Identiﬁcation of phenolic compounds in Australian grown dragon fruits by LC-ESI-QTOF-MS/MS and determination of their antioxidant potential

Dragon fruit is a popular tropical fruit that has a high phenolic content which are the main contributors to the antioxidant potential and health beneﬁts of dragon fruit pulp and peel waste. Although some phenolic compounds in dragon fruit have previously been reported, a comprehensive analysis of complete phenolic proﬁle of the Australian varieties has not been conducted. Thus, the aim of this study was to extract, identify and quantify phenolics from dragon fruits grown in Australia. Phenolic compounds were extracted from the peels and pulps of white and red dragon fruit. Phenolic content was determined by total phenolic content (TPC), total ﬂavonoid content (TFC) and total tannin content (TTC), while antioxidant activities were measured by 2,2-diphenyl-1-picrylhydrazyl (DPPH), ferric reducing antioxidant power (FRAP), 2,2 0 -Azino-bis-3-et hylbenzothiazoline-6-sulphonic acid (ABTS) and total antioxidant capacity (TAC). The results showed that dragon fruit pulp had a higher total phenolic content and stronger antioxidant capacity than peel, while the peel had a higher content of ﬂavonoids and tannins than the pulp. Liquid chromatography electrospray ionization quadrupole time-of-ﬂight mass spectrometry (LC-ESI-QTOF-MS/MS) was used for the characterization of phenolic compounds, a total of 80 phenolics including phenolic acids (25), ﬂavonoids (38), lignans (6), stilbene (3) and other polyphenols (8) were characterized in all dragon fruits. High performance liquid chromatography equipped with photodiode array detector (HPLC-PDA) quantiﬁed the phenolic compounds in different portion of dragon fruit and showed that dragon peel had higher concentrations of phenolics than pulp. The results highlighted that both dragon fruit peel and pulp are potential sources of phenolic compounds, with peel in particular being a source of antioxidant phenolics with potential as ingredients for the food and pharmaceutical industries. an


Introduction
Dragon fruit (Hylocereus spp.) is a widely consumed tropical fruit which is considered healthy partly due to its high content of phenolic compounds (Zain et al., 2019). The global market value of dragon fruit reached 4.9 billion US dollars worldwide in 2016 (Chen, 2018). Dragon fruit pulp is edible and it is usually eaten raw or used for making commercial products such as juices, ice cream, jam and yogurt (Nurul and Asmah, 2014). The phenolic compounds in pulp possess antioxidant activity and have a range of potential health benefits (Som et al., 2019). However, the dragon fruit peel is non-edible, and mostly goes to waste, despite its high phenolic content . Excessive peel waste results in both economic and environmental impacts, particularly as organic waste going to landfill is a major contributor to methane release into the atmosphere (Chen, 2018). Emerging applications to utilise dragon fruit peel waste include fruit spreads and food additives, with isolation or concentration of antioxidants for food, pharmaceutical and cosmetics industries warranting further exploration (Ferreres et al., 2017).
Phenolic compounds are a major group of phytochemical secondary metabolites (Hoda et al., 2019) that exhibit strong antioxidant capabilities due to the presence of phenolic groups that donate electrons or conjugate with metal ions (Hoyweghen et al., 2012). Phenolic compounds can be categorized into different groups such as flavonoids, phenolic acids, stilbenes and lignans based on the number of carbon molecules and the complexity of the structure (Hoda et al., 2019). Each phenolic group has unique attributes due to their specific molecular structure (Campos-Vega and Oomah, 2013). White dragon fruit (Hylocereus undatus) and red dragon fruit (Hylocereus polyrhizus) are two major varieties found to contain large amounts of phenolic compounds. White dragon fruit has red peel and white pulp, where the pulp was used as an indigenous medicine for healing wounds and bruises in Mexico, partly due to its antioxidant capability (Perez et al., 2005). Red dragon fruit has red peel and red pulp, which can be used for making natural color additives for healthy food due to its pulp color and antioxidant properties. The predominant phenolic compounds identified in these two varieties are flavonols, flavanones and hydroxycinnamic acid derivatives (Garcı´a-Cruz et al., 2017). In addition, phenolic acids including gallic acid, syringic acid, caffeic acid, p-coumaric acid, cinnamic acid and quinic acid have also been characterized in white and red dragon fruits (Castro-Enrı´quez et al., 2020;Luo et al., 2014;Zain et al., 2019).
Although phenolic compounds are abundant in dragon fruit, their content and availability can be affected by varieties, plant part, growth conditions, terroir and extraction method (Hoda et al., 2019). Thus, developing an optimum extraction method is important, as it allows the accurate identification and quantification of phenolic compounds from and within extracts. The most widely used extraction method currently is solvent extraction using various proportions of organic solvents, for which variations in solvents and extraction conditions result in different proportions and amounts of phenolics being extracted (Chan et al., 2014;Choo et al., 2016). After extraction, antioxidant activity or capacity can be determined by the estimation of phenolic contents by using selected antioxidant assays. Phenolic content has been measured through determining total phenolic content (TPC), total flavonoid content (TFC) and total tannins content (TTC) assays (Sa´nchez-Rangel et al., 2013). Antioxidant potential can be estimated by 2,2-Diphenyl-1-picrylhydrazyl (DPPH) assay, ferric reducing antioxidant power (FRAP) assay, 3-ethyl benzothiazoline-6-sulphonic acid (ABTS) assay and total antioxidant capacity (TAC) assay (Haida and Hakiman, 2019). For characterization and quantification of phenolic compounds in plant foods, liquid chromatography-mass spectrometry (LC-MS/MS) is the most widely used technique (Lucci et al., 2017). In previous studies, several phenolic compounds had been identified through LC-MS in dragon fruit such as cinnamic acid, quinic acid, quercetin-3-O-hexoside, apigenin, 3,4-dihydroxyvinylbenzene and apigenin (Lira et al., 2020;Zain et al., 2019). However, previous studies on phenolic profile of dragon fruit peels and pulps characterized only some major phenolic compounds, while a complete phenolic profile in dragon fruit peel and pulp is lacking for varieties grown in Australia.
In this study, phenolic compounds were extracted from the pulps and peels of two Australian grown dragon fruit varieties. Phenolic content and antioxidant activity of the extracts were determined by different phenolic estimation methods (TPC, TFC and TTC) and antioxidant assays (DPPH, ABTS, FRAP and TAC), while phenolic compounds were further characterized and quantified through liquid chromatography with electrospray ionization-quadrupole time-of-flight mass spectrometry (LC-ESI-QTOF-MS/MS) and high performance liquid chromatography equipped with photodiode array detector (HPLC-PDA). The aim of this study was to provide relatively comprehensive information for the antioxidant activities and phenolic profiles of Australian dragon fruit, as part of assessing the potential value of dragon fruit peel waste as a source of new nutritional, cosmetic or pharmaceutical antioxidant ingredients.

