A high-throughput analytical strategy based on QuEChERS-dSPE/ HPLC – DAD – ESI-MS n to establish the phenolic profile of tropical fruits

Tropical fruits are a rich source of phenolic compounds which are favorable in defending the human body against damage induced by free radicals ( e.g. , ROS, RNS). In the current work, a high throughput analytical approach based on a simple extraction procedure (QuEChERS-dSPE) combined with high-performance liquid chromatography-diode array detector-electrospray ionization-mass spectrometry (HPLC-DAD-ESI-MS n ) was used to establish the phenolic profile of tropical fruits. The proposed method showed good linearity (r 2 ≥ 0.991), precision (RSD < 8 %), as well as low limits of detection (LOD ≤ 19.7 μ g/L) and quantification (LOQ ≤ 65.8 μ g/ L). Thirty-four phenolic compounds were identified as belonging to different chemical groups, from which only 6 were common to all tropical fruits. Pitanga showed the highest relative phenolic concentration (99.5 mg/100 g of fruit), with the passion fruit (17.5 mg/100 g of fruit) the lowest. Flavonols were the most predominant chemical group in tropical fruits, representing 77.9, 60.1, and 55.8% of the phenolic composition of pitanga, passion fruit and mango, respectively. The data obtained allow deep and comprehensive insights into the phenolic composition of tropical fruits in order to explore its potential bioactive activity. Nevertheless, in vivo assays using fruit extracts will be essential to recognize their potential health-promoting properties.

Specific attention has been paid to the extraction of phenolic compounds from fruits, which represents a critical step in the establishment of the phenolic profile of fruits.Among the extraction procedure, organic solvents (e.g., acetonitrile, ethyl acetate, methanol) and/or solid-phase extraction (SPE) with reversed-phase C 18 are the most commonly used (Barnes et al., 2020;Celli et al., 2011;Russo et al., 2018).Nevertheless, these extraction procedures are solvent-and time-consuming and involve extra steps (e.g., clean-up, solvent evaporation) due to chromatographic incompatibility and/or sample concentration (Rotta et al., 2019).More recently, a quick, easy, cheap, effective, rugged and safe (QuEChERS) method with C 18 as dispersive cleaning sorbent followed by ultra-pressure liquid chromatographic tandem mass spectrometry (UHPLC-MS/MS) has been proposed to identify and quantify nine phenolic compounds in passion fruit pulp (Passiflora spp.).This extraction procedure comprises two steps: (i) an extraction step based on partitioning via salting-out extraction where an equilibrium between an aqueous and an organic layer is promoted, and (ii) a dispersive solid-phase extraction (d-SPE) step that includes a clean-up process applying numerous mixtures of porous sorbents and salts to eliminate matrix interfering constituents (Perestrelo et al., 2019).On the other hand, regarding phenolic compound identification and quantification, several analytical platforms have been purposed such as spectrophotometry (Vasco et al., 2008) and/or high-performance liquid chromatography (HPLC) coupled to ultraviolet (UV) or diode array (DAD) detectors (Dorta et al., 2014;Garmus et al., 2014).Nevertheless, these analytical platforms showed limitations such as coelutions, similar UV-absorption spectra and require standards reference to provide target identification.High-performance liquid chromatography-diode array detector-electrospray ionization-mass spectrometry (HPLC-DAD-ESI-MS n ) appears to be the most suitable analytical platform to establish the phenolic profile of fruits, since it provides useful structural information and allows a tentative target identification when the reference standards are commercially unavailable (Celli et al., 2011;Russo et al., 2018;Spínola et al., 2015;Vu et al., 2019).
Most of the published investigations related to tropical fruits only provide a screening of the phenolic profile and/or reported the total phenolic compounds (TPC), tannins and flavonoids through colorimetric assays (e.g., Folin-Ciocalteu), as well as antioxidant activity through in vitro chemical tests (Hu et al., 2018;Sobeh et al., 2020;Spínola et al., 2015).As far as we know, QuEChERS-dSPE/HPLC-DAD-ESI-MS n is scarcely used to establish the phenolic profile of tropical fruits (Rotta et al., 2019).Therefore, the main goal of this research is to validate a high-throughput analytical approach based on HPLC-DAD-ESI-MS n combined with QuEChERS-dSPE to separate, identify and semi-quantify the phenolic compounds of passion fruit (Passiflora edulis L.), pitanga (Eugenia uniflora L.) and mango (Mangifera indica L.) habitually part of Madeiran diet.Considering the scarce application of QuEChERS-dSPE for the extraction of phenolic compounds from fruits, the current research represents an added value and improved alternative to the most conventional extraction procedures.

