



ORIGINAL ARTICLE 

Year : 2020  Volume
: 10
 Issue : 6  Page : 281292 

Response surface methodologybased optimization of ultrasoundassisted extraction of βsitosterol and lupeol from astragalus atropilosus (roots) and validation by HPTLC method
Perwez Alam^{1}, Nasir A Siddiqui^{1}, Ali S Alqahtani^{1}, Anzarul Haque^{2}, Omer A Basudan^{1}, Saleh I Alqasoumi^{1}, Abdullah A ALMishari^{1}, MU Khan^{3}
^{1} Department of Pharmacognosy, College of Pharmacy, King Saud University, P.O. Box2457, Riyadh11451, Saudi Arabia ^{2} Department of Pharmacognosy, College of Pharmacy, Prince Sattam bin Abdulaziz University, AlKharj, KSA ^{3} Department of Pharmaceutical Chemistry & Pharmacognosy, Unaizah College of Pharmacy, Qassim University, Al Qassim, Saudi Arabia
Date of Submission  02Dec2019 
Date of Decision  06Jan2020 
Date of Acceptance  06Feb2020 
Date of Web Publication  14May2020 
Correspondence Address: Perwez Alam Department of Pharmacognosy, College of Pharmacy, King Saud University, P.O. Box2457, Riyadh11451 Saudi Arabia
Source of Support: None, Conflict of Interest: None
DOI: 10.4103/22211691.283942
Objective: To optimize the ultrasonication method for efficient extraction of Psitosterol and lupeol from the roots of Astragalus atropilosus using BoxBehnken design of response surface methodology (RSM), and its validation by high performance thin layer chromatography (HPTLC) method. Methods: Ultrasonication method was used to extract βsitosterol and lupeol from Astragalus atropilosus (roots). RSM was used to optimize the different extraction parameters viz. liquid to solid ratio (1014 mL/g), temperature (6080 °C) and time (4060 min) to maximize the yield of βsitosterol and lupeol. The quantitative estimation of βsitosterol and lupeol was done in chloroform extract of Astragalus atropilosus by validated HPTLC method on 10 cm × 20 cm glassbacked silica gel 60F_{254} plate using hexane and ethyl acetate (8:2, v/v) as mobile phase. Results: A quadratic polynomial model was found to be most appropriate with regard to R^{1} (yield of total extraction; R^{2}/% CV = 0.994 8/0.28), R_{2} (βsitosterol yield; R^{2}/% CV = 0.992 3/0.39) and R_{3} (lupeol yield; R^{2}/% CV = 0.994 2/0.97). The values of adjusted R^{2}/predicted R^{2}/signal to noise ratio for R_{1}, R_{2}, and R _{3} were 0.978 2/0.955 1/48.77, 0.990 4/0.911 0/31.33, and 0.992 7/0.940 1/36.08, respectively, indicating a high degree of correlation and adequate signal. The linear correlation plot between the predicted and experimental values for R_{1}, R_{2}, and R_{3} showed high values of R^{2} ranging from 0.990 50.997 3. βsitosterol and lupeol in chloroform extract of Astragalus atropilosus were detected at R_{f} values of 0.22 and 0.34, respectively, at X max = 518 nm. The optimized ultrasonic extraction produced 8.462% w/w of R_{l}, 0.451% w/w of R_{2} and 0.172% w/w of R_{3} at 13.5 mL/g liquid to solid ratio, 78 C of temperature and 60 min of time. Conclusions: The experimental findings of RSM optimized extraction and HPTLC analysis can be further applied for the efficient extraction of βsitosterol and lupeol in other species of Astragalus.
Keywords: βsitosterol; Lupeol; BoxBehnken design; Astragalus; High performance thin layer chromatography
How to cite this article: Alam P, Siddiqui NA, Alqahtani AS, Haque A, Basudan OA, Alqasoumi SI, ALMishari AA, Khan M U. Response surface methodologybased optimization of ultrasoundassisted extraction of βsitosterol and lupeol from astragalus atropilosus (roots) and validation by HPTLC method. Asian Pac J Trop Biomed 2020;10:28192 
How to cite this URL: Alam P, Siddiqui NA, Alqahtani AS, Haque A, Basudan OA, Alqasoumi SI, ALMishari AA, Khan M U. Response surface methodologybased optimization of ultrasoundassisted extraction of βsitosterol and lupeol from astragalus atropilosus (roots) and validation by HPTLC method. Asian Pac J Trop Biomed [serial online] 2020 [cited 2020 Nov 30];10:28192. Available from: https://www.apjtb.org/text.asp?2020/10/6/281/283942 
1. Introduction   
βsitosterol, a plant steroid has been extensively studied and shows antiHIV (by immunomodulatory mechanism), antiviral (against tobacco mosaic virus), antihepatotoxic, anticardiotoxic, and anti oxidative activities^{[1]}. The triterpenoids have also been found to possess a wide spectrum of biological activities such as anti inflammatory, hypocholesterolemic, insulinregulating potential, antiviral (particularly lupeol for hepatitis), antiherpes simplex virus, antimicrobial, and antiproliferative acitivities^{[2]}. βsitosterol and lupeol have been analyzed using HPTLC in different plants such as Tephrosia purpurea (L.) Pers. (βsitosterol and lupeol ranged from 0.043% to 0.125%, 0.023% to 0.045% w/w, respectively), Hibiscus species (aerial parts; 1.18% and 0.75% w/w for βsitosterol and lupeol, respectively), Sisymbrium irio L (0.21% w/w for βsitosterol). Some compounds have been analyzed using HPLC in different species of Astragalus like astragalosides and isoflavonoids^{[3],[4],[5],[6]}, but till now, no report has been published on the quantification of these biologically important phytoconstituents in Astragalus atropilosus (A. atropilosus) using HPTLC method.