Sample preparation
White dragon fruit (Hylocereus undatus) and red dragon fruit (Hylocereus polyrhizus) of 2 kg were purchased from the Queen Victoria Market, Melbourne. The fruits were cleaned, and the peel and pulp were separated into white dragon fruit peel (DWL), white dragon fruit pulp (DWP), red dragon fruit peel (DRL) and red dragon fruit pulp (DRP). Samples were trimmed into slices, freeze dried at À20 ℃ for 48 h and lyophilized at À45 ℃/50 MPa by Dynavac engineering FD3 Freeze Drier (W.A., Australia) and Edwards RV12 oil sealed rotary vane pump (Bolton, England). The dried peels and pulps were made into powders and stored at -20 ℃.

Extraction of phenolic compounds
Phenolic compounds were extracted from 1 g of sample by 15 mL 80% ethanol, homogenized by the Ultra-Turrax T25 Homogenizer (IKA, Staufen, Germany) and incubated in a ZWYR-240 shaking incubator (Labwit, Ashwood, Vic, Australia) with 120 rpm at 4 ℃ for 14 h sequentially. When the incubation was finished, samples were centrifuged by the Hettich Refrigerated Centrifuge (ROTINA 380R, Tuttlingen, Baden-Wu¨rttemberg, Germany) at 24400g for 10 min under 10 ℃. After centrifugation, supernatant was collected and filtered with 0.45 lm syringe filter (Thermo Fisher Scientific Inc., Waltham, MA, USA) for antioxidant and LC-MS analysis.

Estimation of phenolic contents and antioxidant assays
For overall phenolic estimation, TPC, TFC and TTC were performed, while for overall total antioxidant capacity determination, DPPH, FRAP, ABTS and TAC were utilized according to the methods of Suleria et al. (2020), Tang et al. (2020). Absorption data was attained using a MultiskanÒ Go microplate photometer (Thermo Fisher Scientific Inc., Waltham, MA, USA).

Determination of total phenolic content
Total phenolic content was determined by following the method of Wang et al. (2021) using Folin-Ciocalteu reagent. Dragon fruit sample of 25 lL was added into a 96-well plate (Corning Inc., Midland, NC, USA) together with 25 lL diluted F-C reagent (1:3 diluted with water) and 200 lL water before incubation at room temperature for 5 min. Then 25 lL 10% (w:w) sodium carbonate was added for basifying the mixture followed by a 60-min incubation in dark condition. The absorbance of the solutions was determined at 765 nm wavelength with a spectrophotometer (Thermo Fisher Scientific Inc., Waltham, MA, USA) and the standard curve of absorbance verse weight of gallic acid (concentrations ranging from 0 to 200 lg/mL) was plotted. The TPC was calculated with the standard curve and expressed in the form of gallic acid equivalents (GAE) per gram (mg GAE/g) of freeze-dried weight sample.

Determination of total flavonoid content
Total flavonoid content was determined by the aluminum chloride method of Stavrou et al. (2018) with some modifications. Dragon fruit sample of 80 lL was added into a 96well plate together with aluminum chloride (2% diluted with ethanol) of 80 lL and sodium acetate solution (50 g/L) of 120 lL, followed by an incubation at 25 •C for 2.5 h. Then, the absorbance of the solution was determined at 440 nm wavelength by a spectrophotometer, and the standard curve of absorbance verse weight of quercetin (0-50 lg/mL) was plotted. The TFC value was calculated based on the standard curve and expressed as mg of quercetin equivalent per gram (mg QE/g) of dry weight samples.

Determination of total tannin content
The total tannins content was determined by the modification of the vanillin and p-dimethylaminocinnamaldehyde methods of Stavrou et al. (2018). Dragon fruit sample of 25 lL was added into a 96-well plate together with 4% vanillin solution (diluted with methanol) of 150 lL and 32% sulfuric acid of 25 lL, followed by an incubation at 25 •C for 15 min. The absorbance was measured at 500 nm wavelength by a spectrophotometer, and the standard curve of absorbance verse weight of catechin (0-1000 lg/mL) was plotted. The TTC value was expressed as mg of catechin equivalent per gram (mg CE/g) of dry weight samples.
2.4.4. 2,2-Diphenyl-1-picrylhydrazyl antioxidant assay DPPH radical scavenging activity was determined by the modification of the DPPH assay method of Sogi et al. (2013). Dragon fruit sample of 40 lL was added into a 96-well plate together with 0.1 mM DPPH methanolic solution of 40 lL, following by a vigorous shake and an incubation at 25 •C for 30 min. The absorbance was measured at 517 nm wavelength by a spectrophotometer, and the standard curve of absorbance verse weight of ascorbic acid (0-50 lg/mL) was plotted. The DPPH radical-scavenging activity of the solution was calculated based on the standard curve and expressed as mg of ascorbic acid equivalents per gram (mg AAE/g) of dry weight samples.
2.4.5. Ferric reducing-antioxidant power assay FRAP assay was performed using a modification of the method of Sogi et al. (2013). The FRAP dye was made by the mix of 300 mM sodium acetate solution, 10 mM TPTZ (2,4,6-tripyridyl-s-triazine) solution as well as 20 mM Fe[III] solution in 10:1:1 ratio. Dragon fruit sample of 20 lL was added into a 96-well plate together with previously prepared FRAP dye solution of 280 lL, followed by a 10 min incubation at 37 •C. The absorbance was measured at 593 nm wavelength by a spectrophotometer, and the standard curve of absorbance verse weight of ascorbic acid (0-50 lg/mL) was plotted. The FRAP results were calculated based on the standard curve and expressed as mg of ascorbic acid equivalents per gram (mg AAE/g) of dry weight samples.
2.4.6. 2,2-Azino-bis-3ethylbenzothiazoline-6-sulfonic acid radical scavenging assay The ABTS radical scavenging activity was determined by the ABTS + radical cation decolorization assay of Sogi et al. (2013) with slight modifications. The ABTS dye was made by mixing of 5 mL ABTS solution (7 mmol/L) with 88 lL of potassium persulfate solution (140 mM) and a 16-hour dark incubation of the mixture at room temperature. Then, an initial absorbance (0.7 at 734 nm) of the prepared ABTS + solution was obtained by diluting with analytical grade ethanol. After that, dragon fruit sample of 10 lL was added into a 96-well plate together with previously prepared diluted ABTS solution of 290 lL, following by a 6-minute dark incubation at room temperature. The absorbance was measured at 734 nm wavelength, and the standard curve of absorbance verse weight of ascorbic acid (0-150 lg/mL) was plotted. The ABTS results were calculated based on the standard curve and expressed as mg of ascorbic acid equivalents per gram (mg AAE/g) of dry weight samples.

