**The E**ff**ect of Di**ff**erent Extraction Protocols on** *Brassica olerace***a var.** *acephala* **Antioxidant Activity, Bioactive Compounds, and Sugar Profile**

**Nikola Major 1,\* , Bernard Prekalj 1,2, Josipa Perkovi´c <sup>1</sup> , Dean Ban 1,2, Zoran Užila 1,2 and Smiljana Goreta Ban 1,2,\***


Received: 26 November 2020; Accepted: 16 December 2020; Published: 17 December 2020 -

**Abstract:** The extraction of glucosinolates in boiling aqueous methanol from freeze dried leaf tissues is the most common method for myrosinase inactivation but can be hazardous because of methanol toxicity. Although freeze drying is the best dehydration method in terms of nutritional quality preservation, the main drawbacks are a limited sample quantity that can be processed simultaneously, a long processing time, and high energy consumption. Therefore, the aim of this study is to evaluate the effects of applying high temperature for myrosinase inactivation via hot air drying prior to the extraction step, as well as the effects of cold aqueous methanol extraction on total antioxidant activity, total glucosinolates, total phenolic content, and sugar profile in 36 landraces of kale. The results from our study indicate that cold aqueous methanol can be used instead of boiling aqueous methanol with no adverse effects on total glucosinolate content. Our results also show that hot air drying, compared to freeze drying, followed by cold extraction has an adverse effect on antioxidant activity measured by DPPH radical scavenging, total glucosinolate content, as well as on the content of all investigated sugars.

**Keywords:** antioxidant activity; extraction; glucosinolates; kale (*Brassica oleracea* var. *acephala*); phenolics

#### **1. Introduction**

Vegetables from the Brassicaceae family are known for their excellent nutritional value and are abundant in carbohydrates, vitamins (ascorbic acid, folic acid, β-carotene, α-tocopherol), macro and micro elements (iron, calcium, selenium, copper, manganese, zinc), as well as secondary metabolites, including glucosinolates, phenolics (tannins, phenolic acids, anthocyanidins, flavonols, coumarins, flavones), and other bioactive molecules, such as phytosterols and terpenoids [1,2]. They are also known for their antimicrobial and anticancerogenic activity [3,4]. One of the well-known members of the Brassicaceae family is kale (*Brassica oleracea* var. *acephala*). It originates and is traditionally used in Mediterranean countries, but has gained special attention and popularity in the U.S., and later worldwide over the last decade [5]. It is characterized by leaves, which do not form a head, unlike other leafy vegetables from the Brassicaceae family as white cabbage, savoy cabbage, Brussels sprouts, and Chinese cabbage [6].

The most abundant sulfur containing compounds in plants from the Brassicaceae family are glucosinolates and S-methylcysteine sulfoxide [7]. Cruciferous vegetables are extensively researched regarding their beneficial compounds especially ones active in cancer prevention. Isothiocyanates are degradation products from glucosinolates are known for their antioxidant, immunostimulatory, anti-inflammatory, antiviral and antibacterial properties [8–10]. Isothiocyanates are converted from glucosinolates in a reaction catalyzed by the enzyme myrosinase which activates and releases upon plant tissue injury. The myrosinase enzyme was first discovered in 1840 from the mustard seed after which the Brassica studies were focused on clinical studies and prospective benefits of cruciferous vegetables [1]. Myrosinase catalyzes glucosinolate hydrolysis after the plant tissue is injured or in any way disrupted. Therefore, accurate glucosinolate analysis depends on myrosinase inactivation [11]. Differences in glucosinolate content regarding Brassica species and environmental factors can be an obstacle when comparing different studies regarding nutritional and bioactive quality of cruciferous vegetables. Moreover, presence of the enzyme myrosinase, activated in plant handling and extraction processes that precede glucosinolate determinations, is also an aggravating circumstance, which can lead to a decrease in total glucosinolate concentration [12]. Most published glucosinolate analysis methods employ dehydration by freeze drying for myrosinase inactivation and denaturation by moderately high temperature during the glucosinolate extraction step [12,13]. This process is outlined in the ISO9167:2019 standard, as well as in the work by Grosser and van Dame [14]. Freeze drying is a dehydration process based on water sublimation under vacuum, which protects the primary structure and shape of the sample [15]. Although freeze drying is considered the best form of dehydration, the process is time consuming and relatively expensive [16], especially for high throughput methods. On the other hand, hot air drying is simple, fast, and can handle large quantities of samples.

Most published methods for glucosinolate analysis employ boiling aqueous methanol (70/30, methanol/water, *v*/*v*) in the glucosinolate extraction step [17–21]. Methanol is a common organic solvent and reagent in organic synthetic procedures [22]. Poisoning can occur via methanol ingestion, skin absorption, or inhalation [23]. Acute methanol toxicity evolves in a well-understood pattern and results in metabolic acidosis via formic acid formation and superimposed toxicity to the visual system [24]. The use of boiling methanol in the glucosinolates extraction step increases the risk of poisoning and researchers offered several less hazardous alternatives to the standard method, including the use of hot water or the use of cold (ambient temperature) aqueous methanol [12]. The 2019 revision of the ISO9167 method for the determination of glucosinolates in rapeseed and rapeseed meals by HPLC replaces aqueous methanol (70/30, *v*/*v*) by aqueous ethanol (50/50, *v*/*v*) for lower toxicity [25]. Although the ISO9167:2019 official method supports the use of ethanol instead of methanol for glucosinolate extraction in rapeseed and rapeseed meals, aqueous methanol is the recommended extraction solvent due to the structural and biological diversity of seed and plant matrices, but can be replaced by other solvents if appropriate validation procedures are applied [26].

The effect of hot air drying versus freeze drying on phenolic compounds and antioxidant activity was a subject of numerous studies [27–30] but there is not enough data on the possible effects on glucosinolate and sugar content. While there are several studies concerning the comparison of cold versus hot methanol extraction [12] on glucosinolate levels, the number of samples involved could have been higher, and the studies did not include other bioactive compounds or antioxidant activity in the comparison.

Therefore, the aim of this work is to evaluate the effect of applying high temperature for myrosinase inactivation via hot air (oven) drying prior to the extraction step, as well as the effect of cold (ambient temperature) aqueous methanol extraction on total glucosinolates content, total phenolic content, total antioxidant activity, and sugar content in 36 landraces of kale (*Brassica oleracea* var. *acephala*).

#### **2. Results**

#### *2.1. The E*ff*ect of Di*ff*erent Extraction Protocols on Antioxidant Activity, Bioactive Compounds, and Sugar Content in Brassica*

The extraction of glucosinolates by the hot methanol extraction step from freeze dried kale leaves yielded significantly lower amounts of total glucosinolates (30.3 ± 0.6 mg sinigrin equivalents (SEQ)/g dry weight (DW), compared to the cold methanol extraction step (34.3 ± 0.9 mg SEQ/g DW) (Figure 1). Significantly higher content of total glucosinolates was found in extracts obtained from freeze dried

kale leaves (34.3 ± 0.9 mg SEQ/g DW) compared to oven dried leaves followed by cold extraction (31.9 ± 0.7 mg SEQ/g DW) (Figure 1).

denote significant difference by Tukey's Unequal N — — — — **Figure 1.** The effect of different extraction protocols on total glucosinolates content, total phenolic content, DPPH radical scavenging activity and ferric ion reducing antioxidant power in *B. oleracea* var. acephala leaf extracts. Values are expressed as mean ± SE (N = 324). The different letters above bars denote significant difference by Tukey's Unequal N Honestly Significant Difference (HSD) test, *p* < 0.05. TGls—total glucosinolates; TPC—total phenolic content; DPPH—DPPH radical scavenging activity; FRAP—Ferric ion Reducing Antioxidant Power.

The same effect was observed for total phenolic content, where significantly lower amounts were extracted with the hot methanol method (10.1 ± 0.2 mg gallic acid equivalents (GAEQ)/g DW) compared to the cold methanol extraction (10.6 ± 0.2 mg GAEQ/g DW) in freeze dried leaves (Figure 1). On the other hand, total phenolic content was found to be higher in oven dried leaves (15.7 ± 0.2 mg GAEQ/g DW) compared to freeze dried B. oleracea var. acephala leaves followed by cold extraction (10.6 ± 0.2 mg GAEQ/g DW) (Figure 1).

As for the antioxidant activity we found significantly higher Ferric Reducing Antioxidant Power (FRAP) values from samples extracted from freeze dried leaves with cold methanol (92.5 ± 1.8 µmol Fe <sup>2</sup>+/g DW) compared to hot methanol extraction (76.5 <sup>±</sup> 1.7 <sup>µ</sup>mol Fe <sup>2</sup>+/g DW) (Figure 1). The hot methanol extraction method yielded higher antioxidant power measured by 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical scavenging activity (28.8 ± 0.7 µmol Trolox equivalents (TEQ)/g DW) compared to the cold methanol extraction method (27.0 ± 0.7 µmol TEQ/g DW) from freeze dried B. oleracea var. acephala leaves (Figure 1). DPPH radical scavenging was found to be significantly higher in freeze dried B. oleracea var. acephala leaves (27.0 ± 0.7 µmol TEQ/g DW) compared to oven dried samples (20.2 ± 0.6 µmol TEQ/g DW) followed by cold extraction (Figure 1). FRAP values were significantly higher in oven dried B. oleracea var. acephala leaves (105.7 ± 1.5 µmol Fe2+/g DW) compared to freeze dried leaves (92.5 <sup>±</sup> 1.8 <sup>µ</sup>mol Fe <sup>2</sup>+/g DW) followed by cold extraction (Figure 1).

