An Optimized Method for Extraction and Characterization of Phenolic Compounds in Dendranthema indicum var. aromaticum Flower

Dendranthema indicum var. aromaticum plant has been widely used as herbal medicine in China, however, the material basis responsible for the therapeutic benefits remains largely unclear. This study aimed to provide an optimized method for extracting and characterizing phenolic compounds in D. indicum var. aromaticum flower. Firstly, an ultrasound-assisted method combined with central composite circumscribed (CCC) design was applied to optimize phenolic compound extraction. Ethanol-acetic acid (70%:2%, v/v) was selected as solvent, and the optimal extraction condition was: extraction temperature, 57 °C; solid/liquid ratio, 1:30 g/mL; extraction time, 20 min. Secondly, an effective and economic HPLC-PDA-ESI-MSn method was established and validated for phenolic compound characterization and quantification. As a result, 14 phenolic compounds were identified, including 8 phenolic acids and 6 flavonoids, and for the first time, oleuropein derivatives, chrysoeriol, and tricin are reported in D. indicum var. aromaticum flower. The content of phenolics identified by HPLC-MSn was 6.42 ± 0.32 mg/g DW. The optimized method for extraction and characterization of phenolic compounds has significant meaning to future pharmaceutical and medicinal research on D. indicum var. aromaticum, and the results in this study can provide references for herbal research.


Results and Discussion
optimization of phenolic compounds extraction. Extraction methods of phenolic compounds in plant material include soxhlet, microwave-assisted extraction, supercritical fluid extraction, ultrasound-assisted extraction, etc. [12][13][14] . With broad literature retrieval, ultrasound-assisted extraction was found to be a simple and effective method of phenoilcs extraction 8,10,15 , which does not require complicated equipment or technique and could greatly increase the extraction efficiency by strengthening the fragmentation process and assisting the release, diffusion, and dissolution of the components inside cells 11,16 . Furthermore, ultrasound-assisted extraction is more moderate and more secure to operate for possible future large-scale extraction. Therefore, ultrasound-assisted extraction was chosen in this study.
Aqueous methanol, ethanol, and acetone were usually used for phenolic compounds extraction from botanical materials, especially from herbs 10,15 . Acid, for instance, hydrochloric acid, formic acid, and acetic acid was often used to acidify the extraction environment for better efficiency 15,17 . Considering the possible future industrial application of the phenolic extraction, relatively low toxic solvents ethanol and acetic acid were chosen in this study. Figure 1 shows the effect of different proportion of ethanol and acetic acid combinations on total phenolic (TP) content. In general, the extraction effect was increased by adding more acetic acid at lower concentration of ethanol (30% or 50%) but not at higher concentration of ethanol (70%). The highest TP content was obtained from ethanol-acetic acid (70%:2%, v/v) solvent, therefore, ethanol-acetic acid (70%:2%, v/v) was selected to be the extracting solvent in the following studies.
To set up an appropriate CCC model, the ideal value range of independent variables should cover the inflection point of each independent variable. Therefore, the effect of each single factor on TP content was tested to obtain an approximate range for the CCC model construction, and the results were shown in Fig. 2. The increase of TP content was observed over the extraction time range (20∼30 min) ( Fig. 2A), solid/liquid ratio range (1:10∼1:20 g/mL) (Fig. 2B), and extraction temperature range (30∼50 °C) (Fig. 2C). Moderate higher temperature and longer extraction time could enhance the solubility of phenolic compounds and accelerate the whole extraction process, however, degradation should always be considered when extraction time and temperature exceed an appropriate range. The inflection point of each variable was selected as central point of the CCC design: extraction time, 30 min; solid/liquid ratio, 1:20 g/mL; extraction temperature, 50 °C.