Total antioxidant capacity assay
Total antioxidant capacity was determined by modifying the phosphomolybdate assay method of Jan et al. (2013); Mashwani et al. (2013). The phosphomolybdate dye was made by mixing 0.6 M H 2 SO 4 , 28 mM Na 3 PO 4 and 4 mM ammonium molybdate in the ration of 1:1:1. Then, dragon fruit sample of 40 lL was added into a 96-well plate together with 260 lL previously prepared phosphomolybdate dye, followed by a 90-minute incubation at 95℃ and a 10-minute cooling at room temperature. The absorbance was measured at 695 nm wavelength, and the standard curve of absorbance verse weight of ascorbic acid (0-200 lg/mL) was plotted. The TAC results were calculated based on the standard curve and expressed as mg of ascorbic acid equivalents per gram (mg AAE/g) of dry weight samples.

LC-ESI-QTOF-MS/MS analysis
The LC-MS determination was conducted using a modification of the method . Phenolic characterization was performed by an Agilent 1200 series HPLC (Agilent Technologies, CA, USA) connected with an Agilent 6520 Accurate-Mass Q-TOF LC-MS (Agilent Technologies, CA, USA). A Synergi Hydro-RP 80A, LC column 250 mm Â 4.6 mm, 4 lm (Phenomenex, Torrance, CA, USA) was utilized for compound separation. Mobile phase A was made by the mix of water and acetic acid (in the ratio of 99.5:0.5, v/v), and mobile phase B was made by the mix of acetonitrile, water and acetic acid (in the ratio of 50:49.5:0.5, v/v/v), followed by a 15-minute degassing at 21 ℃ for both mobile phases. Filtration of the samples was performed with the syringe (Kinesis, Redland, QLD, Australia) coupled with the 0.45 lm syringe filter (Thermo Fisher Scientific Inc., Waltham, MA, USA) before the filtrates were transferred into HPLC vials. The injection volume of each sample was set to be 5 lL and the flow rate was set to be 0.8 mL/min. The program of the gradient elution carried out by a mixture of mobile phase A and B was set as follow: 10% B (0 to 20 min); 25% B (20 to 30 min); 35% B (30 to 40 min); 40% B (40 to70 min); 55% B (70 to 75 min); 80% B (75 to 77 min); 100% B (77 to 79 min); 100% B (79 to 82 min); 10% B (82 to 85 min). For MS/MS, the operational source utilized for both negative and positive modes was electrospray ionization (ESI), and mass spectra in the range 50 to 1300 (m/z) were attained with collision energy (10, 15 and 30 eV) for fragmentation. The nitrogen gas temperature of the mass spectrometry was set to be 300 •C with a flow rate of 5 L/min. The sheath gas temperature was set to be 250 •C with a flow rate of 11 L/min, and a nebulizer gas pressure of 45 psi. A 500 V nozzle voltage and a 3.5 kV capillary were also set. For data collection and analysis, an Agilent MassHunter data acquisition software version B.03.01 was used.

HPLC analysis
Based on the method of Ma et al. (2019), the putative quantification of targeted phenolic compounds was carried out using an Agilent 1200 series HPLC (Agilent Technologies, CA, USA) connected with a PDA detector. Apart from a sample injection volume of 20 lL, the column and conditions utilized in HPLC were the same as that was previously described in LC-ESI-QTOF-MS/MS. The detection was performed under wavelengths of 280, 320, and 370 nm for various phenolic compounds. Specifically, hydroxybenzoic acids were identified under 280 nm wavelength, hydroxycinnamic acids were identified under 320 nm, and flavonol group was identified under 370 nm. Data collection and analysis were carried out by an Agilent LC-ESI-QTOF-MS MassHunter data acquisition software version B.03.01.

Statistical analysis
The mean differences between different samples were analyzed by one-way analysis of variance (ANOVA) and Tukey's honestly significant differences (HSD) multiple rank test at p 0.05. ANOVA was carried out by Minitab for Windows version 19.0 (Minitab, LLC, State College, PA, USA). The results are shown in the form of mean ± standard deviation (SD). Correlations between polyphenol content and antioxidant activities were analyzed by Pearson's correlation coefficient at p 0.05.