Fruit samples
Fresh samples of passion fruit (Passiflora edulis L.), pitanga (Eugenia uniflora L.) and mango (Mangifera indica L.) were purchased from a local market in Funchal, Portugal (32 • 38 ′ 55 ′′ N, 16 • 54 ′ 14 ′′ W).For each fruit sample, approximately 1 kg was randomly sampled from the market shelves, simulating consumer shopping behavior.Fruits were washed in water and all inedible parts were removed manually or using a steel knife.Bruised and/or wounded fruits were discarded.Passion fruit, pitanga and mango were peeled and only the pulp was analyzed.For each independent analysis, at least 250 g of fruit sample, with 50 mL of Milli-Q water added, were put in a commercial juice extractor (Instant pulp, 200 W, Worten, Portugal), obtaining a fluid fruit extract which was stored at − 20 • C until QuEChERS-dSPE extraction procedure.

QuEChERS procedure for extraction of phenolic compounds
A homogenized fluid fruit sample (5 ± 0.1 g) was weighed into a mL PTFE centrifuge tube, followed by addition of 5 mL of MeCN.Then, the tube was shaken vigorously for 2 min with vortex mixer ensuring that the solvent interacted well with the entire sample.Buffered salts, C 6 H 5 Na 3 O 7 .2H 2 O (0.5 g), C 6 H 8 Na 2 O 8 (0.25 g), NaCl (0.5 g) and MgSO (2 g) were added into the homogenized mixture and the shaking step was repeated for 1 min followed by centrifugation at 5000 rpm for 3 min, Fig. 1.
For the dSPE step, an aliquot of the MeCN phase was transferred into a 2 mL PTFE single-use centrifugation tube already containing 25 mg of PSA (removes various polar organic acids, polar pigments, some sugars and fatty acids), 25 mg of C 18 sorbent (removes non-polar interfering substances like lipids) and 150 mg MgSO 4 .The mixture was shaken in a vortex and centrifuged for 2 min at 3000 rpm.Then, a 700 μL aliquot of the extract was evaporated under nitrogen (N 2 ) flow to dryness and the residue was dissolved in 100 μL of mobile phase A. All the samples were filtered through a 0.22 μm Millipore PTFE filter membrane prior to HPLC-DAD-ESI-MS n analysis.

Method validation
The proposed method was validated based on the linearity, sensitivity, precision and selectivity.As not all the phenolic compounds identified in fruits are commercially available and using a regularly adopted approach (Perestrelo et al., 2012;Spínola et al., 2015), six representative standards of the phenolic compounds under study were chosen.These phenolic standards were used to construct the calibration curves and the results for each target phenolic compound were expressed in equivalents of the respective standard.For each of the six standards, an ethanolic stock solution was prepared (200 mg/L).All solutions were stored at − 20 • C in the dark.Then, seven different concentrations, covering the concentration range predictable for each phenolic compound (Table 1), were prepared by diluting suitable amounts of each stock solution in mobile phase A. Each one of these solutions was analyzed in triplicate, using the QuEChERS-d-SPE/HPLC-DAD-ESI-MS n method.
The method sensitivity was measured based on limit of detection (LOD) and limit of quantification (LOQ).The LOD and LOQ were determined through the multiplication by 3 and 10 of the ratios of standard deviation(s) of calibration curve interception and the slope of the regression curve, respectively.
The accuracy of the method was assessed by spiking pitanga sample in triplicate at three concentration levels (low (LL), middle (ML) and high (HL)) and subjecting them to QuEChERS-dSPE procedure.The repeatability of the method was evaluated by analyzing three replicates of standard solutions (5 mg/L) under the same conditions on the same day (intra-day) and over a period of one week (inter-day).The results obtained were expressed as the relative standard deviation (% RSD).
The selectivity was assessed by the absence of interfering peaks at the analyte retention time (RT).To demonstrate the nonexistence of any carryover during injection, a pure solvent was injected directly using the highest calibration point of each phenolic standard.