Extraction conditions and the corresponding results under CCC design were shown in Table 1. Statistical analysis revealed that quadratic model significantly fitted to the CCC design (p = 0.0002, R 2 = 0.9211, statistic results were not shown). As seen in Fig. 3A, at certain extraction time, TP content rose with the increase of solid/liquid ratio. Further, relatively higher TP content was obtained when solid/liquid ratio tending to 1:30 with extraction time tending to 20 min simultaneously. As shown in Fig. 3B, TP content increased following temperature increasing at certain extraction time, and the rising trend was more obvious in 35 to 40 min extraction time range. In Fig. 3C, TP content went up following the rise of solid/liquid ratio under certain temperature. The more the solid/liquid ratio was close to 1:30 g/mL, the higher TP content was obtained in the temperature range of 50∼60 °C.
The optimal ultrasound-assisted phenolic extraction condition obtained from the CCC model was shown in Table 2: extraction temperature, 57 °C; solid/liquid ratio, 1:30 g/mL; extraction time, 20 min. The model predicted a maximum response of 1.29 g GAE/100 g DW under optimal condition. TP content of 1.27 ± 0.08 g GAE/100 g DW obtained from real experiments validated the CCC model (Table 2)  Effect of different proportion of ethanol and acetic acid combinations on TP content. Means of each two treatments were compared using least significant difference (LSD) statistic method. Lowercase letters a, b, and c were used to mark significance of difference (p < 0.05). Same letters between treatments mean insignificant difference. Different letters between treatments mean significant difference. www.nature.com/scientificreports www.nature.com/scientificreports/ GAE/100 g DW phenolics from methanolic extract of D. indicum (L.) Des Moul (another variety of D. indicum) flower through ultrasound-assisted extraction 18 . The TP content we got in this study was about 10 times higher than that of D. indicum (L.) Des Moul, which reveals the potential of developing D. indicum var. aromaticum as a source of phenolics. optimization of HpLC conditions. To analyze phenolic compounds by HPLC, acid was usually added in mobile phase 9,19 . Adding adequate amount of acid into mobile phase is beneficial to achieve complete separation, to lighten peak trailing, and to improve resolution of the compounds 20 . In this study, mobile phase B was fixed as acetonitrile, and mobile phase A was tested by various concentrations of formic acid (0.1%, 0.5%, and 1% v/v). It was found that using 0.1% or 0.5% formic acid did not result in the separation result as satisfactory as using 1% formic acid. Therefore, 1% formic acid was chosen in the following studies. Two 5 μm C18 columns (250 * 4.6 mm and 150 * 4.6 mm) were tested. By comparing with the longer column, using the shorter one could decrease flow rate from 1 mL/min to 0.6 mL/min and greatly shorten the elution time without affecting the separation result much. At last, various gradient methods were tried until a satisfactory chromatogram was obtained. Eventually, an effective and economic HPLC method was established. Figure 4 shows the HPLC-PDA chromatograms of phenolic compounds in D. indicum var. aromaticum flower.
Identification of phenolic compounds. In order to identify phenolic compounds, the external standard retention time and UV spectra from literatures were used to assess the target peaks preliminary, and the molecular structures were deduced based on MS and corresponding MS n fragment signals. The identified phenolic compounds were classified into hydroxybenzonic acid group, hydroxycinnamic acid group, and flavonoid group in this study. Besides, several fatty acids were identified together with the phenolic compounds. A summary of the MS n fragmentations of all compounds were shown (Table 3), and definable structures of identified phenolic compounds were drawn (Fig. 5).  Lowercase letters a and b were used to mark significance of difference (p < 0.05). Same letters between treatments mean insignificant difference. Different letters between treatments mean significant difference. www.nature.com/scientificreports www.nature.com/scientificreports/ Compound 2 was identified as coumaroyloleuropein. As illustrated in Fig. 6B, firstly, the presence of its base ion at m/z 539 in MS2 with next base ion at m/z 377 in MS3 indicated the cleavage of a glucosyl from oleuropein 22,23 . Secondly, base ion at m/z 539 generated from parent ion at m/z 685 revealed the loss of a coumaroyl moiety, therefore, a structure of coumaroyloleuropein was deduced. Oleuropein is a bitter phenolic compound which mainly exists in green olives, olive leaves, and argan oil 24,25 . It is for the first time to report the presence of oleuropein derivative in D. indicum var. aromaticum flower.