Phenolic estimation (TPC, TFC and TTC)
Dragon fruit was reported to contain large amounts of phenolic compounds with strong antioxidant capacity, including fla-vonoids and phenolic acids. The phenolic contents in dragon fruit pulps and peels were determined by TPC, TFC, and TTC assays mentioned in Table 1.
As for TPC results, DRP had a significantly higher value (0. 39 ± 0.02 mg GAE/g) than the rest of the samples, while DWP and DWL has comparative phenolic contents (0.27 ± 0.01 and 0.23 ± 0.01 mg GAE/g) and DRL has the lowest value (0.17 ± 0.01) (p < 0.05). The TPC values from our study are close to the study conducted by Choo et al. (2016), in which they determined the TPC of white and red dragon fruit pulps to be 0.29 ± 0.02 and 0.24 ± 0.01 mg GAE/g. However, the pattern of the TPC results of Nurliyana et al. (2010) was contradictory to our research as they found that white and red peel samples had higher phenolic contents than pulp samples. They attributed the higher phenolic content in peels to the abundance of betacyanins, which contributes to TPC value apart from polyphenols (Tenore et al., 2012). An additional reason for the contradictory results between their study and ours might be the freeze-drying process we applied to the peel samples. Shofian et al. (2011) have suggested that freeze-drying can cause degradation of some oxidatively sensitive phenolic compounds, thus lowering the antioxidant activity in tropical fruits. The different varieties and extraction solvent used in the two studies may also contribute to differences in the TPC observed (Choo et al., 2016).
Peel samples including DWL and DRL has significant higher values for TFC (26.23 ± 1.85 and 21.66 ± 1.91 lg QE/g respectively) than DWP (2.39 ± 0.20 lg QE/g), while there was no significant difference in the flavonoid content in both peels. Previously, Wojdyło et al. (2007) reported that although polyphenols were present in both peel and pulp, flavonoids mostly existed in the peels, which is in agreement with the results we observed. However, Tenore et al. (2012) extracted flavonoids from red dragon fruit peel and pulp by 70% methanol which is much higher than for our results. The difference might be attributed to the sub-fraction method they used for extraction which was able to separate flavonoids from other phytochemicals to give a higher TFC value and the Australian varieties were subjected to the assay specifically in our study (Tenore et al., 2012).
The TTC assay only detected measurable levels for the DWL sample, with a value of 24.26 ± 2.04 lg CE/g. Wu et al. (2006) reported tannin contents in red dragon fruit peel and pulp extracted by 80% acetone (83.3 ± 1.1 and 72.1 ± 0. 2 mg CE/g respectively). Rebecca et al. (2010) measured tannins in red dragon fruit pulp extracted in 96% ethanol (2.3 ± 0.2 mg CE/g), which is also contradictory with our results.
The difference in tannin content may be explained by the difference in variety and the extraction solvents utilized (Sulaiman et al., 2011). Also, the plant varieties may also be an important factor these difference from previous studies, since the dragon fruits studied were from Taiwan and Malaysia, while we used Australian varieties as samples.

Antioxidant activities (DPPH, FRAP, ABTS and TAC)
A combination of antioxidant assays is often used to determine the antioxidant capacity of food samples containing a complex mix of phytochemicals. In this study, the antioxidant capabilities of dragon fruit pulps and peels were determined using DPPH, FRAP, ABTS and TAC assays. The results are shown in Table 1.
DPPH is the most commonly used assay to characterize free radical scavenging capabilities of food samples based on their hydrogen atom donation ability. From Table 1, DRP has significantly higher activity (0.29 ± 0.02 mg AAE/g) than the other three samples (p < 0.05), followed by DWP with 0. 09 ± 0.01 mg AAE/g (p < 0.05), which is also higher than DWL and DRL (both are 0.07 ± 0.01 mg AAE/g) (p < 0.05). Previously, Nurliyana et al. (2010) reported that DRP has higher DPPH value than DWP, which is consistent with our results. The stronger antiradical capability in DRP is likely to be due to the abundance of pigments (betalains) with antioxidant potential. However, these authors indicated that peels have higher antiradical capacities than pulps, which is the reverse of our findings. Kim et al. (2011) also reported higher antiradical capacities in peels compared with pulps, which they attributed to the higher content of phenolic compounds in peels. The reason for the lower DPPH in our peel samples might be plant strain differences (Shofian et al., 2011).
The FRAP assay measures the antioxidant ability of food samples by utilizing a ferric tripyridyltriazine (Fe III -TPTZ) complex to determine their reducing potential. The results of the FRAP assay shared the same pattern as the DPPH results, in which DRP has significantly higher value than the other three samples (53.02 ± 2.76 lg AAE/g), while DWP has a significantly higher value (38.80 ± 0.45 lg AAE/g) than the peels DWL and DRL (25.50 ± 0.73 and 18.12 ± 0.75 AAE/g respectively) (p < 0.05), with no significant difference between peels. Choo et al. (2016) indicated that the ferric reducing capability of dragon fruit was rather weak as the antioxidant compounds in this fruit had stronger antiradical capability than metal reducing ability. In addition, Nurliyana et al. (2010) reported that the ferric reducing capabilities of dragon The data is shown as mean ± standard deviation (n = 3); a,b indicate the means in a row with significant difference (p < 0.05) using one-way analysis of variance (ANOVA) and Tukey's test. DWP, white dragon fruit pulp; DWL, white dragon fruit peel; DRP, red dragon fruit pulp; DRL, red dragon fruit peel; GAE, gallic acid equivalents; QE, quercetin equivalents; CE, catechin equivalents; AAE, ascorbic acid equivalents. fruit peels are stronger than that of pulps, which is contrary to our results, and again may be due to either differences in drying methods or strain variation. The ABTS assay is another widely used method for antiradical capability assessment based on hydrogen atom donation tendency of phenolic compounds. From the ABTS results, pulp samples DWP and DRP has significantly higher value (0.31 ± 0.01 and 0.29 ± 0.01 mg AAE/g respectively) than peel samples DWL and DRL (0.20 ± 0.01 and 0.19 ± 0.01 mg AAE/g respectively) (p < 0.05). The ABTS value of DWP is significantly higher than that of the DRP (p < 0.05), while no significant difference was found between peel samples (p > 0.05). As for former studies, Wu et al. (2006) measured the antiradical capability of dragon fruit peel and pulp by ABTS assay and concluded that the peel extract had better free radical scavenging ability than the pulp extract, which is not consistent with our results. They did however find that the increase of antiradical capability of pulp and peel is positively correlated with the increase in overall antioxidant capacity, which is consistent with our results.
TAC is often used for the determination of total antioxidant capacity of liquid food extracts based on electron transfer mechanism. In this assay, molybdenum (VI) is reduced to molybdenum (V) in the presence of antioxidant compounds (phenolic compounds). The results of TAC indicate that pulp samples DWP and DRP have significantly higher activity (0.32 ± 0.02 and 0.30 ± 0.01 mg AAE/g respectively) than peel samples DWL and DRL (0.19 ± 0.01 and 0.17 ± 0.01 mg AAE/g respectively) (p < 0.05), while there was no significant difference in the TAC results between both peel samples or both pulp samples (p > 0.05). Previously, Abd Manan et al.
(2019) determined the total antioxidant capacity in red dragon fruit pulp by phosphomolybdate assay and indicated that the total antioxidant capacity of this fruit was positively affected by the phenolic content.