Phenolic profile by HPLC-DAD-ESI-MS n
The qualitative and semi-quantitative analysis of the phenolic compounds was carried out on a HPLC system of Dionex ultimate 3000 series (Sunnyvale, CA) instrument equipped with a binary pump, diode array detector (DAD), autosampler and column compartment according to the method previously described (Perestrelo et al., 2012).The equipment was equipped with an Atlantis dC18 column (250 mm × 4.6 mm i.d.× 5 μm) supplied from Waters (Milford, Ma, USA) at controlled temperature (25 • C).The elution was performed using mobile phase A (0.1% FA in aqueous solution) and mobile phase B (0.1% FA in acetonitrile).The flow rate was 300 μL/min.The gradient program was used as follows: 0− 3 min, 100 % A; 3− 10 min, 100− 90% A; 10− 30 min, 90− 80% A; 30− 35 min, 80− 75% A; 35− 50 min, 75-100% A. The fruit extract obtained was dissolved in the initial HPLC mobile phase A and were filtered through 0.22 μm micropore membranes prior to injection into the HPLC system (injection volume of 10 μL).The detection by DAD was carried out by scanning 210-520 nm, with a resolution of 1.2 nm, and the semi-quantification was performed at 280, 320 and 360 nm for phenolic acids, stilbenes and flavonols, respectively.
For identification purposes, mass spectrometry analysis was performed using a Bruker Esquire model 6000 ion trap mass spectrometer (Bruker Daltonics, Bremen, Germany) equipped with an electrospray ionization (ESI) source.Data acquisition and processing were performed using Esquire control software.The mass spectrometer was operated in the negative mode, and the mass range from 50 to 3000 m/z, under the following conditions: nebulizer gas pressure of 50 psi, drying gas flow of 10 mL/min, desolvation temperature of 350 • C, cone voltage between 30 and 50 V, collision energy set between 10 and 45 V, and the capillary voltage ranged from 2.6 to 2.9 kV.

Phenolic identification by HPLC-DAD-ESI-MS n
A total of 34 phenolic compounds were tentatively identified in tropical fruits that included 5 hydroxybenzoic acids, 6 hydroxycinnamic acids, 15 flavonols, 3 stilbenes and 5 others.The peak assignment of phenolic compounds extracted from tropical fruits was carried out by comparison of their retention time (RT) and MS n fragmentation profiles with reference standards, analyzed under the same experimental conditions and/or with published data.In general, in the MS 1 spectrum the most intense peak corresponded to the deprotonated molecular ion [M− H] − .The identification of the phenolic compounds detected in tropical fruit extracts is presented in Table 2.The resulting chromatograms of investigated tropical fruits obtained by HPLC-DAD analysis are shown in Fig. 2.

Hydroxybenzoic acid
Compound  were assigned as caffeic acid derivates due to their typical MS 2 fragments (m/z 179 and 135).
Compound 12 with a pseudomolecular ion [M− H] − at m/z 479 produced product ions at m/z 317 [myricetin acid− H] − and minor ions at m/z271, 171 and 159, suggesting a loss of glucose.On the basis of mass fragments, this compound was identified as myricetin glucoside.Compounds 13 and 16 identified as myricetin arabinopyranoside, yielded an MS spectrum containing [M− H] − at m/z 449, which fragmented on MS 2 to produce a myricetin ion at m/z317 due to a loss of an arabinopyranoside moiety, 132 Da.The myricetin-galloyl-deoxyhexose (compound 28) was identified by its [M− H] − ion at m/z 615 and MS 2 fragment ions at m/z 463 and 317 by neutral loss of galloyl (− 152 Da) and deoxyhexose (− 146 Da) moiety.

Flavan-3-ol
Compound 11 was assigned as epigallocatechin gallate presenting a [M− H] − at m/z 457 that released an MS 2 fragment at m/z 331 and 305 originating from epigallocatechin, while the signal at m/z 169 originated from [gallic acid− H] − .

Table 2
Relative concentration of phenolic compounds identified in tropical fruits using QuEChERS-dSPE/HPLC-DAD-ESI-MS n .Values were expressed as mean ± standard deviation of three replicates (n = 3).< LOQ : lower than limit of quantification; -: not detected.a Expressed in equivalents of protocatechuic acid.b Expressed in equivalents of catechin.c Expressed in equivalents of p-coumaric acid.d Expressed in equivalents of rutin.e Expressed in equivalents of trans-resveratrol.
f Expressed in equivalents of kaempferol.
C. Silva et al.

Flavone
Compound 7 with a pseudomolecular ion [M− H] − at m/z 401 produced product ions at m/z 269, suggesting a loss of glucose (− 162 Da).Based on the mass pattern, this compound was identified as apigenin pentoside.

Stilbenes
Piceatannol glucoside (compound   product ion was further fragmented to form m/z 157 and 143 corresponding to the loss of CO (− 28 Da) and C 2 H 2 O (− 42 Da), respectively.

Lignans
Compound 20 has been identified as a possible syringaresinol glucoside, which exhibited a deprotonated molecular ion [M− H] − at m/ z 579, and released MS 2 fragments at m/z 417 and 166.The same fragmentation pattern was observed for syringaresinol.