Ultrasonicassisted extraction is an economical and easily operated extraction technique in comparison to the other techniques such as supercritical fluid extraction and microwaveassisted extraction. The enhanced extraction by ultrasonic treatment is mainly attributed to its mechanical effects, which largely expedite the mass transfer between immiscible phases at low frequency by super agitation^{[7],[8]}.
Response surface methodology (RSM) is a more economical, convenient, diversified, logical and timesaving statistical technique than the conventional single parameter optimization, and has been used to simultaneously optimize different variables involved in the process. The various extraction parameters such as extraction time, extraction temperature, liquid to solid ratio, solvent ratio, etc. have been optimized for several phytoconstituents viz. betulinic acid from Tecomella undulata, triterpenoids from Jatropha curcas, embelin from Embelia ribes and phytosterol from Saccharum officinarum L^{[9],[10],[11],[12],[13]}. Among various response surface designs available in RSM, BoxBehnken design (BBD) is more labor efficient (requiring the minimum number of experimental runs), and quite suitable for fitting secondorder polynomial equations of three or more experimental factors^{[14],[15],[16],[17]}. BBD is known to be more competent than central composite and threelevel full factorial RSM designs, as it allows estimation of quadratic model parameters, sequential design building and lack of fit determination for the proposed model. In this experiment, only 17 runs were needed for a threefactorial (3^{3}) study. BBD can also help in analyzing the quadratic response surface and generating a secondorder polynomial model. The HPTLC is a widely used chromatographic technique in the quantitative analysis of herbal extracts, herbal drugs, and its supplements because it is rapid, less expensive, highly sensitive, precise, and has the potential to measure a large number of samples efficiently^{[18],[19],[20],[21],[22]}.
In the present experiment, authors planned to optimize various extraction parameters such as liquid to solid ratio, extraction temperature and extraction time for the maximum yield of βsitosterol and lupeol in chloroform extract of A. atropilosus (AACE) by applying BoxBehnken design of RSM along with the quantitative estimation of βsitosterol and lupeol in AACE for the first time by a validated, simple and efficient HPTLC method.
2. Materials and methods   
2.1. Plant material and chemicals
Roots of A. atropilosus (voucher no. 14471) were collected from the Tamniah area of Saudi Arabia. The plant material was authenticated by Dr. Mohamed Yousef, a taxonomist at Pharmacognosy Department, College of Pharmacy, King Saud University, and the voucher specimens were deposited in the herbarium, Department of Pharmacognosy. The roots were washed and dried at room temperature. After drying, roots were broken into small pieces, powdered and stored for further processing. The standards βsitosterol and lupeol [Figure 1] were procured from Sigma Aldrich. The chemicals hexane, ethyl acetate, and chloroform were purchased from BDH chemicals.
2.2. Ultrasonic extraction and determination of βsitosterol and lupeol
The extraction of the powdered root of A. atropilosus was carried out by ultrasonic vibrations (ultrasoundassisted extraction) using Sonics Vibra cell (Model VCX750; Sonics, USA). The effect of single factors on extraction procedures was determined as follows:
(1) Effect of liquid to solid ratio on the extraction:
Root powder (1.0 g) was put into a 50 mL conical flask and extracted with various volumes of chloroform (4, 6, 8, 10, 12, 16, 20 mL) to get different liquidtosolid ratios keeping the extraction time (40 min) and extraction temperature (40 °C) constant throughout the experiment. Each experiment was repeated 5 times (n=5) and the obtained extracts were merged, filtered and dried at low pressure with rotavapour to get the final extractive yield.
(2) The effect of temperature on extraction:
A total of 10 mL of chloroform was added to 1 g of powdered root in a 50 mL flask for each experiment, and the extraction was performed for different extraction temperatures (30, 40, 50, 60, 70, 80 °C) at constant extraction time (40 min). Each experiment was repeated 5 times (n=5) and the obtained extracts were merged, filtered and dried at low pressure with rotavapour to get the final extractive yield.
(3) The influence of time on the extraction:
A total of 10 mL of chloroform was added to 1 g of powdered root in a 50 mL flask for each experiment and the extraction was executed for different time variables (10, 20, 30, 40, 50, 60 min) at constant extraction temperature (40 °C). Each experiment was repeated 5 times (n=5) and the obtained extracts were merged, filtered and dried at low pressure with rotavapour to get the final extractive yield.
On the basis of the above finding of the influence of single factor on extraction yield, the ultrasoundassisted extraction procedure for RSM was set as below: root powder was taken into a conical flask (50 mL) and chloroform was added with different liquidtosolid ratios (1014 mL/g), temperature (6080 °C) and time (4060 min).