LC-ESI-QTOF-MS/MS characterization of phenolic compounds from dragon fruit
In our study, a qualitative analysis of the phenolic compounds from dragon fruit extracts has been conducted using LC-ESI-QTOF-MS/MS in negative and positive ionization modes (Supplementary Materials). Table 2 shows the compounds that were putatively identified in dragon fruit peels and pulps based on their m/z value and MS spectral data using Agilent MassHunter data acquisition software and Personal Compound Database and Library (PCDL) with database of the Kansas State University, USA. Compounds with scores of higher than 80 (PCDL Score) and mass error < ± 5 ppm were selected for m/z verification and MS/MS identification purposes.
In total, 80 different phenolic compounds were tentatively characterized in dragon fruit, which includes 25 phenolic acids, 38 flavonoids, 6 lignans, 3 stilbenes and 8 other polyphenols mentioned in Table 2.

Phenolic acids
Phenolic acids are one of the major classes of phenolic compounds identified in dragon fruit (Garcı´a-Cruz et al., 2017). In our study, four subgroups of phenolic acids were detected in dragon fruit samples, including hydroxybenzoic acid deriva-tives, hydroxycinnamic acid derivatives, hydroxyphenylacetic acids and hydroxyphenylpropanoic acid derivatives. Most of the compounds were identified as hydroxybenzoic acids and hydroxycinnamic acids.

Hydroxybenzoic acids derivatives
Hydroxybenzoic acids are commonly found in red fruits with antioxidant potential such as strawberries and raspberries (El Gharras, 2009). In our study, eight hydroxybenzoic acid derivatives were putatively identified in four dragon fruit samples.
Compound 1 with [M-H]m/z at 169.0138 was detected from DWP, DRL and DRP, and tentatively characterized as gallic acid based on the product ion at 125 m/z, due to the loss of CO 2 (44 Da) from the precursor ion (Escobar-Avello et al., 2019). Previously, Kim et al. had also tentatively identified gallic acid from white and red dragon fruit peel and pulp samples .
Compound 2, 3, 4, 5 and 6 were only detected in DRL and putatively identified as galloyl glucose, 2-hydroxybenzoic acid, 4-hydroxybenzoic acid 4-O-glucoside, 4-O-methylgallic acid and protocatechuic acid 4-O-glucoside according to the precursor ions [MÀH] À at m/z 331.0655, 137.0246, 299.076 and 315.0717 for compounds 2, 3, 4 and 6, and the precursor ion [M+H] + at m/z 185.0444 for compound 5, respectively. The identification of galloyl glucose was confirmed by the product ions at m/z 169 and 125, formed by the neutral loss of a glucose moiety and further loss of CO 2 from the parent ion (Rajauria et al., 2016). The identification of 2-hydroxybenzoic acid was further confirmed by the product ion at m/z 93, formed by the neutral loss of a CO 2 (44 Da) from the parent ion (Escobar-Avello et al., 2019). In the MS 2 experiment of 4hydroxybenzoic acid 4-O-glucoside and protocatechuic acid 4-O-glucoside, the spectra displayed the product ions at m/z 137 and m/z 153 respectively, corresponding to the loss of hexosyl moiety (162 Da) from the precursor ions (Escobar-Avello et al., 2019). Previously, Zain et al. had also tentatively identified protocatechuic in red dragon fruit peels (Zain et al., 2019). Besides, the MS 2 spectrum of 4-O-methylgallic acid displayed the product ions at m/z 170 and m/z 142, indicating the loss of CH 3 (15 Da) and CH 3 CO (43 Da) .
Paeoniflorin (Compound 7) was detected in both negative (ESIÀ) and positive (ESI+) modes in DWP and DRL with an observed [MÀH] À m/z at 479.1558. In the MS 2 spectrum of paeoniflorin, the product ions at m/z 449, 357 and 327 were due to the loss of CH 2 O (30 Da), C 7 H 6 O 2 (122 Da) and CH 2 O plus C 7 H 6 O 2 (152 Da) from the parent ion respectively, which was comparable with the fragmentation rules of paeoniflorin (Wang et al., 2017b). Although paeoniflorin was reported to be abundant in Chinese herbal plants such as Paeonia lactiflora with strong anti-inflammatory and immunomodulatory effects, this compound was tentatively identified in dragon fruit for the first time in the present study to our best knowledge (He and Dai, 2011).

Hydroxycinnamic Acids, hydroxyphenylpropanoic acids and other derivatives
According to previous study, hydroxycinnamic acids are more common than hydroxybenzoic acids in fruits (El Gharras,   Identification of phenolic compounds in Australian grown dragon fruits 2009). This is in consistent with our present study, which detected more hydroxycinnamic acid derivatives (14) as compared to hydroxybenzoic acid derivatives (08). Besides, one hydroxyphenylacetic acid and two hydroxyphenylpropanoic acids were also tentatively identified in our study.