Both sucrose and fructose content was significantly lower in the freeze dried extracts obtained by the hot methanol extraction method (7.8 ± 0.3 g/100g DW and 4.4 ± 0.2 g/100g DW for sucrose and fructose, respectively) compared to the cold methanol extraction method (10.0 ± 0.3 g/100g DW and 9.4 ± 0.2 g/100g DW for sucrose and fructose, respectively) (Figure 2). Glucose content was found to be no different in freeze dried sample extracts obtained by hot or cold methanol extraction (8.0 ± 0.3 g/100g DW and 8.0 ± 0.2 g/100g DW, respectively) (Figure 2). Extracts obtained by cold extraction from oven dried tissues exhibited significantly lower sucrose, glucose and fructose levels (5.2 ± 0.3 g/100g DW, 2.9 ± 0.1 g/100g DW and 4.6 ± 0.1 g/100g DW, respectively) compared to those obtained from freeze dried leaf samples (10.0 ± 0.3 g/100g DW, 8.0 ± 0.2 g/100 g DW and 9.4 ± 0.2 g/100g DW, respectively) (Figure 2).

denote significant difference by Tukey's Unequal N HSD test, **Figure 2.** The effect of different extraction protocols on sucrose, glucose and fructose content in *B. oleracea* var. acephala extracts. Values are expressed as mean ± SE (N = 324). The different letters above bars denote significant difference by Tukey's Unequal N HSD test, *p* < 0.05.

#### *2.2. Multivariate Analysis of the Data Obtained by Di*ff*erent Extraction Protocols*

– The obtained data was processed with Partial Least Squares–Discriminant Analysis (PLS-DA) to further clarify the relations between different extraction protocols. The PLS-DA model was employed due to the supervised nature of the learning algorithm.

Overall, the extracts which included oven drying leaf tissue following cold methanol extraction were separated on the primary axis from the extracts obtained from freeze dried leaf tissue followed by cold or hot methanol extraction (Figure 3). The highest contribution to the sample placement on the primary axis had total phenolic content, DPPH radical scavenging activity, glucose, and sucrose content (Figure 3). On the second axis, the extracts, which were obtained from freeze dried leaf tissues following hot or cold methanol extraction, were separated according to fructose, total glucosinolates, and FRAP values (Figure 3).

– — — — — **Figure 3.** Partial Least Squares–Discriminant Analysis (PLS-DA) analysis of data obtained by different extraction protocols (N = 324); TGls—total glucosinolates; TPC—total phenolic content; DPPH—DPPH radical scavenging activity; FRAP—Ferric ion Reducing Antioxidant Power.

≥ 0.05) were observed b The model showed that the largest difference between extraction protocols was in fructose content where the freeze drying and cold methanol combination yielded the highest amount of the compound. Total phenolic content was the second most important variable in the differentiation between the observed extraction protocols where the oven dried and cold methanol combination exhibited the highest yield of phenolic compounds. The third most important variable in differentiating among extraction protocols was glucose where similar amounts were extracted with either hot or cold methanol from freeze dried leaf tissue, whereas several-fold lower amounts were extracted from oven dried tissue with cold methanol. The next most important variable was FRAP where the highest values were observed in extracts obtained from oven dried tissues by cold methanol. Sucrose was the fifth most important variable where the oven dried leaf tissues had lower yields of the compound compared to extracts from freeze dried tissues. The sixth variable, according to the discriminating power, was DPPH radical scavenging, where again the extracts obtained from oven dried tissues exhibited lower antioxidant activity compared to extracts obtained from freeze dried leaf tissues. The variable with the lowest discriminating power according to the obtained model was total glucosinolates where cold methanol extraction from freeze dried Brassica tissue exhibited the highest yield compared to the other two extraction protocols.

#### *2.3. Correlations between Bioactive Compounds, Antioxidant Activity, and Sugar Content in Brassica Extracts*

Statistically significant correlations (*p* ≥ 0.05) were observed between bioactive compounds content, antioxidant activity, as well as between sugar content (Table 1). Positive correlations were observed between total phenolic content, total glucosinolates, and FRAP values (Table 1). Negative correlations were observed between total phenolic content and sucrose, glucose, and fructose content, as well as total glucosinolates and glucose content (Table 1). DPPH scavenging activity correlated positively with

sugar content but FRAP correlated negatively with sucrose and glucose content (Table 1). Positive correlations were observed between sucrose, glucose, and fructose content (Table 1).


**Table 1.** Pearson's correlations between bioactive compounds content, antioxidant activity and sugar content in B. oleracea var. acephala extracts (N = 324).

<sup>1</sup> Numbers highlighted in red are statistically significant (*<sup>p</sup>* <sup>≤</sup> 0.05); TGls—total glucosinolates; TPC—total phenolic content; DPPH—DPPH radical scavenging activity; FRAP—Ferric ion Reducing Antioxidant Power.

#### **3. Discussion**

Biological activity of kale containing phytochemicals are associated with antioxidant, anti-cancerogenic activity and protection of gastrointestinal and cardiovascular system [4,31–33]. Sikora et al. found that kale had much higher antioxidant activity than broccoli, Brussels sprouts, and cauliflower [34]. Šamec et al. on the other hand found that white cabbage and kale sprouts had significantly higher antioxidant capacity, polyphenol and glucosinolate content compared to arugula, broccoli, and Chinese cabbage sprouts [20]. Kale is also abundant in organic acids, including citric, malic, pyruvic, shikimic fumaric, and aconitic acids [35].

The freeze drying process is regarded as the best method for sample dehydration because it can prolong shelf life without compromising nutritional quality [36]. The major drawback of this method is limited sample quantity that can be processed simultaneously, long processing time (up to several days) and high energy consumption [37]. On the other hand, oven drying is readily accessible to most laboratories, sample quantity is seldom an issue and the drying time can be as low as one hour but rarely exceeds 24 h.

In our study, freeze dried leaf tissues followed by cold extraction yielded significantly higher total glucosinolate content compared to either oven dried leaf tissue followed by cold extraction or freeze dried leaf tissue followed by hot extraction. The results are in accordance with Rutnakornpituk et al., who investigated the effect of freeze and oven drying on phenolic and glucosinolate content, as well as, antioxidant activity in five cruciferous vegetables where the authors concluded that freeze dried leaf tissues yielded higher glucosinolate content in all investigated species [38]. On the other hand, Tetteh et al. studied the effect of different various drying methods on post-harvest glucosinolate content in *Moringa oleifera* leaf tissues where no differences were observed between oven dried and freeze dried leaf tissue samples [39]. Doheny-Adams et al. investigated the effect of hot and cold methanol extraction on glucosinolate content in five Brassicaceae species [12]. The results from their study showed that the hot extraction step can be replaced with cold extraction with no losses in glucosinolate content, which is in line with the findings presented in our study [12].

Oven drying, compared to freeze drying followed by cold extraction yielded significantly higher total phenolic content and FRAP values and lower total glucosinolate content and DPPH scavenging activity in *B. oleracea* var. *acephala* leaf extracts. Similarly, Managa et al. showed that the majority of investigated phenolic compounds in Chinese cabbage significantly increased in content in oven dried compared to freeze dried samples [40]. Although the phenolic compound content in Chinese cabbage was higher, the antioxidant activity measured by FRAP, DPPH, and ABTS, was significantly lower in the oven dried compared to freeze dried samples [40]. On the other hand, Korus found that freeze dried kale had significantly higher total phenolic content and antioxidant activity compared to air dried kale, albeit at a lower temperature (55 ◦C) compared to this study (105 ◦C) [41]. Papoutsis et al. found higher total phenolic content and DPPH scavenging activity in hot air dried compared to freeze dried lemon [28]. Que et al. also found higher total phenolic content and antioxidant activity in hot air dried compared to freeze dried pumpkin flour [27]. Das et al. studied the antioxidant properties of freeze dried and oven dried wheatgrass where higher phenolic content but lower FRAP and DPPH radical scavenging activity was found in hot air dried compared to freeze dried samples [42]. Hot air drying also exhibited elevated total phenolic content and antioxidant activity compared to freeze drying in olive leaves as shown by Ahmad-Qasem et al. [43]. Hot air dried leaf tissue extracts tend to exhibit higher phenolic compounds content, as found in this study, but the link with antioxidant capacity is not clear. In our study we found FRAP having significant correlation with total glucosinolate content as well as total phenolic content, but no significant correlation with DPPH scavenging activity. Interestingly, we found at the same time the highest DPPH radical scavenging activity and lowest FRAP values in freeze dried tissues followed by hot extraction, while exactly the opposite was determined in hot air dried samples followed by cold extraction. The obtained results are contrary to previously published [42,44,45] data where FRAP and DPPH radical scavenging assays have high correlations, especially since it is known that both depend mainly on the electron transfer mechanism which measures the antioxidant's reducing ability [46,47]. FRAP values correlated highly with total phenolic content and negatively with sucrose and fructose content while the data obtained by the DPPH radical scavenging assay correlated positively with the samples' sugar content possibly indicating interferences with the meta-products of Maillard reactions occurring during the hot air drying process [48].