2.3. Experimental design of RSM
A 3^{3} factorial BBD (DesignExpert Software, Trial version 12, Stat Ease Inc., Minneapolis, MN, USA) of RSM (17 runs) was applied to optimize the extraction variables viz. liquid to solid ratio (P_{1} 10, 12, 14 mL/g), temperature (P_{2}: 60, 70, 80 C) and time (P_{3}: 40, 50, 60 min) to get the maximum yield (% w/w) of total extractive matter (R_{1}), βsitosterol (R_{2}) and lupeol (R_{3}). The appropriate range of different variables was determined according to singlefactor experiments. The preparation and analysis of all the samples were carried out in triplicate. A nonlinear quadratic model equation generated by this experimental design is shown below:
Where, R is the response related to each factor level combinations; k_{0} is intercept; q_{1}, q_{2}, q_{3} are linear coefficients; q_{12}, q_{13}, and q_{23} are the interaction coefficients while q_{11}, q_{22}, and q_{33} are the quadratic coefficients. The independent variables were P_{1}, P_{2}, and P_{3} while R_{1}, R_{2} and R_{3} were the dependent variables. The results of various initial trials were used to choose the range of independent variables. Here all the variables including solvent to solid ratio, temperature and sonication time were studied at three levels, low (1), medium (0) and high (+1). The obtained extracts were filtered using filter papers and used to determine the content of βsitosterol and lupeol by using validated HPTLC method.
2.4. HPTLC analyses of βsitosterol and lupeol
All the 17 BBD runs of the AACE (2 mg/mL) were applied as spots (10 μL) on a 10 cm×20 cm glassbacked silica gel 60F_{254} HPTLC plate (Merck, Germany) with a band size of 6 mm using Automatic sampler4 (CAMAG, Switzerland). Before the application, the extract solutions were filtered using a 0.22 μm filter fitted with a microliter syringe (CAMAG). Then the postapplication of the plate was developed in a twin trough glass chamber (Automatic Development Chamber2, CAMAG) saturated with the mobile phase [mixture of hexane and ethyl acetate in the ratio of 8:2 (v/v)] for 20 min at controlled temperature [(25 ± 2) °C] and controlled humidity [(60 ± 5)%] and the chromatogram was developed up to a height of 8.0 cm. The postdevelopment of the TLC plate was airdried (30 min), derivatized with jpanisaldehyde reagent and dried again at 110 °C for 10 min (in a hot air oven) to furnish the clear and compact spots of all the phytoconstituents present in the sample along with the markers (βsitosterol and lupeol). The plate was scanned by using TLC scanner3 (CAMAG) in absorbance mode at X max = 518 nm and the concentrations of βsitosterol and lupeol in all the seventeen runs were quantified by using regression equation obtained from the calibration curve of a βsitosterol and lupeol standards.
2.4.1. Calibration curve preparation
A stock solution (1 mg/mL) of standards βsitosterol and lupeol in chloroform was prepared. The stock solution was further diluted with chloroform in order to get seven different dilutions viz. 10, 20, 40, 60, 80, 100 and 120 μg/mL. All the seven dilutions of βsitosterol and lupeol (10 μL, each) were applied in triplicate on the HPTLC plate to furnish concentrations of 100, 200, 400, 600, 800, 1 000 and 1 200 ng/band. Furthermore, the linear leastsquares regression was used to treat the data of peak area versus the concentration of biomarkers.
2.4.2. Validation
The developed HPTLC method was validated for accuracy, precision, robustness, the limit of detection (LOD) and limit of quantification (LOQ) as per the International Conference on Harmonization guidelines^{[23]}. The recovery as accuracy studies of βsitosterol and lupeol was accomplished by the standard addition method. Additionally, the analyte was spiked with various concentrations (50%, 100%, and 150%) of βsitosterol and lupeol and reanalyzed by using the proposed HPTLC method in triplicate. The % relative standard deviation (RSD) and recovery were calculated, and the intra and interday precisions of three replicates for βsitosterol and lupeol determination were executed at three concentrations (400, 600 and 800 ng/band). The % RSD of peak areas were calculated. However, for the robustness study of the developed HPTLC method, a small intentional modification was applied to the composition of the solvent system, the mobile phase volume (18, 20, 22 mL) and the saturation time (10, 20, 30 min). Moreover, the effects on the result were calculated as SD and % RSD. The LOD and LOQ for the βsitosterol and lupeol were calculated by using equations (1) and (2), respectively:
LOD = (3.3xSD)/α (1)
LOQ = (10×D)/α (2)
Where SD is the least standard deviation and a is the slope of the curve.
The specificity of the developed HPTLC method was confirmed by analyzing βsitosterol and lupeol standards and its presence in AACE. Furthermore, the spots for βsitosterol and lupeol in AACE were established by comparing the R_{f} value, color and the peak of the spot in the samples with those of the standard.
2.5. RSM model and validity testing
To analyze the experimental results, BBD of RSM (Design Expert™ software, version 12) was used, and Pvalue < 0.05 were considered to be significant. Additionally, independent variables of the extraction process such as P_{1}, P_{2}, and P_{3} were concurrently optimized by using BBD. The ultrasonication extraction of the crude drug was executed by using the optimized conditions in triplicate and the yield of R_{1}, R_{2} and R_{3} was compared with predicted values for the model validation.
