Ultrasound-Assisted Extraction of Phenolic Compounds from Moroccan Lavandula stoechas L.: Optimization Using Response Surface Methodology

Biotechnology, Environmental Technology and Valorization of Bio-Resources Team, Department of Biology, Faculty of Science and Technology Al-Hoceima, Abdelmalek Essaadi University, Tetouan, Morocco Laboratory of Applied Organic Chemistry, Faculty of Sciences and Technology, Sidi Mohamed Ben Abdellah University, B.P. 2202–Route d’Imouzzer, Fez, Morocco Physico-Chemical Laboratory of Inorganic and Organic Materials, Materials Science Center (MSC), Ecole Normale Supérieure, Mohammed V University in Rabat, Rabat, Morocco Laboratory of Biotechnology Environment Agri-Food and Health, Faculty of Sciences, Sidi Mohamed Ben Abdellah University, P.O. Box 1796 (Atlas), Fez, Morocco Higher Institute of Nursing Professions and Health Techniques of Fez, Regional Health Directorate Fez-Meknes, El Ghassani Hospital, Fez, Morocco Engineering Laboratory of Organometallic, Molecular Materials and Environment (LIMOME), Faculty of Sciences, Sidi Mohamed Ben Abdellah University, P.O. Box 1796 (Atlas), Fez, Morocco


Introduction
erapeutic activities of plants have long been associated with the production of secondary metabolites, including alkaloids, flavonoids, tannins, terpenoids, coumarins, and mucilages. Polyphenols are abundant phytochemicals and constitute a large class of compounds which have attracted research attention due to their potential human health benefits [1]. ey have attracted the attention of scientists for a long time, and several recent studies have proven the antioxidant activity [2], anti-inflammatory activity [3], prevention against cardiovascular disease [4], antibacterial [5], antifungal [6], hypoglycemic [7], antiviral [8], anticancer [9] activities, and many other diseases [4].
Lavandula stoechas is one of most plants used traditionally in the Mediterranean countries, including Morocco, for the treatment of painful illnesses such as inflammatory diseases, cystitis, nephritis, and rheumatic arthritis [10,11]. Several studies have reported that the aqueous and alcoholic extracts of L. stoechas are characterized by various biological effects. Indeed, [12,13], antifungal [14], antioxidant [14,15], antihyperglycemic activities [16], anticonvulsant, antispasmodic, and sedative effects [17], and anti-inflammatory activity [17][18][19] have been reported. ese activities are related to the phenolic compounds identified in the extracts of L. stoechas. Phenolic acids and flavone glycosides are the most reported phenolic compounds in L. stoechas extracts [14,20,21].
Ultrasound-assisted extraction (UAE) is an innovative and promising technique used in several fields such as cosmetics, pharmaceutical, chemical, and food industries since the second half of the 20 th century [22]. Recently, several studies indicate that ultrasound produces more abundant extracts in biomolecules and in a shorter extraction time compared to conventional extraction techniques [22,23]. A study realized by Chemat et al. [24] showed that the yield of carvone extracted from caraway seeds (Carum carvi L.) is best in ultrasound-assisted extraction as those given by the Soxhlet extract. A comparative study on the extraction of phenolic compounds from the flowers and buds of Acacia confusa was carried out by ultrasound, maceration, and extraction assisted by heat. is study confirmed that ultrasonic extraction is the most rapid and efficient technique, which allows to significantly increase the rate of phenolic content compared to other methods studied [25]. According to several studies, the quality of the phenolic compounds and the total concentration of polyphenols in several matrices extracted by ultrasound depends on the quantity and the polarity of the solvent used, the ultrasonic intensity [26], the extraction temperature [27], and the extraction time [28]. e aim of this work was to maximize the obtained yield and total phenolic contents from L. stoechas extracted by ultrasound method. In order to achieve this objective, we used the Box-Behnken design in conjunction with a response surface methodology (RSM) to optimizing three parameters: the extraction time (min), liquid/solid ratio (ml/ g), and ethanol concentration (%).

2.2.
e Ultrasound-Assisted Extraction. e extraction method was performed with ultrasound cleaning bath Elma Transsonic TI-H-15 (a frequency of 35 kHz and a nominal power of 100 W). e powder of the dried plants (20 g) was placed in a capped tube (100 ml) and mixed with an appropriate amount of the extraction solution (according to experimental design). e tube with the suspension was immersed in water in the ultrasonic device and irradiated for the preset extraction time. After ultrasonic extraction, the sample was filtered by Whatman paper and evaporated under vacuum at 40°C on a rotary evaporator.

Experiment Design.
Response surface methodology (RSM) was used for investigating the influence of three independent variables on yield and total phenols of L. stoechas extracts. e extraction time (min, X1), liquid/solid ratio (ml/ g, X2), and ethanol concentration (%, X3) were selected as independent variables that should be optimized for the extraction. e samples were kept at room temperature to avoid the degradation of temperature-sensitive compounds. In the study, the experiments were performed on the Box-Behnken design (BBD). e level values of the experimental factors are shown in Table 1.

Determination of Total Phenolic Content.
e Folin-Ciocalteu method was used to determine the total phenolic content (TPC) of the extract with slight modifications [29]. TPC was estimated in milligrams of gallic acid equivalents per gram of dry plant material (mg of GAE/g). 100 μl of the sample (1 mg/ml) was transferred to a 10 ml volumetric flask, to which 500 μl undiluted Folin-Ciocalteu reagent was added subsequently. After 1 min, 1.5 ml of 20% (w/v) Na 2 CO 3 was added and the volume was made up to 10 ml with H 2 O. After 2 hours of incubation at 25°C, the absorbance was measured at 765 nm and compared to the gallic acid calibration curve.

Statistical Analysis.
e experimental results of the response surface design were analyzed using JMP software (version 10) [30]. e main effect of regression was considered to be statistically significant for p values less than 0.05. All experiments were conducted in triplicate unless otherwise noted in the text. e complete design was carried out in random order and consisted of 15 combinations, including three replicates in the central point. e data from the Box-Behnken design (BBD) were analyzed by multiple regressions to fit the following quadratic polynomial model: where Y is the studied response; b 0 represents the average value of responses; b 1 , b 2 , and b 3 are coefficients of the main terms; b 11 , b 22 , and b 33 are the coefficients of quadratic terms; b 12 , b 13 , and b 23 are the coefficients of the interaction terms; and ε is an error term. An analysis of variance with F test at a 95% significance level was conducted. e mean squares (MS) were obtained as follows: MS � (SS/DF), where SS is the sum of squares of each variation source and DF is the respective degree of freedom.
e ratio between the mean square regression (MS R ) and the mean square residual (MS r ), F ratio(R/r) , was used in order to establish whether the model was statistically significant [31]. e greater F-value from unity adequately explains the variation of the data around its mean. Besides, the estimated factor effects are real [32]. e quality of fitting the first-order polynomial was also expressed with the coefficient of determination R 2 .
is coefficient measures the proportion of total variation in the mean response explained by the regression. In fact, it represents the correlation between the observed and predicted responses, and it is often expressed as a percentage [33]. e model coefficients were considered significant for p values <0.05. e statistical significance of the model coefficients was determined by using the t-test (only significant coefficients with p value <0.05 are included).

Optimization of Operating Conditions.
e isoresponse curves were used to find the optimal agreement between operating conditions. ese are surface curves that reflect the changes in the response. Moreover, the "desirability" function was used to find the exact optimal combination with a percentage of compromise. Table 2 shows the yield and TPC of L. stoechas extracts obtained from all experiments.

Statistical Validation of the Postulated Model.
According to the analysis of variance table (Table 3), we can conclude that the main effect of regression is significant for both the studied responses since the significance of the risk (p value) is lower than 0.05. Obviously, for the two studied responses, the calculation of F Ratio(R/r) (13.93 for the response yield and 6.59 for the response TPC) has shown that it is higher than the theoretical value of F (0.05;9,5) at 95% confidence level which is equal to 4.77.
e linear regression of the model permitted us to evaluate the quality of the established model graphically. It is represented by the linearity of the points measured by the real model according to the points considered by the quadratic model. e goodness of the model is judged if the coefficient of determination is equal to or higher than 80%. e coefficients of determination R 2 are equal to 96% and 92% for the response yield and TPC, respectively. ese values give a good agreement between the experimental and predicted values of the adapted model. ese results are confirmed by those obtained in the graph (Figure 1), showing a linear curve for the observed values in terms of the predicted ones.

Study of the Factor's Effects and the Fitted Model for the Response Yield.
e effects of all factors studied and the statistical values of Student's t-test and the observed probability (p value) are summarized in Table 4. Student's ttest values are used to determine the significance of the coefficients of each parameter, while p values are defined as the smallest level of importance, leading to rejection of H0 Table 4 shows that all factors can be considered as significant since their p values are lower than 0.05 except factor b 1 and factor b 13 , which can be considered as negligible (p value >0.05).
e equation of the quadratic model used for the modeling extraction conditions of yield, determined according to the results of the JMP 10 software, was represented as follows:

Study of the Factor's Effects and Fitted Model for the Response TPC.
e effects of all factors studied and the statistical values of Student's t-test and the observed probability (p value) are summarized in Table 5. 22 ′ can be considered as statistically significant since their p values are lower than 0.05. e equation of the quadratic model used for the total polyphenols, determined according to the results of the JMP 10 software, was represented as follows: 3.5. Response Surface Optimization. Using the isoresponse curves, we can consider options regarding operating conditions. With manual adjustment of target values of the two studied answers, one is interested in the most advantageous economical experimental conditions. en, it is advantageous to use the least amount of ethanol in a time interval somewhat weak.

Optimization of the Yield Response.
e desirability plot ( Figure 2) shows that the maximum yield that we can have is around 31.88%. We can also see that obtaining this value is possible with a desirability of 98.7%. is same plot also shows that the maximal value of yield requires a minimal concentration of ethanol (40%), a maximum solvent-to-material ratio (30 ml/g), and a time processing higher than 30 min. By fixing the concentration of ethanol at 40% in the isoresponse plot, we can see that the required time for obtaining this yield is between 30 and 40 min when the liquid/solid ratio must be between 28 and 30 ml/ g (Figure 3(a)). e 3D plot (Figure 3(b)) shows the   graphical representation of the response yield in terms of extraction time and liquid/solid ratio and by fixing the parameter ethanol concentration (%) on its low level (40%).

Optimization of the Total Phenolic Content (TPC)
Response. e desirability plot of this second response (TPC) (Figure 4) shows that the maximum value that we can obtain is around 190.14 mg GAE/g. Obtaining this response   Journal of Chemistry is possible with 98% desirability. As in the case of the extraction yield, the maximal TPC value that we can get requires a minimal concentration of ethanol (40%), a minimal time processing, and a maximum liquid/solid ratio. e isoresponse plot ( Figure 5(a)) confirms the results of the desirability plot. us, this plot shows that the compromised area (white area) is accessible by ensuring a time between 20 and 21.5 min and a liquid/solid ratio between 28 and 30 ml/g and by fixing the concentration of ethanol at 40%. e 3D plot ( Figure 5(b)) shows the graphical representation of the response TPC in terms of extraction time and liquid/solid ratio and by fixing the parameter ethanol concentration (%) on its low level (40%).

Optimization of Two Responses Simultaneously.
e obtained results for the two responses separately indicate that the maximization of two responses requires a reduction in the ethanol concentration to 40% and an increase in the liquid/solid ratio (30 ml/g). However, this shows disagreement as regards the time factor. When we have a maximum yield (31%) performance requires a time longer than 30 minutes, the TPC response requires a minimal time (just 20 minutes) to reach its maximum. Since the interest of extraction is on TPC, we will adopt the setting that guarantees the maximization of TPC, as shown in the desirability plot ( Figure 6). By adopting the optimum setting of maximizing TPC (time: 21.5 min, liquid/solid ratio: 30 ml/ g, and ethanol concentration: 40%), we will have a TPC equal to 190 mg GEA/g with a yield of 24% (loss of 7% compared to the maximum possible yield 31%). ese results are obtained with a desirability of 99%. e isoresponse plot for both answers shows a very small area (white area) of compromise between the two studied responses (Figure 7).

Verification of Predictive Model.
To compare the predicted result with the experimental result, a control experiment randomly chosen from the experimental domain was performed. Experimental values of yield (14.85 ± 1.02%) and TPC (130.15 ± 2.35 mg GAE/g) were slightly higher than the predicted ones which were equal to 14.29% and TPC at 129.75 mg GAE/g (Table 6). is result suggests that the optimized model explains appropriately the extraction process of yield and phenolic compounds. Moreover, the experimental validation provides preliminary information that the developed model can be successfully applied for predicting the optimal extraction time, the liquid/solid ratio, and the ethanol concentration.

Discussion
e objective of this study was to determine the levels of experimental factors that would allow to obtain maximum yield and a high concentration of polyphenols. In fact, to get the maximum yield and TPC, a compromise was found in choosing the optimal adjustment for the concentration of ethanol (40%) and liquid/solid ratio (30 ml/g). However, a disagreement in extraction time was observed for the two responses. Response surface analysis demonstrates that the relationship between the total phenolic content and extraction parameters is quadratic with a good regression coefficient. As the extraction of phenolic compounds depends mainly on the polarity of solvents and compounds, a single solvent might not be effective for the extraction of a bioactive compound. Hence, a combination of alcohol with water is more effective in extracting phenolic compounds than alcohol alone [34]. Ethanol/water was chosen as the unique extraction solvent, instead of others, because of the low price of ethanol, low toxicity, easiness of recycling [35], and good polarity to extract the phenolic compounds. e obtained findings from our study are in agreement with previous studies, which reported that adding water to the alcohol shows a synergistic effect, increasing the extraction performance of phenolic compounds from plant samples [36,37]. e results of a study carried out by Guerra et al. [38] on the optimization of the parameters influencing the concentration of polyphenols obtained from mango peel by ultrasound-assisted extraction confirmed that the use of 50% ethanol made it possible to obtain optimum phenolic compounds. e extraction time is an important parameter to minimize the energy cost of the process [39]. In our study, the total phenolic concentration increased with prolonged  is observation was understandable because an extended extraction time favors the extraction of phenolic compounds [40]. Also, another study realized by Hammi et al [41] reported similar results while investigating the effect of time on the extraction efficiency of polyphenols from Zizyphus lotus, while other studies revealed a shorter extraction time to obtain the optimum phenolic compounds using ultrasound-assisted extraction. e optimal time for the extraction of phenolic compounds (1225.7 mg EAG/100 g of extract) from the peel of Litchi chinensis extracted by ultrasound was 30 minutes [42]. e maximum of phenolic compounds (345.81 ± 2.82 mg EAG/100 g of the extract) from the stems of Bactris gasipaes (Arecaceae family) was obtained during 30 min of extraction [43]. Recently, Iftikhar et al. [44] evaluated the total concentration of polyphenols in Secale cereale ethanolic extract (Poaceae), the extraction time to obtain the optimum TPC (245.74 mg GAE/100 g of extract) was 29 minutes. During these two studies, the extraction temperatures were 60 and 66°C, respectively, which could explain that the researchers obtained the optimum TPC in a reduced time compared to our study (temperature of extraction at 25°C). e increased extraction of total phenolic content was observed with an increased solvent-to-material ratio from 20 to 30 ml/g. is is probably due to the fact that more solvent can enter cells while more phenolic compounds can permeate into the solvent under the higher solvent-to-material ratio conditions [45]. A comparison of the literature has shown that extraction assisted by ultrasound contributes to the increase in yields of polyphenols, in addition to the preservation and growth of the biological activities of polyphenol extracts compared to traditional techniques [46].  ese results were obtained according to the following conditions: the temperature at 41°C, the ratio solvent-material at 1/20 (w/v), and a power value of 80%. e scientists proved that the ultrasonic extraction could reduce the extraction time compared to the extraction by maceration while obtaining a better concentration of phenolic compounds. Besides, ethanol was the most suitable solvent for the extraction of phenols [47].

Conclusion
e response surface methodology was successfully employed to optimize the yield and the phenolic compound extraction from Moroccan L. stoechas. e BBD proved to be a powerful tool for the optimization of ultrasonic-assisted extraction parameters. is study has permitted to define the optimal extraction process for yield response as follows: an extraction time of 34 min, a liquid/solid ratio of 30 ml/g, and an ethanol concentration of 40%. e optimal conditions to obtain the highest polyphenols content, 190.14 mg GAE/g, are as follows: extraction time of 21.5 min, liquid/solid ratio of 30 ml/g, and ethanol concentration of 40.5% (v/v). However, for the optimization of two responses simultaneously to get a yield of 24.2% and a total concentration polyphenols of 190.24 mg GAE/g, the following parameters must be respected: extraction time of about 20.73 min, liquid/solid ratio of 29.72 ml/g, and ethanol concentration of 40.5% (v/v). e results indicated that the UAE-RSM approach was effective for maximizing the yield extraction and TPC, and the knowledge gained from this study should be useful for further exploitation and application of the phenolic compounds.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.  Journal of Chemistry 9