Identification of phenolic compounds in Australian grown dragon fruits
Compound 9 was tentatively characterized as 3-pcoumaroylquinic acid found in DWL, DWP, DRL and DRP in both negative and positive modes with an observed [MÀH]m/z at 337.0932. The identification was further supported by the MS 2 spectrum, which exhibited typical product ions at m/z 265, 173, 162 and 127, formed by the neutral loss of four H 2 O, C 9 H 7 O 3 , C 7 H 11 O 5 and HCOOH-C 9 H 7 O 3 from precursor ion respectively (Lin et al., 2019).
Compound 10 . Previously, caffeoyl glucose and caffeic acid derivatives were tentatively identified in fruits such as berries and plums, but these compounds were identified in dragon fruit for the first time to our best knowledge (Fang et al., 2002;Patras et al., 2018).
Compound 11, 16, 19, 22 were putatively identified in peel samples DWL and DRL. Compound 11 was putatively characterized as 3-caffeoylquinic acid found in DWL and DRL in both negative and positive modes with an observed [MÀH]m/z at 353.0873. With the MS 2 spectrum, the identification was further supported by typical product ions at m/z 253, 190 and 144, formed by the neutral loss of three H 2 O (18 Da) and HCOOH (82 Da); three H 2 O (54 Da) and C 6 H 5 O 2 (109 Da); H 2 O (18 Da) and C 7 H 11 O 6 (191 Da), respectively (Lin et al., 2019). The characterization of 3caffeoylquinic acid is in consistency with previous study of Castro-Enrı´quez et al., which also identified caffeoylquinic acid in dragon fruit (Castro-Enrı´quez et al., 2020 (Lin et al., 2019). Compound 22 was also tentatively identified in DWL and DRL, and tentatively characterized as 3-sinapoylquinic acid based on [MÀH]m/z at 397.1135. In the MS 2 spectrum, the product ions at m/z 223 and m/z 179 indicating the presence of sinapic acid ion and the further loss of COO respectively (Lin and Harnly, 2008).
Cinnamic acid (Compound 20) was detected in DWL, DWP and DRP in negative and positive modes and observed [MÀH]m/z at 147.0454. The compound was confirmed by the product ion at m/z 103, due to neutral loss of CO 2 (44 Da) (Lai et al., 2015). The result of our study is inconsistent with that of Zain et al. (2019), who putatively identified cinnamic acid only in red dragon fruit peel by UHPLC-ESI-QTRAP/MS/MS. This difference is probably related to variation in plant variety.
Two hydroxyphenylpropanoic acids were also detected, which were compounds 24 and 25. Compound 24 was tentatively identified as dihydrocaffeic acid 3-O-glucuronide with [MÀH]m/z at 357.0833, and further confirmed with product ions at m/z 181 due to neutral loss of glucuronide from precursor ion (Sasot et al., 2017). Similarly, compound 25 was tentatively identified as dihydroferulic acid 4-O-glucuronide with [MÀH]m/z at 371.0995, and further confirmed with product ion at m/z 175 due to neutral loss of glucuronide from precursor ion (Sasot et al., 2017).

Flavonoids
Flavonoids were previously identified as the major group of phenolic compounds in dragon fruit (Garcı´a-Cruz et al., 2017). The largest number of compounds detected in the dragon fruit samples were from this phenolic class. Eight subgroups of flavonoids were identified, including anthocyanins, dihydrochalcones, dihydroflavonols, flavanols, flavanones, flavones, flavonols and isoflavonoids. Most of the flavonoids detected were in the glycoside forms.

Anthocyanins derivatives
Anthocyanins are a main subclass of flavonoids, which are known to be abundant in red dragon fruit peel and have anti-inflammation and anticarcinogenic potential (Prabowo et al., 2019). In our study, compound 27 with [M+H] + m/z at 521.1295 was only detected from pulp sample DWP, and characterized as petunidin 3-O-(6 0 '-acetyl-glucoside) based on the product ion at 317 m/z, corresponding to the loss of glucose moiety (162 Da) plus acetyl moiety (42 Da) from precursor ion (Tourino et al., 2008).
In DWL, DRL and DRP, compound 28 was detected in both modes with an observed [M+H] + m/z at 465.1033 and exhibited characteristic fragment ion at m/z 303 [M+HÀglucoside], which was tentatively identified as delphinidin 3-Oglucoside (Tourino et al., 2008). Compound 32 was putatively characterized as cyanidin 3,5-O-diglucoside found in DWL, DWP and DRL based on the observed [M+H] + m/z at 611.1612. The identification was further supported by the MS 2 spectrum, which exhibited typical product ions at m/z 449 and 287, formed by the successive loss of two glucosides (Dincheva et al., 2013). Previously, cyanidin derivatives were reported to be identified in white dragon fruit peels by Vargas, Cortez, Duch, Lizama, and Me´ndez (Vargas et al., 2013).

Dihydrochalcones, dihydroflavonols and flavanols derivatives
Dihydrochalcones, dihydroflavonols and flavanols derivatives are widely present in plants, and were reported to possess diverse biological activities including antioxidant, antiinflammatory and antimicrobial effects, which were important and beneficial for plants as stress-resistant agents (Wen et al., 2014). In our study, only one dihydrochalcones was identified, which was compound 34. It was identified as phloridzin in DWL, DWP, DRL and DRP based on the observed precursor ion [MÀH]at m/z 435.1303, with product ion at m/z 273 representing the existence of phloretin aglycon (Kelebek et al., 2017). Prodelphinidin dimer B3 (Compound 37) was a flavanol derivative found in red dragon fruit samples DRL and DRP. It was tentatively identified with a [M+H] + m/z at 611.1363, which yielded product ion at m/z 469 (formed by heterocyclic ring fission followed by removal of phloroglucinol), m/z 311 (formed by the breakdown of dimer into monomer via quinone methide fission cleavage) and m/z 291 (formed by the formation of catechin from gallo-catechin molecule by loss of OH group).

Flavanones derivatives
Flavanones derivatives are flavonoids that possess antioxidant potential, and were identified in fruits such as citrus with the function of imparting bitter taste (Tripoli et al., 2007). Five flavanones derivatives were putatively characterized in the present study.
In pulp samples, hesperidin (Compound 39 with [M+H] + ion at m/z 611.1992) present in DWP and DRP was identified and confirmed by MS 2 experiments. In the MS 2 spectrum of m/ z 611.1992, the product ions at m/z 593, 465, 449 and 303 were due to the loss of H 2 O (18 Da), rhamnose (146 Da), glucose (162 Da) and rhamnosylglucose (308 Da) from the parent ion (Zheng et al., 2013).
In  (Yu et al., 2020). Previously, flavanones were found to be abundant in citrus fruits, however, this is the first time for these flavanones derivatives to be identified in dragon fruit through LC-MS/MS to our best knowledge (Kawaii et al., 1999).