3. Results   
3.1. HPTLC analysis of βsitosterol and lupeol
Out of these solvents, hexane and ethyl acetate in the ratio of 8:2, v/v was found to be quite selective. The developed HPTLC method provided a sharp, compact and welldefined peaks of βsitosterol and lupeol at the R_{f} values of 0.22 and 0.34, respectively [Figure 2]A. [Figure 2]B shows that the selected solvent system had a very good resolution for the separation of βsitosterol and lupeol from other constituents of AACE. The identities of the bands were confirmed by overlaying the spectra of all the extracts with the spectra of βsitosterol and lupeol [Figure 2]C. Furthermore, the linear regression data obtained for the calibration curves (n=6) showed a good linear relationship over a wide range of concentrations (1001 200 ng/band) with respect to peak area [Supplementary Table 1] [Additional file 1]. The linear equation/correlation coefficients (r^{2}) for βsitosterol and lupeol were found as Y= 10.363X + 522.03/0.997 2 and Y= 11.442X + 790.77/0.994 1, respectively. The LOD/LOQ (ng) for βsitosterol and lupeol were found as 10.32/31.28 and 21.17/64.16, respectively.  Figure 2: High performance thin layer chromatography (HPTLC) chromatogram and spectral comparison. A: HPTLC chromatogram of standards βsitosterol and lupeol; B: HPTLC chromatogram of βsitosterol and lupeol estimation in chloroform extract of Astragalus atropilosus (βsitosterol, spot 5, Rf = 0.22; lupeol, spot 7, Rf =0.34) using mobile phase: hexane: ethyl acetate (8:2, v/v) at λ max= 518 nm; C: Spectral comparison of all tracks. AACE: chloroform extract of Astragalus atropilosus.
Click here to view 
On one hand, the % RSD for intraday/interday precision of βsitosterol and lupeol were found as 1.2091.337/1.0871.287 and 1.2311.522/1.2191.447, respectively [Supplementary Table 2] [Additional file 2]. The accuracy was calculated by recovery analysis which afforded recovery of 98.26%99.59% and 98.15%99.43%, respectivelyfor βsitosterol and lupeol [Supplementary Table 3] [Additional file 3]. On the other hand, the % RSD of βsitosterol and lupeol were found as 0.9781.423 and 0.9991.315, respectively for the accuracy of the proposed method. The low values of % RSD of βsitosterol and lupeol in robustness studies are recorded in [Supplementary Table 4] [Additional file 4].
3.2. Effect of singlefactor tests with ultrasonic extraction of AACE
3.2.1. Effect of extraction time (P_{3}) on the yield of AACE
The [Supplementary Figure 1]A [Additional file 6] shows that the extraction yield (% w/w) of AACE was affected by variation in P_{3} (10, 20, 30, 40, 50 and 60 min) [where the other two factors P_{1} (liquid to solid ratio) and P_{2} (extraction temperature) were fixed at 10 mL/g and 40 °C, respectively]. In addition, the (%) extraction yields of the AACE increased significantly from 4.10% to 5.71% when P_{3} increased from 10 to 40 min. However, no change was observed with further increase in P_{3} from 4060 min, which indicated that 40 min was the time limit to get the maximum AACE yield.
3.2.2. Effect of extraction temperature (P_{2}) on the yield of AACE
The selected extraction temperatures (P_{2}) were 30, 40, 50, 60, 70 and 80 °C, respectively, to study the impact of extraction temperature on the AACE extraction yield (%), keeping the other two factors P_{1} (10 mL/g) and P_{3} (40 min) constant [Supplementary Figure 1]B. The result demonstrated that the AACE extraction yield was increased with an increase in P_{2}, reaching the maximum at 60 °C . However, no significant difference was observed by further increasing P_{2} from 60 °C80 °C.
3.2.3. Effect of liquid to solid ratio (P_{1}) on the yield of AACE
The impact of P_{1} on the extraction yield of AACE is shown in [Supplementary Figure 1]C. The AACE extraction yield was significantly increased from 35.1 to 53.3 mg/g as the P_{1} increased within the range of 4 to 12 mL/g, due to the increase of the driving force for the mass transfer. However, as the P_{1} continued to increase, the extraction yields did not differ significantly any longer.
3.3. Model fitting
The (% w/w) quantity of βsitosterol (R_{2}) and lupeol (R_{3}) of each experimental BBD run was estimated by validated HPTLC method, and the results are shown in [Table 1] along with the total extraction yield (R_{1}). A quadratic model was found to be the best fit model and the comparative results of regression analysis for model and response regression equation for the final proposed model are listed in [Supplementary Table 5] [Additional file 5]. The values of adjusted R^{2}/predicted R^{2} for R_{1}, R_{2}, and R_{3} were found as 0.978 2/0.955 1, 0.990 4/0.911 0 and 0.992 7/0.940 1, respectively which were close to 1. This indicated a high degree of correlation between the observed and predicted values. Similarly, the difference between the adjusted R^{2} and predicted R^{2} is less than 2, which is required to fit the model. “Adequate Precision” measures the signal to noise ratio which should be greater than 4 to fit the model. In this experiment, the signal to noise ratios were found as 48.77, 31.33 and 36.08 for R_{1}, R_{2}, and R_{3}, respectively. [Table 2] showed the analysis of variance (ANOVA) for the fitted quadratic polynomial of R_{1}, R_{2}, and R_{3} from A. atropilosus. The “lack of fit Fvalue” was found as 1.38, 3.01, and 2.05 for R_{1}, R_{2}, and R_{3} which showed that the “lack of fit” was not significant and showed the validity of RSM results. Furthermore, in this experiment, the model Fvalue for R_{1}, R_{2}, and R_{3} was found as 149.82/100.33/133.09 which suggests that the model was significant.  Table 1: Response surface central composite design (uncoded) and results for R_{1}, R_{2}, and R_{3}.