Hot air drying *B. oleracea* var. *acephala* leaf tissues followed by cold extraction induced significantly lower content of all investigated sugars compared to freeze drying followed by cold extraction. If freeze dried tissues were subjected to hot extraction sucrose and fructose levels were significantly lower compared to the cold extraction process but higher than oven dried leaf tissues except for fructose. Fante and Zapata Norena studied the quality of hot air dried and freeze dried garlic and found that inulin content decreased with hot air drying while glucose and fructose content increased compared to freeze dried garlic [49]. Iombor et al. studied changes in soursop flour composition as affected by oven and freeze drying and found 40% decrease in carbohydrate content in hot air dried samples [50]. Zhang et al. studied the changes in chestnut starch properties during different drying methods and found lower starch but higher reducing sugar content in oven dried compared to freeze dried samples [51]. Karaman et al. found lower fructose and glucose content in oven dried compared to freeze dried persimmon powders [52]. On the other hand, Gao et al. showed that there was no difference in sucrose, glucose and fructose content in oven and freeze dried jujube samples [53]. The observed decrease in sugar content in our study could be attributed to Maillard reactions occurring at elevated temperatures during hot air drying [54] or possibly the caramelization of sugars [55]. Michalska et al. determined that increased drying temperature of different plum cultivars resulted in decreased total sugar content while early and intermediate Maillard reaction products increased in content [48]. Similarly, Li et al. observed a decrease in reducing sugar content during hot air drying opposed to freeze drying of instant *Tremella fuciformis,* which the authors attributed to an intense Maillard reaction occurring between the carbonyl group of the reducing sugars and amino acids at elevated temperatures [56]. Previously published data on oligosaccharide and simple sugars extraction in various solvent mixtures and extraction temperatures showed that, with increased temperature, especially at the solvent boiling point, sugar content yields were higher [57]. We, on the other hand, observed lower yields of sucrose and fructose in freeze dried leaves hot extracts, compared to the cold extraction step while glucose levels remained unaffected.

#### **4. Materials and Methods**

#### *4.1. Plant Material*

Thirty-six *B. oleracea* var *acephala* ecotypes (accessions IPT379, IPT390, IPT386, IPT391, IPT381, IPT392, IPT393, IPT419, IPT394, IPT395, IPT383, IPT396, IPT384, IPT385, IPT397, IPT398, IPT420,

IPT399, IPT400, IPT387, IPT401, IPT 402, IPT403, IPT404, IPT405, IPT406, IPT174, IPT202, IPT206, IPT407, IPT421, IPT408, IPT409, IPT410, IPT380, IPT411) used in this study were grown under the same agro-climatic conditions on the experimental farm of the Institute of Agriculture and Tourism, Poreˇc, Croatia (N 45 ◦13 ′20.30", E 13 ◦36 ′6.49") and are part of the National Program for the Conservation and Sustainable Use of Plant Genetic Resources for Food and Agriculture (accession IPT399 and IPT403 are shown in Figure 4). The leaves used in this study were harvested when the plants reached technological maturity. Harvested leaves were fully developed without any signs of physiological, pest or diseases injury. Fresh plant samples (three biological replicates per sample), immediately after harvesting, were either kept at −80 ◦C until the freeze drying process or placed in an oven (Memmert UF160, Schwabach, Germany) at 105 ◦C overnight. Frozen plant samples were placed in a freeze dryer (Labogene Coolsafe 95-15 Pro, Allerød, Denmark) and lyophilized over a period of 48 h. Lyophilized or oven-dried samples were ground to powder (0.2 mm) using an ultra-centrifugal mill (Retsch ZM200, Haan, Germany). Poreč, Croatia ′ ″ ′ ″ −8

**Figure 4.** *Brassica oleracea* var. *acephala* accession (**a**) IPT399 and (**b**) IPT403.

### *4.2. Hot Methanol Extraction*

G for 5 min (Domel Centric 350, Železniki, Slovenia) and the − Freeze dried plant material (30 mg) was preheated to 75 ◦C for 3 min in a heating/cooling dry block (Biosan CH100, Riga, Latvia) and 1.5 mL of preheated 70: 30 methanol:water (*v*/*v*) at 75 ◦C was added. The samples were incubated for 10 min at 75 ◦C and manually shaken every 2 min. Afterwards the samples were centrifuged at 15,000 G for 5 min (Domel Centric 350, Železniki, Slovenia) and the supernatant was filtered through a 0.22 µm nylon filter and transferred to a clean tube. The samples were stored at −80 ◦C until further analysis.

### *4.3. Cold Methanol Extraction*

Freeze dried or oven dried plant material (30 mg) was extracted with 1.5 mL of 80:20 methanol: water (*v*/*v*) at 20 ◦C over a period of 30 min in an ultrasonic bath (MRC 250H, Holon, Israel). The samples were centrifuged at 15,000 G for 5 min (Domel Centric 350, Železniki, Slovenia) and the supernatant was transferred to a clean tube. The samples were stored at −80 ◦C until further analysis.

#### *4.4. Determination of Total Antioxidant Activity*

Total antioxidant activity was evaluated using the FRAP assay [58] ant the DPPH radical scavenging activity assay [59]. Briefly, 100 µL of the sample was mixed with 200 µL of either freshly prepared FRAP reagent or 0.02M DPPH radical for the FRAP or DPPH assays, respectively. The antioxidant activity using the FRAP assay was evaluated after 10 min of reaction time at 25 ◦C by reading the absorbance at 593 nm while the DPPH radical scavenging ability was evaluated after 30 min of reaction time at 25 ◦C by reading the absorbance at 517 nm (Tecan Infinite 200 Pro M Nano+, Männedorf, Switzerland). FRAP values were calculated against a Fe2<sup>+</sup> calibration curve (y <sup>=</sup> 0.0168x <sup>−</sup> 0.002; serial dilutions of Fe2+—20, 40, 80, 120, 160, 200, 250 µM; coefficient of determination, R<sup>2</sup> = 0.9999, recovery: 101.8 <sup>±</sup> 1.6 %) and expressed as <sup>µ</sup>mol Fe2+/g DW. DPPH radical scavenging ability values were calculated against a standard curve of Trolox (y = −0.0128x + 0.0125; serial dilutions of Trolox—2, 5, 10, 25, 50, 75, 100 <sup>µ</sup>M; coefficient of determination, R<sup>2</sup> <sup>=</sup> 0.9995, recovery: 103.7 <sup>±</sup> 1.2 %) and expressed as µmol TEQ/g DW, respectively.

#### *4.5. Determination of Total Glucosinolates and Total Phenolic Content*

Total glucosinolates were determined according to Ishida et al. [60] with some modifications. Briefly, 10 µL of plant extract was mixed with 300 µL of 2 mM Palladium (II) chloride and after 30 min of reaction time at 25 ◦C absorbance was read at 425 nm (Tecan Infinite 200 Pro M Nano+, Männedorf, Switzerland). The results were calculated against a standard curve of sinigrin (y = 3.8021x + 0.1682; serial dilutions of sinigrin—0.3, 0.6, 1.2, 1.8, 2.4, 3.0 mg/L; coefficient of determination, R<sup>2</sup> = 0.9995, recovery: 99.7 ± 2.9%) and expressed as SEQ/g DW. Total phenolic content was determined according to Singleton and Rossi [61] with some modifications. The methanolic extracts (20 µL) were mixed with 140 µL of freshly prepared 0.2M Folin–Ciocalteu reagent. After 1 min, 140 µL of 6% solution of Calcium carbonate was added to the mixture. The absorbance was read at 750 nm (Tecan Infinite 200 Pro M Nano+, Männedorf, Switzerland) after 60 min of reaction time at 25 ◦C. The results were calculated against a standard curve of gallic acid (y = 3.7867x − 0.2144; serial dilutions of gallic acid—12.5, 25, 50, 75, 100, 150, 250 mg/L; coefficient of determination, R<sup>2</sup> <sup>=</sup> 0.9999, recovery: 102.0 <sup>±</sup> 2.9 %) and expressed as mg GAEQ/g DW.

#### *4.6. Sugar Analysis by HPLC*

The analysis of sucrose, fructose and sucrose content was carried out using a HPLC system consisting of a solvent delivery unit (Varian 210, Palo Alto, CA, USA), an autosampler (Varian 410, Palo Alto, CA, USA), column oven (Varian CM500, Palo Alto, CA, USA) and a refractive index detector (Varian 350, Palo Alto, CA, USA). Chromatographic separation was achieved by injecting 10 µL of the sample on a 300 × 8 mm, 9 µm particle size, calcium cation exchange column (Dr. Maisch ReproGel Ca, Ammerbuch, Germany) held at 80 ◦C using deionized water as the mobile phase (1 mL/min, isocratic elution). Retention times and peak areas of the investigated sugars were compared to analytical standards for identification and quantification, respectively. Linear calibration curves were obtained with serial dilutions of 0.25, 0.50, 1.00, 2.50, 5.00, 7.50, 10.00 g/L of sucrose (y = 2265.73x + 29.97, coefficient of determination, R<sup>2</sup> <sup>=</sup> 0.9998, recovery: 99.9 <sup>±</sup> 2.3 %), glucose (y <sup>=</sup> 2224.75x <sup>+</sup> 28.33, coefficient of determination, R<sup>2</sup> <sup>=</sup> 0.9999, recovery: 99.8 <sup>±</sup> 1.8 %) and fructose (y <sup>=</sup> 2233.53x <sup>+</sup> 47.37, coefficient of determination, R<sup>2</sup> <sup>=</sup> 0.9998, recovery: 100.0 <sup>±</sup> 0.6 %).

#### *4.7. Statistical Analysis*

To determine the effect of the investigated extraction protocols on total antioxidant activity, total glucosinolate, total phenolic and sugar content in *B. oleracea* var. *acephala* leaves the results were processed by analysis of variance (ANOVA) and reported as mean ± SE. Pearson's correlations were calculated to evaluate the connection between bioactive compounds content, antioxidant activity, and sugar content. Further investigation on the impact of different extraction protocols on the

studied compounds was carried out by employing PLS-DA as a supervised multivariate method. All statistical analyses were performed using Statistica 13.4.0.14. (Tibco Inc., Palo Alto, CA, USA). Significant differences were determined at *p* ≤ 0.05 and homogenous group means were compared by Tukey–Kramer Unequal N HSD test.

#### **5. Conclusions**

The results from our study confirm the results published by Doheny-Adams et al. [12] that cold (ambient temperature) aqueous methanol can be used instead of boiling aqueous methanol with no adverse effects on total glucosinolate content. In addition to higher total glucosinolate content we observed an increase in antioxidant activity measured by FRAP, total phenolic content, sucrose content, and fructose content when the cold extraction step was applied on freeze dried *B. oleracea* var. *acephala* plant tissues while glucose levels remained unaffected. Our results also show that hot air drying, compared to freeze drying, followed by cold extraction has an adverse effect on antioxidant activity measured by DPPH radical scavenging, total glucosinolate content, as well as, on the content of all studied sugars in *B. oleracea* var. *acephala* leaves, indicating it would be inferior to freeze drying. On the other hand, if phenolics are the compounds of interest in *B. oleracea* var. *acephala* leaves, hot air drying may be a viable alternative to freeze drying.

**Author Contributions:** Conceptualization, N.M. and S.G.B.; methodology, N.M., B.P., S.G.B., Z.U., D.B.; investigation, N.M., B.P., Z.U.; data curation, N.M.; writing—original draft preparation, N.M. and J.P.; writing—review and editing, S.G.B. and D.B.; visualization, N.M.; funding acquisition, S.G.B. and D.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the project KK.01.1.1.01.0005 Biodiversity and Molecular Plant Breeding, Centre of Excellence for Biodiversity and Molecular Plant Breeding (CoE CroP-BioDiv), Zagreb, Croatia.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Optimization of a Green Ultrasound-Assisted Extraction of Di**ff**erent Polyphenols from** *Pistacia lentiscus* **L. Leaves Using a Response Surface Methodology**

#### **Cassandra Detti <sup>1</sup> , Luana Beatriz dos Santos Nascimento 1,\*, Cecilia Brunetti 1,2 , Francesco Ferrini 1,2 and Antonella Gori 1,2,\***


Received: 28 September 2020; Accepted: 30 October 2020; Published: 3 November 2020

**Abstract:** *Pistacia lentiscus* leaves are used in several applications, thanks to their polyphenolic abundance. Thiswork aimed to characterize the polyphenols and to optimize the extraction conditions to shorten the time, decrease the consumption of solvent, and to maximize the yield of different classes of phenolics, which have diverse industrial applications. The variables were optimized by applying a Box–Behnken design. Galloyl and myricetin derivatives were the most abundant compounds, and two new tetragalloyl derivatives were identified by LC-MS/MS. According to the models, the maximum yields of polyphenols (51.3 <sup>±</sup> 1.8 mg g−<sup>1</sup> DW) and tannins (40.2 <sup>±</sup> 1.4 mg g−<sup>1</sup> DW) were obtained using 0.12 L g−<sup>1</sup> of 40% ethanol at 50 ◦C. The highest content of flavonoids (10.2 <sup>±</sup> 0.8 mg g−<sup>1</sup> DW) was obtained using 0.13 L g−<sup>1</sup> of 50% ethanol at 50 ◦C, while 0.1 L g−<sup>1</sup> of 30% ethanol at 30 ◦C resulted in higher amounts of myricitrin (2.6 <sup>±</sup> 0.19 mg g−<sup>1</sup> DW). Our optimized extraction decreased the ethanolic fraction by 25% and halved the time compared to other methods. These conditions can be applied differently to obtain *P. lentiscus* extracts richer in tannins or flavonoids, which might be employed for various purposes.

**Keywords:** Anacardiaceae; design of experiments (DOEs); flavonoids; green extraction; HPLC-DAD; LC-MS/MS; tannins; ultrasound assisted-extraction (UAE)

#### **1. Introduction**

*Pistacia lentiscus* L. (Anacardiaceae), known as mastic orlentisk, is an evergreen shrub, widespread over many areas in the Mediterranean basin [1]. This species is largely distributed in dry ecosystems, characterized by nutrient and water scarcity due to the long periods of drought, high irradiation, and temperatures [2,3].

Several studies have demonstrated that *P. lentiscus* leaves are rich in polyphenolic compounds [4,5] including gallotannins and flavonoids (mainly quercetin and myricetin derivatives) [6–9]. These two main classes of compounds have different industrial and commercial applications. Flavonoids are intensively used in the food industry as preservatives and flavoring agents [10], the cosmetic industry as skin protectors [11], and in agricultureas an anti-infective agents [12]. Tannins, otherwise, are widely applied in the leather industry, as well as beverages additives, corrosion inhibitors of metals in shipbuilding, wood adhesives, and foams [13].

From a pharmacological point of view, both classes of phenolics have long been suggested to have high antioxidant capacities and other several biological activities [14]. *Pistacia lentiscus* leaves have traditionally been used in folk medicine for the treatment of various diseases such as hypertension, stomach aches, and kidney stones [15–18]. Moreover, anti-ulcer, anti-inflammatory, cytoprotective, acetylcholinesterase inhibition, and anticancer activities have been already described for its leaf extracts [15,19–21].

Therefore, leaves of *P. lentiscus* represent a reliable source of polyphenols to be exploited by several industries [22]. Thus, obtaining extracts enriched in different classes of these compounds is of high interest.

The quality and the content of polyphenols in plant leaf extracts depend on several factors such as the harvest moment and seasonality, the plant phenological stage, the leaf age, and the applied extraction process [6,23–25]. Well-established conventional extraction methodologies have been associated with significant economic and environmental impacts such as high solvent consumption and prolonged extraction times [23,26]. Nowadays, with the development of the concept of "green extraction", environmentally friendly techniques should be developed, avoiding hazardous reagents and optimizing extraction parameters such as time, temperature, and solvent type [27,28]. These green techniques, include ultrasound-assisted extraction (UAE), enabling the maximum yield of active compounds with low energy and less time consumed [29–31]. Ultrasound-assisted extraction represents one of the best and cheapest technologies with limited instrumental requirements [32,33], being efficiently used for extracting phenolic compounds from several plant materials [23,34,35].

The increasing interest inthe improvement of extraction processes from plants has triggered the application of mathematical models for the optimization of extraction conditions. In this sense, response surface methodology (RSM), widely applied for industrial purposes, has become the most preferable approach for optimizing extraction procedures that apply multiple variables at the same time [36,37]. In this sense, the Box–Behnken design (BBD) is one of the most used RSMs. This design requires a small number of runs and, therefore, avoids time-consuming experiments and has been largely applied for optimizing extractions of single or classes of molecules from different plant materials [38–40].

Response surface methodologieshave already been applied for optimizing the ultrasound-assisted extraction of polyphenols from *P. lentiscus* leaves [41]. In this study, the authors tested the UAE by using a central composite design with high solvent volume and leaf material [41]. In addition, the authors quantified the total phenolic content using the Folin–Ciocalteu reagent, not considered specific for phenolics, since this reagent can be reduced by other compounds that might cause interferences in the results [42]. As such, a detailed optimization of different classes of polyphenols in *P. lentiscus* leaf extracts has not yet been conducted.