Flavones and flavonols derivatives
Flavones and flavonols are the most widely distributed antioxidant flavonoids in plants (Hoda et al., 2019).
In the present study, only compound 44 was identified in both dragon fruit peel and pulp samples DWL, DWP and DRL in both modes. Compound 44 was tentatively characterized as apigenin 6,8-di-C-glucoside based on the observed [M -H]at m/z 593.1531. The MS/MS fragmentation yielded the product ions at m/z 575, 503, 473, exhibiting the fragment pattern of apigenin 6,8-di-C-glucoside (Hussain et al., 2018). Previously, Zain et al. has also reported tentative identification of apigenin derivatives in red dragon fruit peel samples (Zain et al., 2019), while it is the first time to identify this compound in dragon fruit pulp sample.
Compounds 45 (Sekuła and Zuba, 2013). Myricetin derivatives were tentatively identified in peel and pulp samples of white and red dragon fruits by Kim et al. (Kim et al., 2011). Zain et al. also reported myricetin derivatives as well as isorhamnetin derivatives in red dragon fruit peel samples (Zain et al., 2019). Moreover, Lira et al. (2020) also tentatively characterized isorhamnetin derivatives and quercetin-3-O derivatives in red dragon fruit pulp and peel samples. In addition, Yi et al. reported to identify kaempferol-3-O derivatives in red dragon fruit pulp, which was in consistent with our study (Yi et al., 2012).

Isoflavonoid derivatives
Isoflavonoids are heterocyclic phenolic compounds that are present in plants with strong antioxidant potential and important pharmacological activities such as anti-diabetic, anticancer and anti-inflammatory (Raju et al., 2015).
In our study, compounds 61 and 62 were detected in both peel and pulp samples. Compound 61 was putatively characterized as 5,6,7,3 0 ,4 0 -pentahydroxyisoflavone found in DWL, DWP and DRL with an observed [M+H] + m/z at 303.0504. With the MS 2 spectrum, the identification was further supported by typical product ions at m/z 285 and 257, formed by the neutral loss of three H 2 O (18 Da) and H 2 O plus CO (46 Da) respectively (Zain et al., 2019) (He and Dai, 2011).
In peel samples, compound 60 with [M+H] + m/z at 287.055 was only detected from DWL, and characterized as 3 0 -hydroxygenistein based on the product ions at m/z 269 and 259, corresponding to the loss of H 2 O (18 Da) and CO (28 Da) from precursor ion . Although isoflavonoids were widely identified in plants, to our best knowledge, most of the isoflavonoids derivatives characterized were the first time detected in dragon fruits (Barnes et al., 2002).

Lignans and stilbenes
Lignans and stilbenes are commonly present in vegetables and fruits (Cassidy et al., 2000). These compounds can act as phytoestrogens as they have both hormonal and non-hormonal activities in animals (Cassidy et al., 2000). Stilbenes also have antibacterial capability that is essential for plant inducible defense system, but also possess antioxidant potential that benefits human health (Chong et al., 2009). Lignans also have strong antioxidant capabilities with high medicinal value (Cassidy et al., 2000).
Matairesinol (Compound 69 with [MÀH]m/z at 357.1349) was identified in DWL and DRL with the product ions at m/z 342 (MÀHÀ15), 327 (MÀHÀ30), 313 (MÀHÀ44) and 221 (MÀHÀ136), representing the loss of CH 3 , C 2 H 6 , CO 2 and C 8 H 8 O 2 from the parent ion respectively (Wen et al., 2014). Six other lignans were also identified in our study. Lignans were previously found in the Leguminosae, which also have strong antioxidant capability (Cassidy et al., 2000). To our best knowledge, the lignans identified in our study were the first time detected by LC-MS/MS in dragon fruits.

Other polyphenols
Some other phenolic compounds identified from dragon fruit samples could not be categorized in the earlier identified classes.
Compound 75 with [M+H] + m/z at 247.0607 was only detected from DRP, and characterized as isopimpinellin based on the product ions at m/z 232, 217, 205 and 203, corresponding to loss of CH 3 (15 Da), two CH 3 (30 Da), CO-CH 2 (42 Da) and CO 2 (44 Da) from the precursor ion (Esquivel et al., 2007). To our best knowledge, isopimpinellin was identified for the first time in dragon fruit though it was previously identified in other fruit such as citrus (Peroutka et al., 2007).
Dragon fruit contain a wide range of phenolics compounds and is therefore a good source of both individual and mixtures of phenolics that may be utilized in food, feed, cosmetics and medicinal industries.

Distribution of phenolic compounds -Venn diagram
The Venn diagrams summarizes the distribution of phenolic compounds in dragon fruit varieties and the difference between peel and pulp ( Fig. 1). A total of 315 phenolic compounds were identified in dragon fruit samples.
Venn diagram A shows that 200 phenolic compounds were identified in both varieties, while white and red dragon fruits had equivalent amounts (57 and 58 respectively) of exclusive compounds, which showed that there is no significant difference in the quantity of phenolic compounds present in each of the two varieties. Previously, Sekar et al. (2016) reported higher antioxidant activity in red dragon fruit than in white dragon fruit extract. We found that although the number of phenolic compounds are equivalent for the two varieties, red dragon fruit have higher total levels of polyphenols compared to the white variety, resulting in higher antioxidant activities.
Venn diagram B shows that dragon fruit peel and pulp shared 140 common phenolic compounds. However, the peel has more exclusive compounds (138 phenolic compounds) than pulp (37 phenolic compounds), indicating that dragon fruit peel might be a better source for extracting phenolic compounds than dragon fruit pulp. Previously, Kim et al. (2011) found higher quantities of phenolic compounds in dragon fruit peels than in pulps through an HPLC-tandem MS analysis, which is in consistent with our results from HPLC-PDA quantification. The higher amounts of phenolic compounds in dragon fruit peel is consistent with Morais et al. (2015), who suggested that the peel of tropical fruits usually have higher amounts of phenolic compounds than their respective pulps.

