*2.8. Experimental Design*

The CCD with three factors was deployed to determine the optimal extraction conditions for phenolic compounds. The ethanol concentration (%, *X*1), extraction temperature (◦C, *X*2) and extraction time (min, *X*3) were investigated at the five levels (−1.68, −1, 0, +1, +1.68). The levels of coded and actual factors are shown in Table 2. The UAE is commonly performed at the lower temperatures (20–70 ◦C) compared to conventional extraction procedures [32]. This method is desirable for the extraction of thermosensitive phenolic compounds from various plant species. For this reason, the temperature was observed in the range of 33–67 ◦C. The experimental design included the data set that belongs to the points of factorial design (23), axial points (2 × 3), and central point (4). The central point was repeated four times to determine the statistical parameters of the proposed model. Due to statistical calculations, the factors *X*i were coded as *x*i according to Equation (2):

$$\alpha\_{i} = \frac{\chi\_{i} - \chi\_{0}}{\delta \mathcal{K}} \tag{2}$$

where *X*0 is the value of *X*i at the central point and δ*X* is the step change. The Design Expert 12.0.0 (Stat Ease, Minneapolis, MN, USA) software was used to obtain the analysis of variance (ANOVA), regression coefficients, and regression equation. The data of response were fitted using a second-order polynomial equation (Equation (3)):

$$\mathbf{Y} = \beta\_0 + \beta\_1 \mathbf{x}\_1 + \beta\_2 \mathbf{x}\_2 + \beta\_3 \mathbf{x}\_3 + \beta\_{11} \mathbf{x}\_1^2 + \beta\_{22} \mathbf{x}\_2^2 + \beta\_{33} \mathbf{x}\_3^2 + \beta\_{12} \mathbf{x}\_1 \mathbf{x}\_2 + \beta\_{13} \mathbf{x}\_1 \mathbf{x}\_3 + \beta\_{23} \mathbf{x}\_2 \mathbf{x}\_3 + \varepsilon \tag{3}$$

where *Y* is the predicted response; β0 is the intercept; β1, β2, and β3 are the linear coefficients of *x*1, *x*2, *x*3, respectively; β11, β22, and β33 are the squared coefficients of *x*1, *x*2, and *x*3, respectively; β12, β13, β23 are the coefficients of interaction between *x*1 and *x*2, *x*1 and *x*3, *x*2 and *x*3, respectively; ε is the residual.


**Table 2.** Experimental design space for the extraction of phenolic compounds from black locust flowers.

#### 2.8.1. Statistical Analysis of the Regression Model

ANOVA with 95% confidence level was carried out to analyze the significance of the model and equation terms. The sum of squares (SS), degree of freedoms (df), mean squares (MS), *F*- and *p*-values were used as the statistical parameters. The statistical significance of the terms was analyzed based on *p*-value (Prob > *F*). The model terms are statistically significant, if the *p*-value is less than 0.0500. The coefficient of determination (*R*2), adjusted correlation coefficient (Adj-*R*2), and predicted correlation coefficient (Pred-*R*2) were used to express the quality of the regression model. The model's significance was checked using an *F*-test.

#### 2.8.2. Optimization of Phenolic Compounds' Extraction

The extraction was optimized using a numerical optimization method in order to maximize the yield of phenolic compounds. Before optimization, the weighted factor was assigned to 1. The weight is important to define the form of response desirability function. It is desirable to have the value in the range of 1–10. The higher value of the weight indicates the greater importance of the response. The importance of goal was adjusted at the default value of 3. This parameter can have a value between 1 (least important) and 5 (most important).

#### **3. Results and Discussion**

#### *3.1. Modeling of Phenolic Compounds' Extraction Using CCD*

Modeling of the extraction of phenolic compounds from black locust flowers was carried out according to the matrix of CCD with 18 experimental runs. The combinations of different factor levels and TPC are given in Table 3. The coded factors are also presented in parenthesis. TPC in the extracts was in the range 2.33–3.15 gGAE 100 g<sup>−</sup><sup>1</sup> d.p.m.


**Table 3.** Matrix of central composite design for three factors with total phenolic content.

The ANOVA results at 95% confidence level are depicted in Table 4. The significance of terms in the second order polynomial equation was estimated using ANOVA. The model's *F*-value of 34.64 was higher than the critical value of 3.39 so that the model can be considered as statistically significant. Since the lack-of-fit *F*-value of 4.3 was lower than the critical value of 9.01, the lack-of-fit can be considered as not statistically significant relative to the pure error (0.0031). There is a 12.96% chance that the lack-of-fit *F*-value this large could occur due to noise. The interaction between ethanol concentration and extraction temperature, as well as the quadratic terms of ethanol concentration and extraction temperature were not statistically significant terms. The statistically significant model *F*-value and not significant lack-of-fit *F*-value indicate the adequacy of the proposed model [33].

*R*<sup>2</sup> of 0.984 implies that 98.4% of the variation in the yield of phenolic compounds could be explained by the regression model (Table 4). Pred-*R*<sup>2</sup> of 0.821 was in reasonable agreemen<sup>t</sup> with the Adj-*R*<sup>2</sup> of 0.947, while Adj-*R*<sup>2</sup> was close to *R*2. The coefficient of variation of 1.99% (C.V. < 10%) indicates the low deviation between the experimental and predicted values of the response, and the high degree of precision and reliability. Adequate precision of 19.8 indicates an adequate signal so that this model can be used to navigate the design space. This parameter is a measure of the signal to noise ratio, and it is desirable to have the value higher than 4 [34].

The polynomial model that describes the extraction process and represents the interaction between factors and response is presented in Table 5.


**Table 4.** Analysis of variance (ANOVA) of the quadratic response surface model.

\* statistically significant; \*\* not statistically significant.

**Table 5.** Estimated regression coe fficients in the polynomial equation.


The not statistically significant terms could be excluded from the second order polynomial equation in order to improve the prediction ability of the proposed model. The regression coe fficients indicate that the linear e ffects have a positive impact on the response. The quadratic e ffects of ethanol concentration and extraction time, as well as the interaction between ethanol concentration and extraction temperature have a negative impact on the response. The extraction time of the linear e ffects had the highest impact on TPC, followed by extraction temperature and ethanol concentration.

The adequacy of the model was also evaluated by the residuals, which represent the di fference between the observed and predicted values of the response [35]. The residuals are thought of as the elements of variation unexplained by the regression model. The obtained residuals are plotted against the expected values in the normal probability plot (Figure 1). The obtained plots of the model after excluding nonstatistically significant terms indicate that the residuals are normally distributed. The slight deviation of points from the straight line in the reduced model indicates a better prediction of the regression model.

**Figure 1.** Normal probability plot of studentized residuals for the reduced polynomial model.

Cook's distances for the reduced polynomial model is depicted in Figure 2. Based on these values, the regression changes can be estimated when the case is deleted. Cook's distances were less than the limit of 1.0 so that there were no outliers in the given dataset.

**Figure 2.** Cook's distance for the reduced polynomial model.

#### *3.2. The Impacts of Factors on the Response Surface*

Figure 3a illustrates the interaction between ethanol concentration and extraction temperature for extraction time of 25 min. The yield of phenolic compounds increased with increasing ethanol concentration [36]. This impact on TPC is more significant for shorter extraction times. By means of analyzing the response shape, it can be concluded that there is a strong interaction between these factors. The increase of extraction temperature leads to increased TPC [36], but only using lower ethanol concentrations. The interaction between ethanol concentration and extraction time at 50 ◦C is depicted in Figure 3b. The impact of ethanol concentration on the response is significant at longer extraction times, while the effect of ethanol concentration at the shorter extraction times is almost negligible. The increase of extraction time is also significant at higher ethanol concentration levels and has a positive impact on the TPC [33]. Saturation in the response is achieved after extraction time of 30 min. The impacts of extraction temperature and time using 50% (v/v) ethanol are presented in Figure 3c. The extraction time has a more pronounced impact at higher extraction temperatures.

**Figure 3.** The impacts of: (**a**) ethanol concentration and extraction temperature for 25 min; (**b**) ethanol concentration and extraction time at 50 ◦C; and (**c**) extraction temperature and extraction time using 50% (v/v) ethanol on the TPC.

#### *3.3. Optimization of the Extraction*

The yield of phenolic compounds was maximized to obtain the optimal conditions for the extraction of these bioactive compounds [35]. Prior to applying the optimization method, the factor levels were ranged between −1 and +1. The optimal conditions were achieved for 60% (v/v) ethanol, 59 ◦C, and 30 min at the liquid-to-solid ratio of 10 cm<sup>3</sup> g<sup>−</sup>1. The predicted TPC under these conditions was 3.17 gGAE 100 g<sup>−</sup><sup>1</sup> d.p.m., while the TPC was found to be 3.12 gGAE 100 g<sup>−</sup><sup>1</sup> d.p.m. Based on the good agreemen<sup>t</sup> between obtained and predicted TPCs, it can be concluded that the proposed model is adequate.

Sarikurkcu et al. [8] determined the TPC of 56.74 mgGAE g<sup>−</sup><sup>1</sup> acetone extract, 36.42 mgGAE g<sup>−</sup><sup>1</sup> methanol extract, and 27.17 mgGAE g<sup>−</sup><sup>1</sup> aqueous extract of black locust flowers. Ji et al. [11] found the highest TPCs of 47.30 mgGAE g<sup>−</sup><sup>1</sup> d.p.m. for the freeze-drying method, and the lowest TPC of 29.15 mgGAE g<sup>−</sup><sup>1</sup> d.p.m. for sun drying method. The results of TPC obtained in this paper are in accordance with available data for different extraction techniques. Unlike previous studies, the extraction time for UAE is shorter so that the energy-efficient procedure was developed. The reduction in energy consumption was achieved by applying the ultrasound and advanced mathematical approach compared to other available procedures.

## *3.4. HPLC Analysis*

The identification and quantification of phenolic compounds in the extract of black locust flowers were carried out based on the retention times and UV spectra of the standards using the reversed-phase high-performance liquid chromatographic method with ultraviolet detection (RP-HPLC–UV) The contents of rutin (56.9 mg 100 g<sup>−</sup><sup>1</sup> d.p.m., Rt = 32.478 min, λmax = 250 nm), epigallocatechin (10.10 mg 100 g<sup>−</sup><sup>1</sup> d.p.m., Rt = 18.180, λmax = 250 nm), ferulic acid (6.76 mg 100 g<sup>−</sup><sup>1</sup> d.p.m., Rt = 30.789 min, λmax = 320 nm), and quercetin (2.44 mg 100 g<sup>−</sup><sup>1</sup> d.p.m., Rt = 50.096, λmax = 250 nm) were quantified. The lowest content of identified phenolic compounds was in the case of quercetin [10]. Veitch et al. [6] identified flavonol 3,7-di-O-glycosides, flavonoid robinin, glucosyl analogue of robinin, kaempferol, and isorhamnetin in methanolic extracts obtained by maceration of black locust flowers. Truchado et al. [4] also found robinin in nectar collected from black locust flowers (Bologna, Italy) using the HPLC–MS method. In addition to these studies, there are no available data related to the chromatographic analysis of given plant material.

#### *3.5. Comparison of Ultrasound-Assisted Extraction with Maceration and Soxhlet Extraction*

The TPC and half maximal inhibitory concentration (IC50) were determined for the extracts obtained by UAE, maceration, and Soxhlet extraction to compare the e fficiency of extraction techniques (Table 6). The results, which refer to the TPC, indicate that the yields of phenolic compounds are almost the same for the UAE and Soxhlet extractions. The UAE has been proven to be a more e fficient and profitable extraction technique of phenolic compounds compared to the Soxhlet extraction, since the extraction time was significantly shorter [25,37]. This fact is the result of cavitation, which causes the enhancement of mass transfer of bioactive compounds through the destroyed cell walls.


**Table 6.** The comparison of UAE with maceration and Soxhlet extraction.

The extract obtained by UAE gave better antioxidant activity than those obtained by Soxhlet extraction and maceration. The higher temperatures used in the Soxhlet extraction can probably cause the degradation of thermolabile phenolic compounds, leading to weak antioxidant activity. In the literature, there are data indicating that the methanolic extract of black locust flowers has the highest antioxidant activity (471.75 mg Trolox g<sup>−</sup><sup>1</sup> extract) compared to the extracts obtained using ethyl acetate, acetone, and water [8]. Ji et al. [11] determined that the ethanolic extracts of black locust flowers have the highest antioxidant activity previously dried by lyophilization.

The results of this study are hard to compare with available data since the extracts were obtained using di fferent solvents and assay for the determination of antioxidant activity. The obtained black locust extracts were represented as the main source of phenolic compounds with expressed antioxidant activity that can be used in the pharmaceutical and cosmetic industries.