Click here to view 
 Table 2: ANOVA for the fitted quadratic polynomial model of R_{1}, R_{2} and R_{3}
Click here to view 
3.4. Effect of extraction parameters (P_{1}, P_{2}, P_{3}) on R_{1}, R_{2} and R_{3}, and RSM analysis
The contributions of each independent variable are shown in [Table 3]. The linear variables (P_{1}, P_{2}, P_{3}), the interaction variables (P_{1}P_{2}) and the quadratic variables (P_{1}^{2}, P_{3}^{2}) were found significant (P<0.05), and affected the R_{1} whereas other variables (P_{2}P_{3}, P_{1}P_{3} and P2^{2}) were found insignificant (P>0.05). All the linear and the quadratic variables along with the interaction variables (P_{1}P_{2}, P_{2}P_{3}) were found significant (P<0.05) and affected the R _{2} except the interaction variable P_{1}P_{3} (P>0.05). In the case of R_{3}, all the variables (the linear, quadratic and interaction variables) were found significant (P<0.05) and affected it. Furthermore, the R^{2}/coefficient of variation (% CV) of the model for R_{1}, R_{2} and R_{3} were found as 0.994 8/0.28, 0.992 3/0.39 and 0.994 2/0.97, which indicated a good precision and reliability of the experimental values. Moreover, three dimensional (3D) plots were constructed to visualize the relationship between independent variables and R_{1}, R_{2} and R_{3} according to the generated quadratic polynomial model equation of the coded factors:  Table 3: Significance of each response variable effect showed by using the F ratio and Pvalue in the nonlinear secondorder model
Click here to view 
A positive value of the variables’ coefficients indicated that it is in favor of optimization. However, a negative value indicated a reverse relationship between the independent variables and the response (R_{1}, R_{2} and R_{3}). Therefore, it is evident from the equation 1, 2 and 3 that the variables such as P_{1}, P_{2} and P_{3} had a positive effect on R_{1}, R_{2} and R_{3}, respectively. It also revealed that the relationship between the response and the variables was not constantly linear. When more than one variable is changed simultaneously, the variables can show various degrees of response.
The combination ratio of all the variables (P_{1}, P_{2} and P_{3}) for the extraction was selected based on the results of R_{1}, R_{2} and R_{3} using threedimensional response surface plots. As shown in [Figure 3]A & [Figure 3]C, [Figure 4]A & [Figure 4]C and [Figure 5]A, [Figure 5]C the R_{1}, R_{2} and R_{3} were increased positively with the increase in P_{2} up to 78 C when P_{1} and P_{3} were fixed at 12 mL/g and 50 min, respectively. [Figure 3]B, [Figure 4]B and [Figure 5]B showed an increase in R_{1}, R_{2} and R_{3} at longer P_{3} and lower P_{1} when P_{2} was fixed at 60 °C.  Figure 3: Response surface model 3D plots showing the effects of P_{1}, P_{2} and P_{3} on R_{1}. (A) the effect of P_{1} and P_{2} on R_{1}; (B) the effect of P_{1} and P_{3} on R_{1}; (C) the effect of P_{2} and P3 on R_{1}
Click here to view 
 Figure 4: Response surface model 3D plots showing the effects of P_{1}, P_{2} and P_{3} on R_{2}. (A) the effect of P_{1} and P_{2} on R2; (B) the effect of P_{1} and P_{3} on R_{2}; (C) the effect of P_{2} and P_{3} on R_{2}.
Click here to view 
 Figure 5: Response surface model 3D plots showing the effects of P_{1}, P_{2} and P_{3} on R_{3}. (A) the effect of P_{1} and P_{2} on R_{3}; (B) the effect of P_{1} and P_{3} on R_{3}; (C) the effect of P_{2} and P_{3} on R_{3}.
Click here to view 
3.5. RSM validation
For the R_{1}, R_{2} and R_{3} checkpoints, the yield evaluation result was found to be within the limits. For the validation of the RSM results of R_{1}, R_{2} and R_{3}, the experimental values of the responses were compared with the anticipated values and the percentage prediction errors were found to be 1.40%, 1.60%, and 1.05%, respectively. This helps in establishing the validity of the generated equation and describing the domain of applicability of the RSM model. The linear correlation plot between predicted and experimental values for R_{1}, R_{2} and R_{3} showed a high value of R^{2} (ranging from 0.990 50.997 3), indicating excellent goodness of fit (P<0.001) [Supplementary Figure 2] [Additional file 7].
3.6. Optimization and verification of the model for the extraction parameters
The optimum extraction process parameters were determined by maximizing the responses R_{1}, R_{2} and R_{3}. During the optimization stage, the desirability function of the DesignExpert™ (version 12) statistical software was applied to obtain the best compromise of response. The predicted optimal condition for the extraction process was found at 13.0 mL/g (P_{1}), 76 °C (P_{2}), and 56.5 min (P_{3}) which resulted in the extraction of 8.496%, 0.445% and 0.169% w/w of R_{1}, R_{2} and R_{3}, respectively. The extraction process once more repeated by modifying the optimum extraction conditions viz. 13.5 mL/g (P_{1}), 78 °C (P_{2}) and 60 min (P_{3}) and the total extraction yield (R_{1}), βsitosterol yield (R_{2}) and lupeol yield (R_{3}) were found as (8.462±0.440)% w/w, (0.451±0.020)% w/w and (0.172± 0.010)% w/w, respectively. There was no significant difference (P>0.05) between the predicted and obtained values. Therefore, this model may be applied for the optimization of the extraction process of βsitosterol and lupeol from the roots of A. atropilosus.