Heatmap and hierarchical cluster analysis of quantified phenolic compounds in dragon fruit
A heat map was constructed along with hierarchical clusters for further analyzing HPLC-PDA quantified phenolic compounds in dragon fruits Fig. 2. Correlation was used as the dis-tance measure for determining the similarity between dragon fruit samples and compounds. For columns and rows, clustering method was used based on average. For tree ordering, tightest clusters were grouped first.
In the heat map, four clusters in rows and two clusters in columns were generated and highlighted by the hierarchical clustering, which indicated the differences and similarities in phenolic profiles among samples. The color difference showed the concentrations of flavonoids and phenolic acids in different fruit peels. From the results, two clusters of samples were generated and highlighted by the hierarchical clustering, which were DS-1 (including DWP and DRP) and DS-2 (including DWL and DRL). These two clusters indicated significant differences in phenolic profiles between dragon fruit peel and pulp. The color difference showed higher abundance of phenolic compounds in dragon fruit peels than in the pulp samples. This result agreed with the previous study of Kim et al. (2011), who reported higher phenolic contents and stronger antioxidant activities in red and white dragon fruit peels than pulp extracts. Some compounds with significant high concentrations in a certain sample are highlighted by the red color, including quercetin-3-galactoside in DRL as well as epicatechin derivatives, ferulic acid, diosmin and kaempferol in DWL. A comparative study of Sekar et al. (2016) suggested that red dragon fruit extract have higher antioxidant activities than the white variety. However, from our heat map result, DWP and DWL showed more red zones than DRP and DRL, respectively, indicating higher phenolic content in the white variety, which differs from the previously published result. The differences might be attributed to the difference in varieties and maturity of the dragon fruit (Hoda et al., 2019).
Selected phenolic compounds were grouped into four clusters (CP 1-4) and were further grouped into different subclusters according to the differences of their concentration patterns in the dendrogram. Two phenolic acids (phydroxybenzoic acid and coumaric acid) formed the cluster CP-1, both of which showed the highest concentration in DWP and the lowest in DRL. Protocatechuic acid and caftaric acid made their own clusters (CP-2 and CP-3, respectively), while six other phenolic acids, ten flavonoids and two stilbenes formed the cluster CP-4, and were further grouped into different sub-clusters according to the similarity of their concentration pattern among the four samples.

Correlation between phenolic compounds; targeted phenolics quantified through HPLC-PDA and antioxidant assays
Correlations between phenolic contents (TPC, TFC, TTC, phenolic acids and flavonoids-quantified through HPLC-PDA) and antioxidant activities (DPPH, FRAP, ABTS, and TAC) were performed with a Pearson's correlation test (Table 3). The phenolic acid content and flavonoid content were calculated by summarizing the content of ten selected phenolic acids and ten flavonoids, as an estimate for correlation between overall phenolics and their antioxidant activities.
A strong positive correlation between total phenolic content and FRAP was observed, with a Pearson's correlation coefficient r = 0.982 (p < 0.01). The correlation of FRAP with TPC showed that the reducing capability of dragon fruit is mainly attributed to the phenolic contents of the extracts. This result is in agreement with Mokrani and Madani (2016).
The TAC was observed to be strongly correlated with ABTS (r = 0.999, p < 0.01). ABTS determines the hydrogen donation and chain-breaking capabilities of antioxidants by scavenging ABTS radicals. TAC estimates the total antioxidant activity of a sample by reducing phosphomolybdate ions. The correlation indicates that the antioxidants with strong hydrogen donation capabilities that scavenge ABTS radicals can also effectively reduce phosphomolybdate ion and are the major contributors to the total antioxidant capacity of dragon fruit. The results agree with Farkas and Moha´csi-Farkas (2011), in which they reported a good correlation between ABTS and TAC. However, the DPPH activity, which also determines the antiradical capability of antioxidant, is not sig-nificantly correlated with TAC in this study. The reason might be that the ABTS assay was reported to be more effective than the DPPH assay when the food sample contains lipophilic, hydrophilic, and high-pigmented antioxidant compounds (Floegel et al., 2011).
Significant negative correlations were observed between total flavonoid content with ABTS and TAC (r = À0.957 and r = À0.953, p < 0.01). The result is similar to the study of Fidrianny et al. (2014), who reported a negative correlation between TFC and overall antioxidant capability. The TFC assay only targets specific flavonoids including flavonols and flavone luteolin (Pe z kal and Pyrzynska, 2014). Previously, Mokrani and Madani (2016) reported a strong negative correlation between TFC and antiradical capability in peach samples. They concluded that the negative correlation showed the antioxidant capacity of peach might come from the synergism of different polyphenols or other antioxidant compounds present in the extract rather than flavonoids. In our study, the negative correlation indicates that the overall antioxidant capacity and the antiradical capacity of dragon fruit are not caused by the presence of flavonoids, it can be postulated that the main compounds contribute to the antioxidant capabilities might be other phenolic compounds such as phenolic acids or non-phenolic compounds such as betalains.
In our study, no significant difference was observed between phenolic acids and DPPH, FRAP and ABTS. The result was contradictory with the correlation results between the TPC value and FRAP. Besides, these is no significant correlation found between flavonoids and antioxidant assays, which was contradictory with the correlation results between the TFC value and ABTS or TAC. The reasons might be that only 10 of the most abundant phenolic acids and 10 most abundant flavonoids were selected for quantification purposes, while TPC and TFC assays specifically react with all types of phenolic acids and flavonoids respectively.

Conclusion
In conclusion, dragon fruit pulp was found to have higher content of phenolic compounds and stronger antioxidant activities than dragon fruit peel. The LC-ESI-QTOF-MS/MS technique was successfully applied for separation and characterization of the phenolic compounds in dragon fruits, with 80 phenolic compounds tentatively identified in total. The quantification by HPLC-PDA showed that dragon fruit peel has higher levels of most of the selected phenolic compounds, while the pattern of phenolic composition is different between pulps and peels. The obtained results indicated that Australian dragon fruit peel by-products and pulp waste are potential sources of phenolic compounds, with potential as antioxidants for the food, cosmetic, pharmaceutical and nutraceutical industries. Supplementary Materials: