*2.5. Statistical Analysis*

Each extraction trial and all the analyses were carried out in three independent analysis performed in triplicate. The influence of individual factors on the TPC yield (single-factor experiment) was estimated by Analysis of Variance (ANOVA) and Tukey's post hoc test with a 95% confidence level, while data obtained from the BBD and Central Composite Design (CCRD) trials were analyzed through ANOVA for the response variable to evaluate the model significance and suitability. Significant and highly significant levels were set for *p* < 0.05 and *p* < 0.01, respectively. The John's MacIntosh Product (Version 7.0, SAS, Cary, NC, USA) and Design-Expert (Trial version 10.0, SAS, Cary, NC, USA) software packages were used to construct the BBD and CCRD and to analyze all the results. Principal Component Analysis (PCA) was applied to detect the relationships between contents of phenolic compounds, flavonoids, anthocyanins, tannins, as well as antioxidant activity and their extraction methodologies i.e., UAE, MAE and CSE. All tests were done in triplicate.

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

#### *3.1. Optimization of UAE Conditions*

#### 3.1.1. Modeling and Fitting the Model Using RSM

The experimental design and subsequent response allied to TPC are summarized in Table 1, with results from TPC recovery varying in the range of 79–235 mg GAE/g DW.

**Table 1.** Central composite design with the observed responses and predicted values for yield of total phenolic compounds of *Myrtus communis* pericarp using the UAE method.


TPC results are expressed as means ± standard deviation; GAE, gallic acid equivalent; UAE, ultrasound-assisted extraction; DW, dry weight.

The least square technique was used to calculate the regression coefficients of the intercept, linear, quadratic, and interaction terms [26] (Table 2). Notably, the linear parameters, namely ethanol concentration, irradiation time and liquid–solid ratio (*p* < 0.0001), followed by amplitude (*p* = 0.0421) significantly affected the extraction content of phenolic compounds. The quadratic terms *X2*<sup>2</sup> and *X4*<sup>2</sup> were highly significant at the level *p* < 0.001, while the *X1*<sup>2</sup> and *X3*<sup>2</sup> terms were insignificant

(*p* > 0.05). Regarding TPC yield, the interaction of ethanol concentration with amplitude of ultrasound (*X1*–*X3*) and with liquid to solid ratio (*X1*–*X4*), and that of irradiation time amplitude of ultrasound (*X2*–*X3*) were highly significant (*p* < 0.0001), followed by irradiation time with liquid-to-solid ratio (*p* = 0.0054), amplitude of ultrasound with liquid-to-solid ratio (*p* < 0.0094) and ethanol concentration with irradiation time (*p* = 0.0367). Those significant terms played a dominant role in myrtle pericarp extraction by ultrasound. Indeed, those significant terms played a dominant role in myrtle pericarp extraction by ultrasound. This is justified by the analyses of variance, as represented in Table 2, with significant *p*-values (*p* < 0.0001) for the linear parameters, namely ethanol concentration (*X1*), irradiation time (*X2*) and liquid–solid ratio (*X4*), the quadratic terms *X2*<sup>2</sup> and *X4*2, and the interaction terms (*X1*–*X3*), (*X1*–*X4*) and (*X2*–*X3*). However, insignificant *p*-values (*p* > 0.0001) were obtained for the third linear parameter, amplitude (*X3*), the quadratic terms *X1*<sup>2</sup> and *X3*<sup>2</sup> and the interaction terms (*X1*–*X2*), (*X2*–*X4*) and (*X3*–*X4*).


**Table 2.** Estimated regression coefficients for the quadratic polynomial model and analyzes of variance (ANOVA) for the experimental results.

a Degree of freedom; b the model mean square to error mean square ratio; c corrected total. DF, degree of freedom; FRatio,freedomratio;Prob,probability;C.V.,coefficient of variation.

Based on the significant terms, the regression equation for the UAE efficiency was obtained as follows:

$$\begin{array}{c} Y = 205.032 + 10.998X\_1 + 14.043X\_2 - 3.994X\_3 + 27.390X\_4 + 5.049X\_1X\_2 - 11.460X\_1X\_3 - 10.588X\_1X\_4 \\ 10.588X\_1X\_4 + 15.196X\_2X\_3 - 7.198X\_2X\_4 - 6.580X\_3X\_4 - 1.526X\_1^2 - 11.561X\_2^2 - 0.888X\_3^2 - \\ 17.876X\_4^2 \end{array} \tag{17.876X\_1}$$

Note that the *p*-value can be employed to check the interaction strength between independent factors. From this analysis, *p*-value <0.0001 indicated that the response surface quadratic model was significant, which means that the model represented the data satisfactorily. The adjusted coefficient of determination *(R*2adj) and the coefficient of determination (R2) were 0.9553 and 0.9776, respectively, which implied that the sample variations of 97.76% for the UAE efficiency of myrtle pericarp phenols were attributed to the independent variables, and only 2.24% of the total variations could not be explained by the model, indicating a good degree of correlation between experimental and predict values of the TPC yield. In addition, the low value of coefficient of variance (3.71%) clearly indicated that the model was reproducible and reliable [27]. All these results indicate that the model could work well for the prediction of TPC in the myrtle pericarp extracts.

#### 3.1.2. Response Surface Analysis (RSA)

To provide a better understanding of the interaction between factors, the 3D response surface plot was constructed (Figure 1) using Equation (4). The graphs were generated by plotting the response using the z-axis against two independent variables, while keeping the other independent variable at the fixed level. Figure 1A–C shows the interactions between the ethanol concentration and each of the three other factors, namely irradiation time, amplitude and liquid-to-solid ratio, respectively, on the recovery of TPC. As shown, an increase of ethanol concentration from 20% to 80% (*v*/*v*), or extraction time from 5 to 10 min resulted in a rapid enhancement of TPC with a maximum of 235.21 mg GAE/g being recovered with an irradiation time of 7.5 min and ethanol concentration of 70% (*v*/*v*).

The high phenolic content indicates that the mixture ethanol/water at 70% (*v*/*v*) allowed the solubilization of phenolics from *M. communis* pericarp, thus confirming the results of the single factor experiments [28] that explained the efficiency of the ultrasonic method by the fact that sonication improved the hydration and fragmentation process and hence facilitates the mass transfer of solutes to the extraction solvent. For the extraction yield of TPC performed at fixed extraction time and liquid-to-solid ratio, with varying ethanol concentration and amplitude (Figure 1B), it was possible to conclude that maximum recovery (210.05 mg GAE/g) was achieved for 70% (*v*:*v*) of ethanol and an ultrasound amplitude of 35%. This fact can be explained by the larger amplitude ultrasonic wave that promotes the liquid medium to produce more cavitation bubbles, thus resulting in a stronger pressure, capable of destroying the cell wall and accelerating mass transfer [29]. Figure 1C shows an enhancement of TPC that reached a peak value of 230.15 mg GAE/g, for 70% (*v*:*v*) ethanol and about 30 mL/g of liquid-to-solid ratio. A higher ratio corresponds to a greater concentration difference between the exterior solvent and the interior tissues of *Myrtus* pericarp. It prominently prompted the TPC to be rapidly dissolved, which resulted in an increase in the extraction yield. The response surface plot for the significative interactive effect of irradiation time and amplitude of ultrasound on the response value at a fixed ethanol concentration and liquid-to–solid ratio is shown in Figure 1D. A higher TPC was obtained with the irradiation time at 10 min and amplitude of 30%; these results confirm those reported in the literature [30,31]. Figure 1E shows an interaction between extraction time and the liquid-to-solid ratio (*p* < 0.05). The best content (148 mg GAE/g) was found with the solid–liquid ratio of about 30 mL/g and the radiation time of 10 min. The increase of the ethanol proportion required high sonication intensity to generate the cavitation bubbles. However, a higher increase in the liquid-to-solid ratio diminished the supply of ultrasonic energy density and negatively affected the extraction yield.

The yield of TPC constantly improved with the increase of both amplitude of ultrasound and liquid-to-solid ratio, reaching a maximum when *X3* and *X*4 became 32% and 20% (*v*/*v*), respectively (Figure 1F). Beyond this level, the yield of TPC reduced with the increase of *X*1 and *X*4. Hence, the interactive effect of *X3* and *X4* was remarkable. Overall, these results indicate that the TPC extraction yield was more significantly affected (*p* < 0.0001) by linear parameters, namely ethanol concentration, irradiation time and liquid-to-solid ratio.

**Figure 1.** Response surface analysis for the Total phenolic compounds (TPC) with UAE with respect to: (**a**) ethanol concentration and irradiation time; (**b**) ethanol concentration and amplitude; (**c**) ethanol concentration and solvent-to-solid ratio; (**d**) extraction time and amplitude; (**e**) extraction time and solvent-to-solid ratio; and (**f**) amplitude and solvent-to-solid ratio.

#### 3.1.3. Validation and Verification of the Predictive Model

According to the result of response surface and prediction by this built model, the optimal conditions were thus obtained for the following conditions: ethanol at 70% (*v*/*v*), 7.5 min extraction time, 30% amplitude and a liquid-to-solid ratio of 28 mL/g. To ensure that the predicted result was not biased to the practical value, experimental rechecking was performed using these deduced optimal conditions. The predicted extraction yield of TPC in UAE-OPT was 235.52 ± 9.9 mg GAE/g, that was consistent with the experimental yield of 241.66 ± 12.77 mg GAE/g DW (Table 3). The results showed no significant di fference between the experimental and the predicted values. This strong correlation between experimental and the predicted values indicates that the response of regression model is adequate to reflect the expected optimization for the extraction of antioxidants from *M. communis* pericarp.

#### *3.2. Comparison between UAE, MAE and CSE Methods*

Remarkably, the highest TPC was obtained by UAE (241.60 ± 12.77 mg GAE/g). This corresponded to four and three times higher than that obtained by MAE and CSE, respectively, thus indicating that the application of UAE has a positive e ffect on the extraction of TPC (Table 3). The highest levels of TPC in UAE-OPT extract was reflected by its higher amounts of flavonoids, anthocyanins and tannins (18.99 ± 1.31 mg QE/g; 25.06 ± 0.36 mg/g; 35.56 ± 0.36 mg CE/g, respectively). These findings are consistent with those reported in the literature [32] and are mainly attributed to the fact that ultrasound radiation can facilitate mass transfer and accelerate the extracting process so that the extraction of bioactive compounds may be improved. Hence, according to the overall data, it is possible to conclude that the herein optimized UAE process yields higher levels of bioactive compounds in a short time and requires less solvent consumption than MAE and CSE. Note that, in this study, the operating temperature in the UAE-OPT was kept constant at room temperature, excluding any heating e ffect. This might positively or negatively influence the polyphenols recovery depending on the applied amplitude.


QE, quercetin; RP, ferric reducing antioxidant power; US amp, ultrasound amplitude.

**Table 3.** Comparison of extraction yield of polyphenols obtained by optimized ultrasound-assisted (UAE-OPT), microwave-assisted (MAE) and conventional

344

The antioxidant capacity of the extracts was assessed by DPPH - scavenging and ferric reducing antioxidant power assays. The results show that UAE-OPT extract presented higher DPPH - scavenging ability (90.71% inhibition) when compared to CSE (88.03% inhibition) and MAE (87.16% inhibition) extracts. The same tendency was also observed for reducing power, since the absorbance at 700 nm for UAE-OPT extract was considerably higher than those obtained for MAE and CSE (0.439 ± 0.006 and 0.429 ± 0.01, respectively). This means that UAE method is more e fficient for the recovery of antioxidants than the herein tested microwave and conventional solvent extraction methods, a fact that is probably due to its superior richness in phenolic components, including flavonols [33] and as evidenced in the following section. This information was also confirmed by PCA analysis. PCA was applied to the extracts (UAE-OPT, MAE and SCE) for phenolic compounds (TPC, flavonoids, anthocyanins and tannins) and antioxidant activity, where the two chosen factors justified 100.0% of total variance. The resulting plots allowed selecting the better extraction method of di fferent compounds of myrtle pericarp, and clearly divided the samples into three groups, depending on the extraction method (Figure 2). For PC1, which explains 95.91% of the total variance, the first group showed a positive correlation with PC1, thus confirming that UAE was the best extraction method for phenolic compounds with potent antioxidant activity. The highest correlation was found between antioxidant activity (DPPH• and RP essay) and anthocyanins, hence suggesting that these compounds might have a key influence on the antioxidant capacity of the extracts. The best correlation between the MAE and TPC, flavonoids and tannins were observed in the second group. PC2 explains only better 4.09% of the experimental variability, which could essentially be associated to the CSE method and anthocyanins content (the third group).

**Figure 2.** Principal component analysis of phenolic compounds for *M. communis* pericarp with UAE, MAE and CSE. FLAV, flavonoids; ANTHO, anthocyanins; TANN, tannins.

#### *3.3. Identification of Phenolics by UHPLC-DAD-ESI-MSn Analysis*

The UAE-OPT extract was analyzed by UHPLC-DAD-ESI-MSn to further elucidate its phenolic profile. The registered chromatogram at 280 nm is shown in Figure 3 and the UV-Vis and MSn spectral data of eluted peaks are summarized in Table 4.

**Figure 3.** Chromatographic profile at 280 nm of *M. communis* pericarp extract obtained by UAE extraction at optimized conditions. Numbers in the figure correspond to the eluted UHPLC peaks for which UV and MS data are summarized in Table 4.

Among the distinct phenolic groups found in the extract, flavonols were the prevalent components. Overall, eleven flavonol glycosides were detected, being myricetin glycosides, namely myricetin-*O*-hexoside and myricetin-*O*-deoxyhexoside (eluted in Peaks 10/11 and 14/15, respectively) the major abundant ones, which probably correspond to myricetin-3- *O*-galactoside and myricetin-3- *O*-rhamnoside, since these are known to be present as main phenolic components in distinct organs of *M. communis* plant [34–38].

Besides the above compounds, four othermyricetin glycosides were foundin the extract. The compound eluted in Peak 9, showing a [M-H]− at *m*/*z* 631, corresponded to myricetin-*O*-galloyl-hexoside, since the main fragments in MS<sup>2</sup> spectrum were formed by the loss of 152 Da (equivalent to a galloyl moiety) and 332 Da (equivalent to the simultaneous loss of galloyl and hexosyl units). This could possibly correspond to myricetin 3-(6"- *O*-galloyl galactoside), which has been previously reported in leaves [32,34,36,38] and berries [16]. In addition, the compounds with [M-H]− at *m*/*z* 449 (Peak 13) and at *m*/*z* 625 (co-eluted in Peak 18) were, respectively, assigned to myricetin-*O*-pentoside and myricetin-*O*-hexosyl-deoxyhexoside, according to their fragmentation pattern, which showed the loss of a pentosyl (132 Da) and deoxyhexosyl plus hexosyl (308 Da) moieties, respectively. In turn, the compound eluted in Peak 19 at 14.6 min with a pseudomolecular ion at *m*/*z* 569 and fragment ions at *m*/*z* 485 (equivalent to galloyl ester moiety) and 317 (myricetin) was tentatively assigned to a galloylester of myricetin.

The three remaining flavonols detected in the UAE-OPT extract were assigned to quercetin and kaempferol derivatives. From those, the compound eluted in Peak 12 was characterized by a [M-H]− at *m*/*z* 615 and fragment ions at *m*/*z* 463 (−152 Da, loss of galloyl group) and 301 (−162 Da, loss of an hexosyl group), and was tentatively assigned to quercetin-*O*-hexoside-gallate on the basis of data reported in the literature [39–41]. This compound has been already reported in Myrtaceae family, namely in *Eucalyptus* species [42–44] and two other species from the same family, namely *Myrcia multiflora* extracts [45] and *Eugenia edulis* [46]. In addition, the compound eluted in Peak 17 with a deprotonated ion at *m*/*z* 447 and a base peak fragment ion at *m*/*z* 301 (−146, equivalent to the loss of a deoxyhexose unit), was identified as quercetin-*O*-deoxyhexoside according to literature data, probably corresponding to quercetin-3- *O*-rhamnoside [39,40]. This last flavonoid was previously detected in pericarp [29], berries [31,47] and leaves [30,31,35] of *M. communis*. Finally, the flavonol eluted in Peak 16 ([M-H]− at *m*/*z* 447) presented the main fragment ion at *m*/*z* 285 in the MS<sup>2</sup> spectrum, which in turn showed a fragmentation pattern coherent with kaempferol. Based on UV-Vis spectra (UVmax at 265 and 353) and MSn spectral data, this compound was assigned to kaempferol-*O*-hexoside.


**Table 4.** UHPLC-DAD-ESI-MSn data for *M. communis* pericarp extract obtained under optimized UAE conditions.

 *8*, 205

λmax, wavelength of maximum absorbance.

Besides flavonols, other flavonoids in UAE-OPT extract corresponded to anthocyanins that were eluted from 7.6 min to 9.7 min (Peaks 6–8). Note that, in general, anthocyanins are preferred detected as [M]+ in ESI in the positive mode, while typically they show [M-2H]− in the negative mode [47], as represented in Table 4. Overall, according to UV-Vis and MSn spectral data, these compounds were assigned to delphinidin, petunin and malvidin derivatives. In more detail, the compound in Peak 6 exhibiting a [M-2H]− at *m*/*z* 463 and a base peak MS<sup>2</sup> fragment ion at *m*/*z* 301 (−162 Da) was assigned to delphinidin-*O*-hexoside by comparison with data reported in the literature [48–50]. In turn, petunidin-*O*-hexoside and a petunidin-*O*-hexoside derivative were eluted in Peaks 7 and 8, respectively. The first showed a [M-2H]− at *m*/*z* 477 and a main MS<sup>2</sup> fragment ions at *m*/*z* 315/314 [48–50] while ions corresponding to petunidin-*O*-hexoside and its hydrated form (at *m*/*z* 477 and *m*/*z* 495, respectively) were predominant in MS<sup>2</sup> spectrum of the latter compound. The petunidin-*O*-hexoside derivative was co-eluted with malvidin-*O*-hexoside ([M-2H]− at *m*/*z* 477→329). Note that, except for petunidin-*O*-hexoside, hexosides of delphinidin, petunin and malvidin have already been described in distinct organs of *M. communis*, including pericarp [14,30,31,51,52].

Several non-flavonoid compounds could also be observed in UEA-OPT extract, including caffeoyl hexoside, gallic acid and galloyl derivatives. The first ([M-H]− at *m*/*z* 341→179, eluted in Peak 1) was the only hydroxycinnamic acid found in the extract. Gallic acid ([M-H]− at *m*/*z* 169→125, eluted in Peak 4), has been described in the literature for extracts obtained from the pericarp [16] berries [16,30] and leaves [38,52].

Regarding galloyl derivatives (typical UVmax at 273–276 nm), these enclosed esters of monoor di-galloyl groups with a hexose or quinic acid unit, or even with myrtucommulone-type groups. In detail, the compound eluted in Peak 2 with a [M-H]− at *m*/*z* 331 and corresponding fragments at *m*/*z* 271, 169, 241, 211, 193 and 125, was assigned to a galloyl hexoside [53], presumably galloyl-3-*O-*β-D-galactoside-6-*O*-gallate, since this latter has been previously reported in *M. communis* leaves [2,38]. Besides, two isomers of galloyl quinic acid ([M-H]− at *m*/*z* 343→191, 169, 125) could be found in Peaks 3 and 5, while a digalloyl hexoside ([M-H]− at *m*/*z* 483→271, 331, 313, 439, 193, 169) and digalloyl quinic acid ([M-H]− at *m*/*z* 495→343, 325, 191, 169) were detected as co-eluted compounds in Peak 6. All these galloyl derivatives have been previously detected in *M. communis* leaves [35,38,52].

Moreover, four gallomyrtucommulone-type derivatives were found in UAE-OPT extract. All these compounds showed a UVmax at 276 nm, and similar fragment ions in MSn spectra, including ions at *m*/*z* 331, 313, 271 and 211, which are typically formed in galloylhexoside [43]. Indeed, the ion at *m*/*z* 331 correspond to the galloyl hexoside moiety, while ions at *m*/*z* 271 and *m*/*z* 211 result from the cross-ring fragmentation of the hexose unit in the galloyl hexoside moiety and that at *m*/*z* 313 can be formed due to the loss of water molecule from the latter. Among these compounds, those eluted in Peaks 20 and 21 ([M-H]− at *m*/*z* 583 and 567, respectively) were assigned to gallomyrtucommulone F and gallomyrtucommulone C, in accordance to previous data reported in *M. communis* leaves [36]. Besides these two compounds, the extract also contained two isomeric unidentified gallomyrtucommulone-type derivatives (MW 468 Da) that presumably vary in their acyl chain regarding those previously identified.
