataegus pinnatifida and Salvia miltiorrhiza. BMC Complement Altern Med 2013;13:99. 
7.  Gan CY, Latiff AA. Optimisation of the solvent extraction of bioactive compounds from Parkia speciosa pod using response surface methodology. Food Chem 2011;124:127783. 
8.  Renjie L. Optimization of extraction process of Glycyrrhiza glabra polysaccharides by response surface methodology. Carbohydr Polym 2008;74:85861. 
9.  Kang DG, Sohn EJ, Moon MK, Mun YJ, Woo WH, Kim MK, et al0. Yukmijihwangtang ameliorates ischemia/reperfusioninduced renal injury in rats. J Ethnopharmacol 2006;104:4753. 
10.  He H, Yang X, Zeng X, Shi M, Yang J, Wu L, et al. Protective effect of Liuwei Dihuang decoction on early diabetic nephropathy induced by streptozotocin via modulating ETROS axis and matrix metalloproteinase activity in rats. J Pharm Pharmacol 2007;59:1297305. 
11.  Liu Y, Zhao H, Zhang J, Zhang P, Li M, Qi F, et al. The regulatory effect of liuwei dihuang pills on cytokines in mice with experimental autoimmune encephalomyelitis. Am J Chin Med 2012;40:295308. 
12.  Huang Y, Zhang H, Yang S, Qiao H, Zhou W, Zhang Y. Liuwei Dihuang decoction facilitates the induction of longterm potentiation (LTP) in senescence accelerated mouse/prone 8 (SAMP8) hippocampal slices by inhibiting voltagedependent calcium channels (VDCCs) and promoting Nmethyldaspartate receptor (NMDA) receptors. J Ethnopharmacol 2012;140:38490. 
13.  Sangha JS, Sun X, Wally OS, Zhang K, Ji X, Wang Z, et al. Liuwei Dihuang (LWDH), a traditional Chinese medicinal formula, protects against βamyloid toxicity in transgenic Caenorhabditis elegans. PLoS One 2012;7:e43990. 
14.  Zhao AB, Yu B, Wu XL, Cao KJ, Li EQ, Li QM, et al. Protective effects on myelosuppression mice treated by three different classic Chinese medicine formulae. Pharmacogn Mag 2011;7:13340. 
15.  Perry B, Zhang J, Sun C, Saleh T, Wang Y. Liuwei dihuang lowers body weight and improves insulin and leptin sensitivity in obese rats. Evid Based Complement Alternat Med 2012;2012:847167. 
16.  Szeto YT, Lei PC, Ngai KL, Yiu AT, Chan CS, Kok EW, et al. An in vitro study of the antioxidant activities and effect on human DNA of the Chinese herbal decoction 'Liu Wei Di Huang'. Int J Food Sci Nutr 2009;60:6627. 
17.  Zhao X, Wang Y, Sun Y. Quantitative and qualitative determination of Liuwei Dihuang tablets by HPLCUVMSMS. J Chromatogr Sci 2007;45:54952. 
18.  Wang B, Shen L, Cong W, Lin X, Feng Y, Zhu Y, et al. A simple HPLC method for simultaneous analysis of seven bioactive constituents for Liuwei Dihuang pill and its application in quality consistency evaluation. Anal Methods 2013;5:238490. 
19.  Won JB, Ma JY, Ma CJ. Simultaneous determination of four marker components in Yukmijihwang Tang by high performance liquid chromatography/diode array detector. Arch Pharm Res 2010;33:61923. 
20.  Zhao X, Wang Y, Sun Y. Simultaneous determination of four bioactive constituents in Liuwei Dihuang Pills by micellar electrokinetic chromatography. J Pharm Biomed Anal 2007;44:11836. 
21.  Sun Y, Liu J, Kennedy JF. Application of response surface methodology for optimization of polysaccharides production parameters from the roots of Codonopsis pilosula by a central composite design. Carbohydr Polym 2010;80:94953. 
22.  Hossain MB, BarryRyan C, MartinDiana AB, Brunton NP. Optimisation of accelerated solvent extraction of antioxidant compounds from rosemary (Rosmarinus officinalis L.), marjoram (Origanum majorana L.) and oregano (Origanum vulgare L.) using response surface methodology. Food Chem 2011;126:33946. 
23.  Guo X, Zou X, Sun M. Optimization of extraction process by response surface methodology and preliminary characterization of polysaccharides from Phellinus igniarius. Carbohydr Polym 2010;80:3449. 
24.  Xie JH, Xie MY, Shen MY, Nie SP, Li C, Wang Y ×. Optimisation of microwaveassisted extraction of polysaccharides from Cyclocarya paliurus (Batal.) Iljinskaja using response surface methodology. J Sci Food Agric 2010;90:135360. 
25.  Kim SH, Kim HK, Yang ES, Lee KY, Kim SD, Kim YC, et al. Optimization of pressurized liquid extraction for spicatoside A in Liriope platyphylla. Sep Purif Technol 2010;71:16872. 
26.  Chen XP, Wang WX, Li SB, Xue JL, Fan LJ, Sheng ZJ, et al. Optimization of ultrasoundassisted extraction of Lingzhi polysaccharides using response surface methodology and its inhibitory effect on cervical cancer cells. Carbohydr Polym 2010;80:9448. 
27.  Zhao LC, He Y, Deng X, Yang GL, Li W, Liang J, et al. Response surface modeling and optimization of accelerated solvent extraction of four lignans from Fructus schisandrae. Molecules 2012;17:361829. 
28.  Muralidhar RV, Chirumamila RR, Marchant R, Nigam P. A response surface approach for the comparison of lipase production by Canida cylindracea using two different carbon sources. Biochem Eng J 2001;9:1723. 
29.  Li W, Zhao LC, Sun YS, Lei FJ, Wang Z, Gui XB, et al. Optimization of pressurized liquid extraction of three major Acetophenones from Cynanchum bungei using a boxbehnken design. Int J Mol Sci 2012;13:1453344. 
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7], [Table 8]