For compound 8, its parent ion at m/z 569 was 30 Da higher than that of oleuropein at m/z 539, which means that compound 8 might be methoxyoleuropein, but its MS n spectrum did not give fragment ions as m/z 539 or 377 as expected according to literature reports 22,23 . Compound 8 was tentatively named as methoxyoleuropein isomer.
Spectrum at m/z 125, corresponding to MS signals produced by gallic acid standard. Furthermore, parent ion of compound 6 at m/z 187 (18 Da higher than that of gallic acid) indicated the peak was gallic acid monohydrate 19 .
Compound 7 had the same parent ion with shikimic acid at m/z 173, but its UV absorbance and MS2 fragments did not fit to previous report on shikimic acid 27 . The obtained information of MS fragments of compound 7 is insufficient to deduce its precise structure, so compound 7 was called shikimic acid isomer tentatively.  www.nature.com/scientificreports www.nature.com/scientificreports/ Through comparison of retention time, UV spectrum, and MS n patterns with external standards, luteolin, apigenin, and acacetin were identified for compounds 9, 11, and 17, respectively. UV spectrum of compound 3 was pretty similar as luteolin, and the cleavage of a glucose from luteolin molecule could produce MS2 ion at  100)) have been reported as chrysoeriol (3′-methoxy derivative of luteolin) and tricin in previous studies 19,28 . Chrysoeriol exists in many species of Artemisia genus 29 , and tricin occurs in rice bran and other grass specie such as wheat, maize, and barley 30 , both of these two flavones are for the first time to be reported in D. indicum var. aromaticum flower.
Other compounds. Several hydroxy fatty acids were identified together with phenolic compounds. Compounds 15, 19, 20, 21, 22, and 23 were characterized to be hydroxy fatty acids, according to the phenomenon that water  Table 2. Optimal condition and TP content obtained from prediction and real experiment under optimal condition. a Gallic acid equivalent. b Mean ± SD (n = 3).
www.nature.com/scientificreports www.nature.com/scientificreports/ molecules cleaved from the aliphatic moiety consecutively in their MS2 spectra 31,32 . These peaks show up at end of the chromatogram, and most of them have longer retention time than the phenolic compounds except monohydroxy-octadecaditrienoic (compound 15) which comes earlier than peak of acacetin (compound 17) in the chromatogram (Fig. 4). Hydroxy fatty acids were found to have many biological activities, for instance, cytotoxicity and anti-inflammatory activity 33 , which might contribute to the therapeutic effects of D. indicum var. aromaticum plant, too.

Method validation and phenolic compounds quantification.
In this study, parameters of linearity, sensitivity, accuracy and precision were used for HPLC method validation. Results for method validation were summarized in Table 4. The correlation coefficients of the calibration curves were all over 0.99, which reflected that the detected concentration values were highly coincident with the real values. The method sensitivity was confirmed to be adequate as LOD and LOQ values were lower than 0.036 μg/mL and 0.109 μg/mL. Recovery rates of apigenin, acacetin, and luteolin obtained here varied from 99.50% to 102.19%, and the RSD values were lower than 1.84% and 2.70% for intra-day and inter-day experiments, respectively, indicating high accuracy and good precision of the analytical method in this study.
As shown in Table 5, content of phenolics in D. indicum var. aromaticum flower identified by HPLC-MS n was 6.42 ± 0.32 mg/g DW, consisting of 3.63 ± 0.17 mg/g DW of flavonoids and 2.79 ± 0.15 mg/g DW of phenolic acids. Among all the phenolic compounds in D. indicum var. aromaticum flower, luteolin (1.61 ± 0.11 mg/g DW) took up the highest proportion, around 25% of the TP content, which was found to be similar to the luteolin content (0.5∼2.1 mg/g DW) in flower of D. indicum (original variety of D. indicum var. aromaticum) reported in previous study 34 . Gong. et al. have isolated 20 mg acacetin from 1000 g dry flower of D. indicum var. aromaticum by absolute ethanol cold extraction and column chromatographyl 35 . Comparing with the acacetin content in this study (0.92 ± 0.02 mg/g DW, 0.92‰), the low yield of acacetin (0.02‰) in Gong's work could be mainly caused by low extraction efficiency and high waste ratio during the isolation procedure. Previous studies stated that change of climate conditions (temperature, irradiation, rainfall, ect.) between years and different harvest periods could influence chemical composition and component content in plant material 36,37 . There were two phenolic compounds acacetin-7-O-β-D-glucopy ranoside and apignein-7-O-β-D-glucopy ranoside isolated from D. indicum var. aromaticum flower in Lu's research by column chromatography and semi-preparative HPLC in 2009 2 but not found in this study. Even though plant materials were harvested from the same location, certain degree of difference of chemical profile would exist between different material batches, which should be considered and accepted in research work.