4. Discussion   
The simultaneous quantification of βsitosterol and lupeol in all the fractions of AACE collected during BBD runs was carried out by validated HPTLC method using hexane and ethyl acetate as suitable mobile phase, which showed a good resolution and a separation of βsitosterol and lupeol along with the other phytoconstituents available in all the fractions of AACE. The developed method was validated as per the guideline of WHO. The low values of % RSD of βsitosterol and lupeol indicated an excellent precision for intraday/interday study and the highest accuracy of the proposed method. Furthermore, the low values of % RSD for βsitosterol and lupeol obtained by deliberately changing the mobile phase composition and the time and temperature of the saturation clearly indicate that the mobile phase is robust.
The effect of extraction time as a single factor for the ultrasonic extraction of AACE was tested. The result showed that the (%) extraction yields of the AACE increased significantly with the increase in time from 10 to 40 min. However, no significant increase was seen after that. Consequently, the time affects the liquid circulation and turbulence produced by the cavitation, which causes an increase in the extraction efficiency by increasing the contact surface area between the solvent and the targeted compounds^{[24]}. The increase in the extraction time may lead to the degradation of the triterpenoidal compounds. Similarly, in the case of the extraction, the temperature affects the yield increased with temperature till 60 °C . However, no significant increase was observed after that. The increase in the extraction yield with increasing temperature is because of the higher mass transfer rate, which leads to higher molecular diffusion^{[25]}. The effect of liquid to solid ratio on the extraction yield of AACE was also studied. It was observed that yield increased with the increase in liquid to solid ratio, which may be due to the increase of the driving force for the mass transfer. Therefore, it is consistent with the fact that higher liquid to solid ratios increases the contact surface between the plant material and the solvent, which enhances the mass transfer of soluble compounds from material to solvent^{[26],[27]}. Based on these observations, the ranges of the three independent variables for the optimization of the ultrasonic extraction method by BBD of RSM were selected as liquid to solid ratio:1014 mL/g, the extraction temperature: 6080 °C and the extraction time: 4060 min.
The seventeen runs of BBD were carried out and analyzed with the help of validated HPTLC method to find out the quantity of βsitosterol and lupeol. Consequently, a quadratic model was found to be the best fit model for the BBD analysis. The values of the adjusted R^{2}/predicted R^{2} for R_{1}, R_{2} and R _{3} were found to be close to j, which indicated a high degree of correlation between the observed and predicted values. Furthermore, the difference between the adjusted R^{2} and the predicted R^{2} is less than 2, which is required to fit the model. The “Adequate Precision” measuring the signal to noise ratio was found more than 4, which indicated an adequate signal and can be used to navigate the design space. In addition, the low “lack of fit Fvalue” was found for R_{1}, R_{2} and R_{3}, which indicated that the “lack of fit” is not significant and showed the validity of RSM results. The “lack of fit Fvalue” test for the model explains the deviation in the data around the fitted model. If it is significant, it means that the model does not fit the data well, hence the insignificant lack of fit is good to fit the model. In this experiment, the model Fvalue for R_{1}, R_{2} and R_{3} was found to be high, which suggests that the model was significant.
The significance of each extraction variables (P_{1}, P_{2}, P_{3}) effects on R_{j:} R_{2} and R_{3} and the RSM analysis was evaluated. The interactions of P_{1} and P_{3} and the square root of P_{1} produced negative effects on R_{1} which indicated that if P_{1} were to be doubled then R_{1} will robustly decrease. Moreover, the interactions of P_{1} and P_{2}, P_{1} and P_{3}, and P_{2} and P_{3} along with the square roots of P_{1}, P_{2} and P_{3} produced positive effects on the R_{2}, which suggested that the increase in any variable will increase the R_{2}. The interactions of P_{1} and P_{2}, P_{1} and P_{3}, and P_{2} and P_{3} produced negative effects on R_{3} while the square root of P_{1}, P_{2} and P_{3} produced the positive effects, which indicated that if the square root of variables P_{1}, P_{2} and P_{3} were to be doubled then R_{3} will greatly increase.
The 3D plots were constructed to visualize the relationship between the independent variables (P_{1}, P_{2}, P_{3}) and R_{1}, R_{2} and R_{3}. It was clear from the 3D plot that P_{2} (extraction temperature) had a more significant effect on the R_{1}, R_{2} and R_{3}. The maximum yield of R_{1}, R_{2} and R_{3} were obtained at an optimum temperature of 78 C . This proves that a higher temperature is helpful in enhancing the compound yield as it increases the diffusion coefficient and solubility, although it may also cause compound degradation[28]. For a high yield of R_{1}, R_{2} and R_{3}, the optimum extraction temperature, the extraction time and the liquid to solid ratio were found as 78 °C, 60 min and 13.5 mL/g, respectively.
To validate the RSM results, the experimental values of the responses were compared with the anticipated values. In addition, the percentage prediction errors were evaluated which established the validity of the generated equation and the applicability of the RSM model. The low magnitudes of error, as well as the significant values of R^{2} in the present experiment, prove the high prognostic ability of the RSM.
In summary, the experimental findings indicated that BBD for RSM and a validated HPTLC method may be highly efficient and promising techniques for optimizing the extraction conditions and the quantitative analysis of βsitosterol and lupeol from A. atropilosus roots. All the selected variables, their interactions, and quadratic terms had a significant impact on the yield of the total extraction (R_{1}), βsitosterol (R_{2}) and lupeol (R_{3}). The model prediction can be used to optimize the yield of R_{1}, R_{2} and R_{3} from A. atropilosus (roots) within the limits of the experimental variables. The modified optimal extraction conditions for R_{1}, R_{2} and R_{3} in the A. atropilosus root were found as P_{1} (liquid to solid ratio) of 13.5 mL/g, P_{2} (extraction temperature) of 78 °C and P_{3} (extraction time) of 60 min. Under these optimal extraction conditions, the experimental yield of R_{1}, R_{2} and R_{3} was found as (8.462±0.440)% w/w, (0.451±0.020)% w/w and (0.172±0.010)% w/ w, respectively, which agreed closely with the predicted yield value. The quadratic polynomial model was most appropriate with regard to R_{1} (R^{2}/% CV= 0.994 8/0.28), R_{2} (R^{2}/% CV= 0.992 3/0.39) and R_{3} (R^{2}/% CV= 0.994 2/0.97). The values of adjusted R^{2}/predicted R^{2} for R_{1}, R_{2} and R_{3} were found as 0.978 2/0.955 1, 0.9904/0.911 0 and 0.9927/0.940 1, respectively (close to 1) and its difference was less than 2. This indicated a high degree of correlation and good model fitting. The signal to noise ratio were 48.77, 31.33 and 36.08 for R_{1}, R_{2} and R_{3}, respectively, which indicated an adequate signal and can be used to navigate the design space. Furthermore, the linear correlation plot between the predicted and experimental values for R_{1}, R_{2} and R_{3} showed a high value of R^{2} (ranging from 0.99050.997 3), indicating the prognostic ability of the RSM design. In this study, the solvent system developed for the HPTLC analysis of βsitosterol and lupeol was found to be excellent in resolving their peaks efficiently and the low values of LOD and LOQ showed the great sensitivity of the developed method.
In the future, the extraction of βsitosterol and lupeol from the A. atropilosus (roots) using the ultrasonic extraction can be used as an alternative natural source of βsitosterol and lupeol for the pharmaceutical industries. The findings of the RSM analysis can be applied in the future for the maximum extraction of the βsitosterol and lupeol in other species of genus Astragalus. The obtained statistical data supports the applicability of the developed HPTLC method for the quality control of herbal preparations containing βsitosterol and lupeol.
Conflict of interest statement
We declare that there is no conflict of interest.
Acknowledgments
The authors are grateful to the Researchers Supporting Project Number (RSP2019/132), King Saud University, Riyadh, Kingdom of Saudi Arabia.
Authors’ contributions
PA: Planning, execution, manuscript writing and correspondence; NAS: HPTLC analysis of different exttracts; ASA: BBD analysis; AH: literature survey; OAB: ultrasonic extraction of crude drug; SIA: manuscript writing; AAM: collection, drying and storage of crude drug; MUK: data analysis.
References   
1.  Parvez MK, Alam P, Arbab AH, AlDosari MS, Alhowiriny TA, Alqasoumi SI. Analysis of antioxidative and antiviral biomarkers Pamyrin, Psitosterol, lupeol, ursolic acid in Guiera senegalensis leaves extract by validated HPTLC methods. Saudi Pharm J 2018; 26 (5): 685693. 
2.  Arbab AH, Parvez MK, AlDosari MS, AlRehaily AJ, Ibrahim KE, Alam P, et al. Therapeutic efficacy of ethanolic extract of Aerva Javanica aerial parts in the amelioration of CCl _{4}induced hepatotoxicty and oxidative damage in rats. Food Nutr Res 2016; 60: 3086430873. 
3.  Katakam S, Sharma P, Anandjiwala S, Sharma S, Shrivastava N. Investigation on apposite chemical marker for quality control of Tephrosia purpurea (L.) Pers. By means of HPTLCchemometric analysis. J Chromatogr B Analyt Technol Biomed Life Sci 2019; 11101111: 8186. 
4.  Alam P, AlYousef HM, Siddiqui NA, Alhowiriny TA, Alqasoumi SI, Amina M, et al. Anticancer activity and concurrent analysis of ursolic acid, Psitosterol and lupeol in three different Hibiscus species (aerial parts) by validated HPTLC method. Saudi Pharm J 2018; 26 (7): 10601067. 
5.  AlMassarani SM, El Gamal AA, Alam P, AlSheddi ES, AlOqail MM, Farshori NN. Isolation, biological evaluation and validated HPTLC quantification of the marker constituent of the edible Saudi plant Sisymbrium irio L. Saudi Pharm J 2017; 25 (5): 750759. 
6.  Zhang X, Xiao HB, Xue XY, Sun YG, Liang XM. Simultaneous characterization of isoflavonoids and astragalosides in two Astragalus species by highperformance liquid chromatography coupled with atmospheric pressure chemical ionization tandem mass spectrometry. J Sep Sci 2001; 730 (13): 20592069. 
7.  Chen L, Jin H, Ding L, Zhang H, Li J, Qu C. Dynamic microwave assisted extraction of flavonoids from Herba epimedii. Sep Purif Techno 2008; 59: 5057. 
8.  Vinatoru M, Toma M, Radu O, Filip PI, Lazurca D, Mason TJ. The use of ultrasound for the extraction of bioactive principles from plant materials. Ultrason Sonochem 1997; 4: 135139. 