In this context, this study aimed to:


#### **2. Results and Discussions**

#### *2.1. Screening Design and Determination of the Important Factors*

Several factors can influence the efficiency of an extraction such as solvent type, time, particle size, and temperature [43]. Consequently, it is important to verify how different variables affect the extraction of target compounds [23,44]. The UAE is one of the most appropriate extraction processes due to the fact of its efficacy, cleanliness, facility of use, and speed [45,46]. Among the different factors that should be considered in ultrasound-assisted extraction, the polarity of the solvent and the solvent ratio is very important [45,47]. Moreover, time and temperature also affect the yield of the compounds and the costs of the whole process. Indeed, it is desirable to develop methods that lead to a higher extraction of target compounds using lower temperatures, shorter time, and lower concentration of organic solvent than possible [43,47].

The calculated coefficients for the different answers (i.e., total tannins content (TTC), total flavonoids content (TFC), and total polyphenolic content (TPC)) are shown in Table 1. The solvent ratio (x3) showed to be the most important factor affecting positively all the responses as can be inferred by the positive and significant value of the b<sup>3</sup> coefficient. As such, an increase in the solvent ratio from 0.06 to 0.1 L g−<sup>1</sup> led to higher amounts of polyphenols. Therefore, even higher solvent ratios (solvent volumes) were chosen for the further optimization steps.


**Table 1.** Screening fractional factorial design matrix (FFD-24−<sup>1</sup> ) of trials conducted (trials 1 to 9, this last the central point), with the independent variables (x<sup>1</sup> to x<sup>4</sup> ) and answers (y).

The trials were conducted in triplicate. The coefficients (b<sup>1</sup> to b4), correspondent to each variable (x<sup>1</sup> to x4), were calculated by regression. The asterisks (\*) indicate significant coefficients (*p* ≤ 0.05).Total tannins content (TTC); total flavonoids content (TFC); total polyphenolic content (TPC).

The fraction of ethanol (%) (x4) showed to be a determinant for the extraction of tannins (TTC) and total polyphenols (TPC), negatively affecting both. The b<sup>4</sup> coefficient was significant and negative, indicating that the extractions conducted using a smallerpercentage of ethanol resulted in higher yields of these compounds. This phenomenon could be explained by the fact that with an increase in ethanol concentration, the solvent polarity may decrease as well as the molecular movements, reducing the solubility of the polar compounds [48]. In addition, by raising the surface tension of the solvent, an increase in the molecular interactions is induced, consequently raising the extraction [32], while the addition of water to the organic solvent may help break the hydrogen bonding and facilitate the extraction of polyphenols [49]. Based on these results, lower percentages of ethanol were studied in the optimization step in order to conduct a greener extraction [32,50]. In fact, ethanol and water are solvents widely used by food and pharmaceutical industries due to the fact of their safer handling [51].

For the temperature (x1), the calculated coefficients (b1) were also high for TTC and TPC. For both, this factor showed a positive effect. According to this, higher temperatures (30, 40, and 50 ◦C) were evaluated during the optimization, which should also be considered when higher amounts of water are applied. The use of higher temperatures during the UAE can increase the efficiency of the extraction process due to the increase in the number of cavitation bubbles formed [45,52]. Moreover, temperature influences the mass transfer process by improving the solvent penetration in plant cells due to the reduction in its viscosity. In addition, higher temperatures increase the degradation of the plant matrix, and weaken the interactions of the polyphenols with other cell constituents, making their extraction easier [49,53].

The coefficient b<sup>2</sup> showed to be low for all the answers, indicating that the variable x<sup>2</sup> (time) has no effect on the answers. Therefore, this factor was kept constant in the optimization step, and the shorter extraction time tested (15 min) was chosen. For further industrial purposes, this is desirable, since less time can reduce the energy consumed [54].

#### *2.2. Optimization Design: Models and Response Surfaces Analysis*

For an efficient extraction process, not only the method used (e.g., UAE, microwave assisted or conventional extractions), but also the variables applied are of a great importance as well as their linear, quadratic, and interactive effects. The multi-factorial study of them, such as applying experimental designs and RSM, allows the maximization of responses with minimal energy loss and solvent consumption [47].

The results of tannins (TTC), flavonoids (TFC), myricitrin (MYC), and total polyphenols (TPC) (calculated as the sum of individual phenolics, Supplementary Materials Tables S1 and S2) obtained for the experimental trials conducted are presented in Table 2 and were used to obtain regression equations (models, Supplementary Materials Table S3). Trials 9 to 11 were shownto have the best conditions for achieving higher amounts of TPC and TTC (Table 2), all of them, interestingly using 40% ethanol as solvent. For the flavonoids content (TFC and MYC) instead, the best trials were from trials 5 to 8, all using 0.15 L g−<sup>1</sup> of solvent ratio (Table 2). It is interesting to note that for all the responses, an extraction conducted using a solvent ratio of 0.2 L g−<sup>1</sup> at 40 ◦C resulted in the lowest amounts of polyphenols (Table 2).


**Table 2.** Box–Behnken design (BBD) matrix with natural and coded values for the independent variables and responses. Fifteen experimental trials were conducted with triplicates of the central point (trial 13).

For the central point (trial 13), values of the mean ± SD are presented. All the values of the responses are expressed in mg g−<sup>1</sup> DW. Total tannins content (TTC); total flavonoids content (TFC); myricitrin content (MYC), total polyphenolic content (TPC).

According to the results, second-order polynomial regression models based on the coded coefficient values were obtained for each response (Supplementary Materials Table S3). To verify the fitting of the mathematical models, the data were statistically analyzed. The quadratic model applied is usually assumed to fit the data sufficiently well to indicate the more suitable and the better regions of work. Statistically, the quality of a model is evaluated by the significance of the regression according to the ANOVA test; the lack of fit (LOF), used to measure the adequacy of these models [55]; the multiple determination coefficient (R<sup>2</sup> ), which represents the variation of the response explained by the model; and the adjusted multiple determination coefficient (R<sup>2</sup> adj), which indicates the capacity of the model to be predictive [55–57]. The analysis of the models obtained for the different responses (TTC, TFC, MYC, and TPC) are presented in Table 3.


**Table 3.** Statistical parameters after data analysis and fit of the models obtained for the different responses.

\* Significant values (*p* ≤ 0.05).

For a good fit, R<sup>2</sup> should be at least 80% [58]. In all our analysis, the R<sup>2</sup> values were higher than 0.89 (89%, Table 3), suggesting that the models described well the behavior of these responses. Moreover, for all of them, the R<sup>2</sup> adj were higher than 0.71, indicating a good predictive power, since in a good statistical model R<sup>2</sup> adj should be comparable and similar to R<sup>2</sup> , with differences less than 0.2–0.3. Furthermore, the values for the LOF were not significant to an extent with the pure error (*p*>0.05, for all). A model will fit the experimental data when a significant regression and a non-significant LOF are found [59]. Therefore, considering these results, as well as the *p*-value (all *p*≤0.05) (Table 3), the models showed to be suitable and appropriate to well describe the relationship betweenthe responses (TTC, TFC, MYC, and TPC) and the independent variables (x<sup>1</sup> to x3).

The significance of the coefficients was also determined (Supplementary Materials Table S3). For the response of total polyphenols (TPC) and total tannin (TTC) contents, the coefficient b<sup>2</sup> (x2–solvent ratio) showed to be the highest, compared with the coefficients b<sup>1</sup> (x1–ethanol fraction, %) and b<sup>3</sup> (x3–temperature). Besides, b<sup>2</sup> was negative, suggesting that the use of less solvent is better for the extraction of TPC and TTC. Similarly, for the total content of flavonoids (TFC), the coefficient b<sup>2</sup> showed higher values compared to b<sup>1</sup> and b3, also negatively affecting the response. It means that less solvent should be better for the extraction of flavonoids (Supplementary Materials Table S3). The analysis of variance (ANOVA) for the polynomial models indicated that the quadratic terms of the variable x<sup>1</sup> (ethanol%, b11) and x<sup>3</sup> (temperature, b33) were the most important, significantly influencing the responses (*p* < 0.05) (Supplementary Materials Table S3). In fact, temperature and type of solvent are important factors to be considered in UAE [32].

To provide a better visualization of the effects of the factors in the responses, contour response surface plots were generated from the models (Figure 1—TTC and TPC and Figure 2—MYC and TFC), by plotting the responses with regard to ethanol concentration (x1) and solvent ratio (x2) at each temperature 30, 40, and 50 ◦C (x3). These response surfaces can be used for the prediction of the responses (polyphenols contents) in the investigated experimental domain.

For the total polyphenolic content (TPC, Figure 1a–c), 30 ◦C and 50 ◦C predicted maximum amounts (>50.0 mg g−<sup>1</sup> DW, Figure 1a,c), with 50 ◦C being better, since a more extended and stable optimal region of extraction wasobtained (Figure 1c). At this temperature, 35% to 45% of ethanol in a solvent ratio of 0.1 to 0.15 L g−<sup>1</sup> should be used (Figure 1c). At 40 ◦C, a good region was also found, however, resulting in lower amounts (~44.0 mg g−<sup>1</sup> DW). In fact, the optimal conditions proposed by the model were 40% ethanol in a ratio of 0.12 L g−<sup>1</sup> at 50 ◦C, resulting in 51.3 <sup>±</sup> 1.8 mg g−<sup>1</sup> DW of TPC.