Conclusions
The phenolic compound composition of D. indicum var. aromaticum flower was studied extensively for the first time. Firstly, ethanol-acetic acid (70%:2%, v/v) was selected as extraction solvent. Ultrasound-assisted extraction method was optimized by CCC design, and under optimal condition (extraction temperature, 57 °C; solid/ liquid ratio, 1:30 g/mL; extraction time, 20 min), 1.27 ± 0.08 g GAE/100 g DW TP content was obtained from D. indicum var. aromaticum flower. Secondly, an effective and economic HPLC-PDA-ESI-MS n method was established, and the analytical method was validated by sensitivity, accuracy, and precision. As a result, 14 phenolic compounds were identified and quantified, including 8 phenolic acids and 6 flavonoids. For the first time, oleuropein derivatives, chrysoeriol, and tricin are reported in D. indicum var. aromaticum flower. In summary, the optimized method for extracting and characterizing phenolic compounds in D. indicum var. aromaticum flower www.nature.com/scientificreports www.nature.com/scientificreports/ has significant meaning for future pharmaceutical and medicinal research on D. indicum var. aromaticum plant, and the results in this work would provide references for future herbal research.

Material and Methods
Chemicals and materials. Eight standards were used in this work: gallic acid, caffeic acid, luteolin were purchased from Adamas Reagent (Shanghai, China); acacetin was purchased from TCI (Tokyo, Japan); apigenin was purchased from WAKO Pure Chemical Industries (Osaka, Japan); protocatechuic acid, ferulic acid, and hesperidin were purchased from Sigma-Aldrich (MO, USA). Acetonitrile and formic acid in HPLC grade were purchased from Fisher Chemical (Geel, Belgium). Ethanol and acetic acid in AR grade were purchased from Sinopharm Chemical Reagent Co., Ltd (Shanghai, China). Folin-Ciocalteu and Na 2 CO 3 were purchased from Sigma-Aldrich (MO, USA).
Flower of D. indicum var. aromaticum was harvested in Shen Nongjia area of Hubei province, China. The plant species was identified by Prof. Keli Chen, Hubei University of Chinese Medicine, and authenticated by Herbarium, Kunming Institute of Botany, Chinese Academy of Sciences. The certificate of plant material authentication was provided as Supplementary. Plant material was harvested freshly, and spread at a shady, cool and well-ventilated place to dry naturally. Dry material was kept in −80 °C freezer before using. sample extraction. Dry flowers were milled to a fine and uniform particle size by a YB-500A grander (Shanghai Lijian Machinery Co., Ltd, Shanghai, China). Half gram of the powder was mixed with certain volume of solvent, and followed with an ultrasound bath extraction (KQ-600DB, 40 kHz, Kunshan Ultrasonic Instruments Co., Ltd, China). The ultrasound working power was set at 360 W. External water circulated from a water bath to keep the extraction temperature stable. Centrifugation of 10,000 g was operated at 4 °C for 10 min   Table 3. HPLC-MS n information on phenolic compounds and hydroxy fatty acids identification in D. indicum var. aromaticum flower. a No data available.
Selection of extraction solvent. Different concentrations of ethanol (30%, 50%, and 70%) and acetic acid (2%, 5%, and 10%) were mixed in orthogonal design. Half gram of weighed samples were ultrasound extracted with 10 mL various solvents at 40 °C for 30 min. Supernatant was collected after centrifugation as described above.