9.  Yin GH, Dang YL. Optimization of extraction technology of the Lycium barbarum polysaccharides by BoxBehnken statistical design. Carbohydr Polym 2008; 74: 603610. 
10.  Siddiqui N, Aeri V. Optimization of betulinic acid extraction from Tecomella undulata bark using a BoxBehnken design and its densitometric validation. Molecules 2016; 21 (4): 393. 
11.  Wei L, Wei Z, Li Y, Fang Y, Ying X, Fang C. Extraction optimization of total triterpenoids from Jatropha curcas leaves using response surface methodology and evaluations of their antimicrobial and antioxidant capacities. Electron J Biotechnol 2015; 18 (2): 8895. 
12.  Alam MS, Damanhouri ZA, Ahmad A. Development of response surface methodology for optimization of extraction parameters and quantitative estimation of embelin from Embelia ribes Burm by high performance liquid chromatography. Pharmacogn Mag 2015; 11 (1S): S166S172. 
13.  Simin F, Zisheng L, Zhou Z, Lei J, Kaichen T. Extraction optimization by responsesurface methodology: Purification and characterization of phytosterol from sugarcane (Saccharum officinarum L.) rind. J Sep Sci 2014; 37: 13081314. 
14.  Abdulhameed AS, Jawad AH, Mohammad AT. Synthesis of chitosan ethylene glycol diglycidyl ether/TiO _{2} nanoparticles for adsorption of reactive orange 16 dye using a response surface methodology approach. Desalin Water Treat 2017; 62: 208220. 
15.  Mohammad AT, Abdulhameed AS, Jawad AH. BoxBehnken design to optimize the synthesis of new crosslinked chitosanglyoxal/TiO( _{2}) nanocomposite: Methyl orange adsorption and mechanism studies. Int J Biol Macromol 2019; 129: 98109. 
16.  Abdulhameed AS, Mohammad AT, Jawad AH. Application of response surface methodology for enhanced synthesis of chitosan tripolyphosphate/TiO _{2} nanocomposite and adsorption of reactive orange 16 dye. J Clean Prod 2019; 232: 4356. 
17.  Abdulhameed AS, Jawad AH, Mohammad AT. Synthesis of chitosan ethylene glycol diglycidyl ether/TiO( _{2}) nanoparticles for adsorption of reactive orange 16 dye using a response surface methodology approach. Bioresour Technol 2019; 293: 122071. 
18.  Alam P, AlRehaily AJ, Siddiqui NA, AlSheddi ES, Shakeel F. A stabilityindicating assay of biomarker bergenin in the aerial parts of Flueggea virosa by a validated highperformance thinlayer chromatographicdensitometric method. J Planar Chromatogr 2015; 28 (1): 5460. 
19.  Alam P, Basudan OA, Siddiqui NA, Alqasoumi SI, AbdelKader MS, Donia Abd El Raheim M, et al. Development of a densitometric high performance thinlayer chromatographic method for the quantitative analysis of biomarker lupeol in the leaves of different species of genus Ficus. J Planar Chromatogr 2015; 28 (1): 3035. 
20.  Alam P, Siddiqui NA, Basudan OA, AlRehaily AJ, Alqasoumi SI, Alam P, et al. Comparative profiling of biomarker psoralen in antioxidant active extracts of different species of genus Ficus by validated HPTLC method. Afr J Tradit Complement Altern Med 2015; 12 (1): 5767. 
21.  Siddiqui NA, Alam P, AlRehaily AJ, AlOqail MM. Simultaneous quantification of biomarkers bergenin and menisdaurin in the methanol extract of aerial parts of Flueggea virosa by validated HPTLC densitometric method. J Chromatogr Sci 2015; 53: 824829. 
22.  Alajmi MF, Alam P, Siddiqui NA, Basudan OA, Hussain A. Quantitative analysis of biomarker rutin in different species of genus Ficus by validated NP and RPHPTLC methods. Pak J Pharm Sci 2015; 28 (6S): 22132220. 
23.  ICH guideline Q2B on validation of analytical procedures: Text and methodology. Geneva: International Conference on Harmonization; 2006. 
24.  Zhang H, Cao X, Liu Y, Shang F. Rapid recovery of high content phytosterols from corn silk. Chem Central J 2017; 11 (1): 108. 
25.  Wang T, Guo N, Wang SX, Kou P, Zhao CJ, Fu YJ. Ultrasoundnegative pressure cavitation extraction of phenolic compounds from blueberry leaves and evaluation of its DPPH radical scavenging activity. Food Bioprod Process 2018; 108: 6980. 
26.  Yang L, Cao YL, Jiang JG, Lin QS, Chen J, Zhu L. Response surface optimization of ultrasoundassisted flavonoids extraction from the flower of Citrus aurantium L. var. amara Engl. J Sep Sci 2010; 33: 13491355. 
27.  Xu DP, Zheng J, Zhou Y, Li Y, Li S, Li HB. Ultrasoundassisted extraction of natural antioxidants from the flower of Limonium sinuatum: Optimization and comparison with conventional methods. Food Chem 2017; 217: 552559. 
28.  Palma M, Taylor LT. Extraction of polyphenolic compounds from grape seeds with near criticalcarbon dioxide. J Chromatogr A j999; 849: 117124. 
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5]
[Table 1], [Table 2], [Table 3]