A similar percentage of ethanol was also proposed by a previous study focused on the optimization of the phenolic extraction of *P. lentiscus* leaves using a microwave-assisted method [22]. The authors showed that percentages of ethanol around 30% to 40% significantly raised the total phenolic content, spectrophotometrically quantified by the Folin–Ciocalteu reagent [22]. In addition, a similar effect of ethanol percentage was also reported for the extraction of phenolic compounds from other plant sources such as green tea [60]. Considering temperatures analogous to our findings, 45–50 ◦C was shown to maximize the extraction of polyphenols in *Pistacia atlantica* leaves [36].

−

− **Figure 1.** Response surface (contour plots) for predicting the TPC (**a**–**c**) and TTC (**d**–**f**) in *Pistacia lentiscus* leaf extracts with regard to the ethanol fraction (%, x<sup>1</sup> ) and solvent ratio (L g−<sup>1</sup> , x<sup>2</sup> ), at each temperature (x<sup>3</sup> , 30, 40, and 50 ◦C). The regions with the darkest-gray color represent the domains of working conditions assuring the maximum values for the evaluated compounds (total polyphenols and tannins).

The total tannin content (TTC, Figure 1d–f) showed very similar response surfaces and optimal conditions to TPC. Indeed, *P. lentiscus* leaves are rich in tannins, which represent around 70% of the TPC [6]. Therefore, a similar behavior should be expected. For TTC, temperatures of 30 ◦C and 50 ◦C should be also used to reach higher amounts of tannins (~40.0 mg g−<sup>1</sup> DW). However, different from total polyphenols, a narrow optimal region was observed at 50 ◦C (Figure 1f). At any temperature, an extraction using around 0.13 L g−<sup>1</sup> of 40% ethanol provided better results (Figure 1d–f). Indeed, the optimal conditions for maximization of the content of tannins were the same for TPC (0.12 L g−<sup>1</sup> , 40% ethanol, 50 ◦C), yielding 40.2 <sup>±</sup> 1.4 mg g−<sup>1</sup> DW.

− − − − For TFC and MYC (Figure 2), at any temperature, a decrease in the content of these compounds was observed around the medium percentages of ethanol with the optimal regions being obtained when extreme values of ethanol fractions (30% or 50%) and temperatures (30 or 50 ◦C) are chosen (Figure 2). Therefore, higher contents of flavonoids (~10 mg g−<sup>1</sup> DW, Figure 2a,c) and myricitrin (>2.5 mg g−<sup>1</sup> DW, Figure 2d,f) are predicted under these conditions. As such, extractions conducted at 50 ◦C, using 50% ethanol in 0.1 to 0.17 L g−<sup>1</sup> result in greater amounts of flavonoids (Figure 2c). Decreasing the temperature to 40 ◦C, lesser amounts of flavonoids are obtained (~8.5 mg g−<sup>1</sup> DW) (Figure 2b). The increase in the temperature can cause higher solubility and diffusion coefficients of polyphenols, such as flavonoids, which result in a higher extraction rate [61]. The optimal conditions predicted by the model for maximization of the flavonoid content in *P. lentiscus* leaf extracts are 50% ethanol in 0.13 L g−<sup>1</sup> at 50 ◦C, resulting in 10.2 <sup>±</sup> 0.8 mg g−<sup>1</sup> DW.

−

−

−

−

− −

−

**Figure 2.** Response surface (contour plots) for predicting the TFC (**a**–**c**) and MYC (**d**–**f**) in *P. lentiscus* leaf extracts with regard to the ethanol fraction (%, x<sup>1</sup> ) and solvent ratio (x<sup>2</sup> ) used at each temperature (x<sup>3</sup> , 30, 40, and 50 ◦C). The regions with the darkest-gray or black color represent the domains of working conditions assuring the maximum values for the evaluated compounds (total flavonoids and myricitrin).

− − − − Myricitrin content (Figure 2d–f) showed similar response surfaces to the TFC (Figure 2a–c). The extraction conducted at 30 ◦C with 30% ethanol in 0.1 L g−<sup>1</sup> should result in maximal contents of this flavonoid (2.6 <sup>±</sup> 0.19 mg g−<sup>1</sup> DW). However, at 50 ◦C very similar amounts were also obtained, but 50% ethanol in slightly higher volumes should be used (Figure 2f). As can be noticed, the TFC and MYC have similar behaviors. This could be explained because myricitrin is the most abundant flavonoid detected in *P. lentiscus* leaves (Figure 3), also justifying our choice in maximizing its content. This compound has also been described as a major compound in lentisk leaf extracts in previous studies [6,7,62].

To validate the adequacy of the mathematical models, verification experiments were carried out in triplicate under the optimal conditions. Mean values of 39.8 <sup>±</sup> 4.1 mg g−<sup>1</sup> DW for TTC, 50.9 <sup>±</sup> 4.9 mg g−<sup>1</sup> DW for TPC, and 9.9 <sup>±</sup> 1.4 mg g−<sup>1</sup> DW for TFC were obtained from the real experiments and demonstrated the validation of the models for these three responses (*p*TTC = 0.88; *p*TPC = 0.90; *p*TFC = 0.76).

We observed that among the variables tested, the same solvent ratio (~0.13 L g−<sup>1</sup> ) and temperature (50 ◦C) should be used during the UAE process to obtain the maximal yields of tannins and flavonoids. However, the ethanol percentage showed to differ between both classes of compounds. While for tannins (TTC) 40% ethanol should be used, 50% is preferable for the extraction of flavonoids (TFC). This difference can be explained by the distinct solubility of these compounds [47]. The polarity of the ethanol–water mixture decreases with the addition of ethanol, stimulating the extraction of less polar compounds from plant cells. Flavonols, such as myricetin derivatives, show higher solubility with increasing concentration of alcohol, consequently reaching greater extraction yields when less polar solvents are used [23]. Tannins with low molecular weight (galloyl derivatives) occurring in *P. lentiscus* leaf extracts (Figure 3, Table 4) are more polar than the flavonoids detected. Therefore, it is reasonable that the extraction of these types of tannins (and consequently the overall polyphenolic content) is stimulated by the utilization of more polar solvents (i.e., 40% ethanol). Indeed, higher concentrations

of ethanol and methanol are more beneficial for the extraction of flavonoids than for tannins, which generally need higher amounts of water [47]. In agreement with our results, Barbouchi et al. [63] obtained higher phenolic contents in *P. lentiscus* leaf extracts when more polar extraction solvents were used. In addition, ethanol was considered the most suitable solvent for the recovery of flavonoids from this species [2].

The extraction conditions optimized here are suitable for experimental and further industrial applications, since they apply a green solvent (ethanol:water) in low quantity (~0.13 L g−<sup>1</sup> ) for a short time (15 min), using moderate temperatures (50 ◦C). Considering that more toxic solvents, such as methanol, chloroform, and ethyl acetate, are intensively used to extract *P. lentiscus* leaves [19,64–66] and that the typical extraction methods apply higher percentages of ethanol for longer times [2,6,7], our optimization led to a greener extraction procedure if compared to conventional extraction methods, by decreasing the ethanolic fraction by at least 25% and halved the time used. −

A green extraction is defined as a procedure able to reduce energy consumption and use of organic solvents, saving the quality of the process [29]. In particular, three major points should be considered: the improvement and the optimization of the existing methods; the use of simple equipment; and the innovation in the use of alternative solvents [67]. In this sense, the optimization of a standard extraction procedure, especially employing the minimum amount of organic solvent, could be considered green, even if moderate temperatures are applied.

#### *2.3. Polyphenolic Composition of the Richest P. lentiscus Extract*

Figure 3 shows the polyphenolic profile of *P. lentiscus* leaf extract obtained using the conditions of trial 9 (BBD, Table 2), corresponding to the extract with the highest content of total polyphenols (TPC). The LC-MS/MS analysis was performed to provide a more comprehensive characterization of the polyphenols present in the leaves of the species as well as to confirm previous characterizations reported in the literature [7,9,62,64].

**Figure 3.** Chromatograms of *P. lentiscus* leaf extracts obtained using the extraction conditions of trial 9 (BBD, Table 2) acquired at 280 nm (above) and 350 nm (below).

The UV-Vis and MS/MS spectra allowed us to identify 19 compounds (Table 4), classified into three main classes: gallic acid derivatives (peaks 1 and 2), gallotannins (peaks 3–9), and flavonoids (peaks 10–19). For flavonoids, three peaks were identified as myricetin derivatives (peaks 10, 11, 14), six as quercetin derivatives (12, 13, 15–18), and one as a kaempferol derivative (19).

Among the four main peaks detected, three of them (peaks 3, 4, and 7) showed the fragmentation correspondent to mono-, di-, and trigalloylquinic acids, respectively (Table 4). These metabolites have already been described in the literature using different kinds of detectors such as triple quadrupole (QQQ) [7,9,64] and quadrupole time-of-flight (Q-TOF) mass spectrometers [62]. Indeed, according to these previous reports, the monogalloylquinic acid (peak 3) was defined by the fragments *m*/*z* 343 [M − H]<sup>−</sup> and 191; this lastresulted from the loss of the galloyl moiety [M152-H]. In addition, the digalloylquinic acid and its isomer (peaks 4 and 5) were characterized by the fragments *m*/*z* 495 [M − H]−, 343, 191, and 169, that are consistent with the successive loss of two galloyl units, and correspondent to the gallic acid itself (*m*/*z* 169). Finally, the trigalloylquinic acid (peak 7) and its isomer (peak 6) showed the fragments m/z 647 [M − H]−, 495, 343, 191, and 169, consistent with a trigalloyl substitution. Two minor peaks with the UV spectra and the mass fragmentation typical of quinic acid derivatives were also detected, with precursor ions of *m*/*z* 799 and fragments 495, 343, 191, and 169, correspondent to four consecutive losses of galloyl moieties (Supplementary Materials Figure S1). As such, these peaks (peaks 8 and 9) were tentatively identified as tetragalloylquinic acid (and its isomer), here, firstly reported in *P. lentiscus* leaf extracts.


**Table 4.** LC–DAD-MS/MS characterization of the main polyphenols present in extracts of *P. lentiscus* leaves. Compounds numbers correspond to those indicated in Figure 3 (sh, shoulder).

Among the gallic acid derivatives (peaks 1 and 2), the peak 1 was assigned as monogalloyl glucose (glucogallin), based on the literature [7] and according to its ion fragments *m*/*z* 331 [M − H]−, 169 (resulted from the loss of the glucose, *m*/*z* 162) and 125 (derived from the decarboxylation of galloyl). The peak 2, instead, was identified as gallic acid, based on its characteristic mass spectra, with the precursor ion at *m*/*z* 169 [M − H]<sup>−</sup> and the fragment *m*/*z* 125 (decarboxylation of the galloyl). The identification of this compound was also confirmed based on the comparison with the specific external standard.

Flavonoids (from peak 12 to 18) were identified based on the mass fragment of their corresponding aglycon units, namely, myricetin (*m*/*z* 317), quercetin (*m*/*z* 301), and kaempferol (*m*/*z* 285). This was confirmed by the injection of the external standards myricitrin, rutin, and kampferol-3-*O*-rutinoside. The sugar moieties were characterized based on the neutral losses of 132 (presence of pentosides: xylose or arabinose), 162 (hexosides: galactose or glucose), and 146 (deoxyhexoside: rhamnose). Thus, in agreement with the fragmentation patterns described in the literature [62,64], the following flavonoids were tentatively identified as: myricetin-3-*O*-galactoside (peak 10), myricetin-3-*O*-rutinoside (peak 11), quercetin-*O*-hexosides 1 and 2 (peaks 13 and 15), quercetin-*O*-galloyl-pentoside (peak 16), quercetin-3-*O*-arabinoside (peak 17), quercetin-3-*O*-rhaminoside (peak 18), and kaempferol-*O*-hexoside (peak 19). The identification of the major flavonoidic peak (14), myricetin-3-*O*-rhamnoside (myricitrin), was obtained by comparison with the specific analytical standard. The remaining peak (12) was tentatively identified as a quercetin derivative based on the UV–Vis spectra, in the absence of conclusive mass-spectrometric data and reference in the literature.

High contents of gallotannins (galloylquinic acid derivatives) and myricetin derivatives were previously described in *P. letiscus* leaves [6,7,9,17,68]. These compounds represent approximately 90% of the polyphenolic composition of the leaf extracts [6] and are possibly the main responsible for their biological properties such as: anti-inflammatory, cytoprotective, hepatoprotective, enzymatic-inhibitory, antitumor, and anti-diabetes [7,62,69–71]. It is noteworthy that the two molecules here tentatively identified as tetragalloylquinic acid derivatives have not been described yet in lentisk leaves. These compounds have shown to possess high activity against bronchial hyperreactivity and allergic reactions [72].

All these reports show the importance of developing new methodologies in order to increase the content of active compounds in *P. lentiscus* extracts. This could lead to a wider application of the extracts as nutraceuticals, medicines, or as sources of substances for different commercial and industrial applications.

In fact, the compounds detected here showed important biological activities. In particular, galloyl derivatives of quinic acid have been shown to have effective inhibition of Fe<sup>2</sup> <sup>+</sup>-induced lipid peroxidation in cells [73], anti-HIV, anti-allergic [74], and high antioxidant activities [68]. This class of molecules is among the most pharmacologically active natural products detected in several plant species [75]. In addition, gallotannins are applied as wood adhesives in the leather manufacturing as well as in the construction sector [13].

Flavonoids, especially with quercetin and myricetin skeleton, are considered powerful antioxidants distributed in several plant species with proven anti-inflammatory and anti-cancer actions [76–79]. Besides their application as medicinal compounds, they are specially utilized in cosmetic and nutraceutical products [11].

Moreover, the most abundant flavonoid detected in all extracts, the myricetin-3-*O*-rhamnoside, showed a noticeable lipid peroxidation inhibitionin in vitro tests, with very low IC<sup>50</sup> (inhibitory concentration at 50%) [80], being even more effective as antioxidant than vitamin C [81]. In addition, this molecule has demonstrated positive effects against the oxidative stress induced by hyperglycemia in C2C12 cells [82] and has shown significant inhibition in peroxynitrite-mediated DNA damage [83].

#### **3. Materials and Methods**

#### *3.1. Plant Material*

Fully expanded leaves from branches at the top of the canopy were randomly collected from adult plants of *P. lentiscus* growing in the coastal dunes of Southern Tuscany, Italy (42◦46′ N, 10◦53′ E). Harvesting was conducted in July 2019, around midday in order to ensure the high polyphenolic composition of the leaves [6].

After collection, the leaves were cleaned (to remove damaged parts, dust, and other contaminants from the natural habitat), immediately frozen in liquid nitrogen, freeze-dried, ground into a fine powder, and kept at −80 ◦C until the moment of extraction.

#### *3.2. Ultrasound-Assisted Extraction (UAE) Procedure*

Freeze-dried ground leaves (0.15 gweighted on a digital analytical balance Precisa® 125A) were extracted using ethanol in different percentages and volumes according to the design matrixes (Tables 1 and 2). The UAE was conducted in an ultrasonic bath (BioClass® CP104) using a constant frequency of 39 kHz and an input power of 100 W. The different temperatures and times, according to each trial (Tables 1 and 2), were monitored with a thermometer (Weber® 6750, Springfield, Illinois, USA) and a timer (Fisher Scientific®, Los Angeles, CA, USA). After the extraction, the samples were centrifuged (5 min, 9000 rpm, 5 ◦C-ALC® 4239R, Milan, Italy) and the supernatants were partitioned with 3 × 5 mL *n*-hexane, in order to remove lipophilic compounds that could interfere with the analysis. The hydroethanolic phase was reduced to dryness using a rotavapor (BUCHI® P12, Cornaredo, Italy; coupled to a vacuum controller V-855), and the residue was resuspended with 1.0 mL of MeOH: Milli-QH20 solution (1:1 *v*/*v*, pH 2.5 adjusted with HCOOH). These samples were used to conduct the HPLC-DAD analysis for the quantification of the different classes of polyphenols to construct the model. In addition, the extract with the highest polyphenolic content was chosen for the LC-MS/MS analysis in order to furnish a detailed characterization of its chemical composition.

#### *3.3. HPLC-DAD Quantification and LC-MS*/*MS Characterization of the Extracts*

High performance liquid chromatography coupled to diode array detection (HPLC-DAD) was used for quantification of the different polyphenolic classes of the extractsobtained at the different conditions tested.

The samples (5 µL) were injected into a Perkin® Elmer Flexar liquid chromatography equipped with a quaternary 200Q/410 pump and an LC 200 diode array detector (DAD) (all from Perkin Elmer®, Branford, Connecticut, USA). The stationary phase consisted in a Zorbax® C-18 column (250 mm × 4.6 mm, 5 µm particle size) and the eluents were (A) acidified water (0.1% HCOOH) and (B) acetonitrile (0.1% HCOOH).The following gradient was applied: 1 min (3% B), 1–55 min (3–40% B), 55–60 min (40% B), 60–61 min (3% B), with 62 min of total analysis time, in a flow rate of 0.6 mL min−<sup>1</sup> . Ten minutes of conditioning step were used to return to the initial conditions of the method.

The identification of the polyphenols with HPLC-DAD was carried out based on the retention time, UV-Vis spectral characteristics, comparison with those of the authentic standards acquired at 280 and 350 nm, as well as on the subsequent LC-MS/MS analysis. Quantifications were made by HPLC-DAD. The standards (gallic acid, myricitrin, rutin, and kaempferol-3-*O*-rutinoside from Sigma–Aldrich®–Merck®KGaA, Darmstadt, Germany) were used to obtain five-point calibration curves. If a commercial standard was not available, quantification was performed using the calibration curve of standards from the same phenolic group. The linearity of these calibration curves was determined by the coefficient of determination (R<sup>2</sup> ), being higher than 0.999 for all the three standards. The limit of detection (LOD) and quantification (LOQ), both expressed as µg/mL, were calculated using signal-to-noise ratio of 3 and 10, respectively [84]. The following limits of detection and quantification were found for the standards: LODgallic acid = 0.3 and LOQgallic acid = 0.85; LODrutin = 0.28 and LOQrutin = 0.6; LODmyricitrin = 0.12 and LOQmyricitrin = 0.38; LODkaempferol-3-O-rutinoside = 0.21 and LOQkaempferol-3-O-rutinoside = 0.49).

All the extracts were analyzed in triplicate. The quantitative results of the polyphenols (reported as mg per g of dry weight, DW) were expressed as: myricitrin (the most abundant flavonoid detected in the *P. lentiscus* leaf extracts), total tannin, total flavonoid, and total polyphenols contents, represented as the sum of individual tannins (TTC), flavonoids (TFC) and polyphenols (TPC) detected by HPLC-DAD analysis in each extract (Supplementary Materials Tables S1 and S2).

The characterization of polyphenols was conducted utilizing a LC–DAD-MS/MS system consisted of a Shimadzu® LCMS-8030 triple quadrupole mass spectrometer (Kyoto, Japan) operated in the electrospray ionization (ESI) negative mode and a Shimadzu®Nexera HPLC system (Kyoto, Japan), coupledto a diode array detector (DAD). A reversed-phase Waters® Nova-Pak C18 column (4.9 <sup>×</sup> 250 mm, <sup>4</sup> <sup>µ</sup>m; Waters®, Milford, MA, USA) was used. The mobile phase consisted of water (1% HCOOH, solvent A) and acetonitrile (1% HCOOH, solvent B) and the separation was conducted using the following gradient: 2% B isocratic (10 min), from 2% to 98% B (30 min), 98% B isocratic (7 min) in a flow rate of 0.6 mL min−<sup>1</sup> and 10 µL of injection volume. The conditions for MS analysis were nitrogen as nebulizing and drying gas (at flow rates of 3.0 and 15.0 L min−<sup>1</sup> , respectively); interface voltage of –3.5 kV; desolvation line temperature of 250 ◦C; heat block temperature of 400 ◦C. The spectrometer operated in product ion scan mode using analyte-specific precursor ions; and argon was used as collision-induced dissociation (CID) gas (at 230 kPa). Identification of individual phenolics was carried out by comparison with retention times, UV-Vis, MS and MS/MS spectra, bibliographic data, and available external standards injected in the same conditions (gallic acid, myricitrin, rutin, and kampferol-3-*O*-rutinoside, all from SigmaAldrich®–Merck®KGaA, Darmstadt, Germany).

#### *3.4. Experimental Designs: Optimization Procedure and Data Analysis*

The factors affecting the ultrasound-assisted extraction (Section 3.2.) were firstly screened using a fractional factorial design (FFD) (24−<sup>1</sup> ) in order to select the variables and levels to be applied during the optimization step. Based on the results, a Box–Behnken design was conducted to determine the best combination of the important variables selected [56].

#### 3.4.1. Screening Fractional Factorial Design: Selection of the Important Variables for the Extraction Optimization

Four factors: temperature (x1; 5 ◦C or 30 ◦C), time (x2; 15 or 30 min), solvent ratio (x3; 0.06 or 0.1 L g−<sup>1</sup> ) and ethanol fraction (x4; 50% or 75% *v*/*v*) were chosen as independent variables and analyzed in two levels (+1, <sup>−</sup>1; FFD 24−<sup>1</sup> ; Table 1). The variables and their levels were initially chosen based on their importance for the UAE of plant materials [47]. Nine trials were conducted (8 trials + central point), in different combinations of the variables (x<sup>1</sup> to x<sup>4</sup> inTable 1), all in triplicate. The details of the UAE process conducted are described in the Section 3.2.

The main effects of each factor (x<sup>1</sup> to x4) in the following responses: total tannins (TTC), total flavonoids (TFC), and total polyphenolic contents (TPC) were estimated by the calculation of the coefficients of each variable (b1, b2, b3, and b4) using the statistical software Minitab® 18 (LCC, Pennsylvania, USA). The factors that were significant in the regression analysis (*p* ≤ 0.05) were considered to have an impact on the responses and selected for the optimization step.

#### 3.4.2. Box–Behnken Design for Optimization of the Extraction Conditions

After the determination of the most important factors, these variables were optimized using a Box–Behnken design, a simple and more efficient three-level factorial design in comparison to other 3<sup>3</sup> designs [39]. Three independent variables (factors) were analyzed in three levels: temperature (x1; 30, 40, and 50 ◦C), solvent ratio (x2; 0.1, 0.15, and 0.2 L g−<sup>1</sup> ), and ethanol fraction (x3; 30, 40 and 50%, *v*/*v*).

Fifteen experimental trials, resulted from the combination of the three levels (−1, 0, 1) of each variable and three replicates of the central point were thus conducted in triplicate, following the BBD matrix (Table 2). The response variables (i.e., TTC, TFC, MYC, and TPC) were fitted to a second-order polynomial model equation (Equation(1)) that was used to predict the optimum conditions of extraction process and to construct the response surfaces (RSM).

$$Y = \beta 0 + \sum\_{i=1}^{k} \beta\_i X\_i + \sum\_{i=1}^{k} \beta\_{ii} X\_i^2 + \sum\_{i=1}^{k} \sum\_{j=i+1}^{k-1} \beta\_{ij} X i j \tag{1}$$

where *Y* represents the response variables, *Xi* and *Xj* are the independent variables affecting the response, β0, β*i*, β*ii*, and β*ij* are the regression coefficients of the model (intercept, linear, quadratic, and interaction terms, respectively), and k is the number of variables (*k* = 3). The variables and their levels, with both coded (−1, 0, 1) and uncoded (real values) are given in Table 2.

The Minitab®18 software (LCC, State College, PA, USA) was used for the RSM data analysis. To test the significance of the models, an ANOVA with 95% confidence level was carried out for each response. Furthermore, a lack of fit (LOF) test was performed to check the variability of the residues of the proposed models. The estimated coefficients of multiple determination (R<sup>2</sup> ) of the quadratic models and the adjusted coefficients of multiple determination (R<sup>2</sup> Adj) were also calculated. These coefficients reflect the fraction of the total variability in the response that is explained by the model.

In order to verify and validate the predicted optimal UAE conditions, experimental extractions were under the conditions selected as optimal for TTC, TPC (0.12 L g−<sup>1</sup> of 40% ethanol, at 50 ◦C), and TFC (0.13 L g−<sup>1</sup> of 50% ethanol, at 50 ◦C). The predicted and experimental responses were compared by a *t*-test and the model validation was confirmed if *p* > 0.05.

#### **4. Conclusions**

In conclusion, this study was able to define the optimal UAE conditions to obtain higher amounts of different polyphenolic classes from *P. lentiscus* leaves in a greener way when compared to conventional extraction methods, using a low percentage of organic solvent and less time consumed.

According to our findings, among the variables tested (i.e., temperature, ethanol fraction, and volume), the optimal conditions were slightly different only in terms of ethanol percentage: 40% for tannins and 50% for flavonoids but similar in solvent ratio (~0.13 L g−<sup>1</sup> ), temperature (50 ◦C), and time (15 min). A good agreement between the experimental and the predicted values at these optimal conditions showed the adequacy of the models obtained. Furthermore, this work brings novelty in the characterization of *P. lentiscus* leaf extracts, putatively identifying for the first time the presence of two tetragalloyllquinic acid derivatives.

These results are important considering the wide commercial applications of the different polyphenolic classes of this species, as well as the new trend in the green chemistry. Moreover, they may constitute the basis for future UAE processes applied in larger scale conditions.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2223-7747/9/11/1482/s1, Table S1:Polyphenolic content (in mg g−<sup>1</sup> DW) of individual tannins and gallic acid derivatives obtained in each trial of the optimization BBD design; Table S2: Flavonoidic content (in mg g−<sup>1</sup> DW) of individual compounds obtained in each trial of the optimization BBD design; Table S3: Regression coefficients (intercept, linear, quadratic, and interaction) of the models obtained for each response (total tannin—TTC, total flavonoids—TFC, myricitrin—MYC, and total polyphenols—TPC contents); Figure S1. MS/MS spectra of tetragalloyl quinic acids (isomer 1, A and isomer 2, B) corresponding to peak 8 and 9 of Table 4, respectively

**Author Contributions:** Conceptualization, A.G., C.B. and L.B.d.S.N.; methodology, A.G., C.B. and L.B.d.S.N.; software, L.B.d.S.N. and C.D.; formal analysis, L.B.d.S.N. and C.D.; investigation, L.B.d.S.N., C.D., A.G.; data curation, L.B.d.S.N.; writing—original draft preparation, C.D. and L.B.d.S.N.; writing—review and editing, C.B., A.G., L.B.d.S.N. and F.F.; supervision, A.G., C.B., F.F. and L.B.d.S.N. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by "Progetto FOE ECONOMIA CIRCOLARE, CNR" project: FOE-2019 DBA.AD003.139.

**Acknowledgments:** The authors extend their gratitude to "Analytical Food, Firenze" and to Cristian Marinelli for the technical support during the LC-MS/MS analyses.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