Selection of the range of extraction temperature. Half gram of weighed samples were ultrasound extracted with 10 mL ethanol-acetic acid (70%:2%, v/v) at different temperature (30,40,50, and 60 °C) for 30 min. Supernatant was collected after centrifugation as described above.
CCC design. Experiment was designed according to CCC model with five levels (−1.68, −1, 0, 1, and 1.68) and three variables (extraction temperature, solid/liquid ratio, and extraction time) to achieve the best variable combination for TP extraction. The CCC design consists of six central points, eight factorial points, and six axial points, generating 20 sets of experiments. Both coded and uncoded form of independent variables were shown in Table 1.
Determination of TP content. Folin-Ciocalteu method reported in Cicco's study 38 was adopted to measure the TP content, with slight modifications. In brief, 150 μL of appropriately diluted extract, 150 μL of Folin-Ciocalteu (50%, v/v), and 1.2 mL of Na 2 CO 3 (5%, m/v) were mixed. After two hours' incubation at room temperature, absorbance under 760 nm was measured with a blank (150 μL of extraction solvent instead of the extract) using a nucleic acid/protein analyzer (Beckman Coulter, DU 730, CA, USA). Calibration curve was established using gallic acid standard. Results were recorded as gallic acid equivalent (g GAE/100 g DW). Absorbance experiments were operated in triplicate.    15,39,40 .
HPLC-ESI-MS n (Thermo Fisher Scientific, Waltham, MA, USA) was used to identify and characterize phenolic compounds. The same HPLC condition described above was used. Selected compounds were analyzed by MS2, MS3, till MS4 as needed. Negative mode of ionization was performed, and a mass range of m/z 50∼700 was covered for full scan. Collision gas was ultrahigh pure helium (He). The ionization parameters were set as follows:, 40 units/min of sheath gas (N 2 ); 2 units/min of auxiliary gas (N 2 ); 4.50 kV of spray voltage; 300 °C of capillary temperature; −1.00 V of capillary voltage; −8.77 V of tube lens offset voltage.
To determine intra-day and inter-day precision, spiked samples with standards were analyzed at five different time points in one day and at the same time point for five consecutive days.
Since not all standards for each compound were commercially available, only apigenin, acacetin, and luteolin were quantified directly with their authentic standards, while other compounds were quantified by internal standards with similar structure and properties using relative response factor (RRF). The internal standard protocatechuic acid was used to quantify 3-O-caffeoyl-5-O-p-coumaroylquinic acid, gallic acid monohydrate, and shikimic acid isomer. The internal standard ferulic acid was used to quantify coumaroyloleuropein, 3,5-dicaffeoylquinic acid monohydrate, 3,5-dicaffeoylquinic acid, methyoxyoleuropein isomer, and prenyl-dimethoxy-caffeoyl-p-coumaric acid. The internal standard hesperidin was used to quantify tricin, luteolin-O-glucoside and chrysoeriol.
Fixed concentration (20 μg/mL) of internal standards were mixed with the sample extract and with the known compounds (25 μg/mL) mentioned above. Quantification of the unknown compounds was carried out according to the following formula: where RRF = (A k /A i )/(C k /C i ); A k means peak area of known phenolic compound; A i means peak area of internal standard; C k means concentration of known phenolic compound; C i means concentration of internal standard; A means peak area of unknown phenolic compound; R means recovery rate.

Statistical analysis.
For the single factor extraction tests, ANOVA was adopted and the TP content of different extraction treatments were analyzed under Tukey's test by Minitab 17 (Minitab Inc., State College, PA, USA). Data were expressed as mean ± SD (n = 3). Statistic of the CCC design was performed using Design Expert 10 (Stat Ease Inc., Minneapolis, USA). ANOVA was adopted for analyzing model main effect and interaction between variables. P < 0.05 was considered as significant.  Table 5. Content of phenolic compounds in D. indicum var. aromaticum flower. a Mean ± SD (n = 3).