Method validation and semi-quantification of phenolic compounds in tropical fruits
According to the HPLC-DAD-ESI-MS n analysis, as well as the distribution and structure of phenolic compounds in the chromatogram, the phenolic compounds tentatively identified in tropical fruits are organized into 4 main chemical families, namely hydroxybenzoic acids, hydroxycinnamic acids, flavonols and stilbenes.The tropical fruits revealed dissimilar phenolic profiles based on the phenolic compounds identified and their relative concentration.A total of 25 phenolic compounds were identified in pitanga fruit, whereas in passion fruit and mango 15 and 12 compounds, respectively, were identified.From these, only 6 were identified in all tropical fruits, namely galloyl glucose, sinapic acid glucoside, myricetin arabinopyranoside, quercetin glucoside and caffeic acid derivate.Most of the phenolic compounds identified in tropical fruits were already found in pitanga, passion fruit and mango fruits (Hu et al., 2018;Shanmugam et al., 2018;Sobeh et al., 2020).On the other hand, some phenolic compounds expected to be found in pitanga (e.g., gallic acid, quinic acid, p-coumaryl quinic acid) (Sobeh et al., 2020), passion fruit (e.g., gallic acid, artepellin C, daidzein, narigenin) (Shanmugam et al., 2018) and mango (e.g., gallic acid, ellagic acid, mangiferin and mangiferin gallate) (Hu et al., 2018) were not identified may be the methodology applied in this research was not the most appropriate to identify and quantify trace amounts of phenolic compounds.
Each phenolic compound was relatively quantified using the calibration curves, using a set of 6 reference phenolic compounds chosen according to the principle of structure-related target analyte/standard (functional group and/or chemical structure).Each concentration level was processed following the proposed QuEChERS-dSPE procedure followed by HPLC-DAD-ESI-MS n in triplicate.The important data regarding the method validation is displayed in Table 1.The method exhibited good linearity with a regression coefficient (r 2 ) higher than 0.991.The LOD values ranged from 0.99 to 19.7 μg/L, whereas the LOQ values ranged from 3.30-65.8μg/L.Regarding the accuracy of the QuEChERS-dSPE/HPLC-DAD-ESI-MS n method, satisfactory recovery values, for the three concentration levels (LLlow level, MLmiddle level, HLhigh level), were obtained, ranging from 76 to 119%.The repeatability was assessed in terms of by intra-and inter-days, and the values for both were lower than 8%.The literature has described that a quantitative method should be proved as being capable of showing mean recoveries with the range of 70-120%, and precision with %RSD values lower or equal to 20% (Nantia et al., 2017;Rotta et al., 2019).Thus, the data obtained for repeatability and accuracy indicating the stability and robustness of the proposed method.
The analytical method developed was compared with other liquid chromatography (LC) methods reported in the literature for phenolic compounds quantification in tropical fruits (Barnes et al., 2020;Celli et al., 2011;Hu et al., 2018;Nguyen et al., 2019;Rotta et al., 2019;Shanmugam et al., 2018).It should be pointed out that the proposed method required lower sample amount (5 g) and solvent volumes (5 mL), instead of the higher amount of sample (up to 10 g) and solvent volume (10 mL) applied in QuEChERS-dSPE (Rotta et al., 2019) and other extraction procedures for the determination of phenolic compound in fruits (Celli et al., 2011;Nguyen et al., 2019).Moreover, QuEChERS-dSPE contrarily to other extraction procedures does not require any previous sample treatment (e.g., lyophilization) before the extraction procedure (Celli et al., 2011;Shanmugam et al., 2018).Regarding analytical performance, the LOD, LOQ, precision and recovery obtained with the proposed method were similar to the reference methods (Barnes et al., 2020;Rotta et al., 2019).
Table 2 reports the qualitative and semi-quantitative data related to the tentatively identified phenolic compounds, and it was performed at 280, 320 and 360 nm for the phenolic acids, stilbenes and flavonols, respectively.The relative quantified phenolic compounds accounted for 99.5 mg/100 g of fruit for pitanga, whereas for passion fruit and mango it was 17.5 mg/100 g and 24.0 mg/100 g of fruit, respectively.
Myricetin arabinopyranoside (18.7 mg /100 g of fruit), quercetin glucoside (15.1 mg/100 g of fruit), quercetin pentoside (15.6 mg/100 g of fruit) and quercetin rhamnose (19.2 mg/100 g of fruit) were the most abundant phenolic compounds identified in pitanga, which represent 68.9% of the total phenolic composition.This result is in agreement with a previous study that reported the quantification of phenolic compounds in two varieties of Brazilian cherry (Eugenia uniflora L.) using organic solvent followed by HPLC-MS/MS analysis (Celli et al., 2011).In addition, the phenolic content and antioxidant activity of tropical fruits (mango, passion fruit, range, acerola, longan, rambutan) were previously assessed, and the results showed that among of the investigated tropical fruits, mango presented the highest phenolic content and antioxidant activity (de C. Albuquerque et al., 2019;Nguyen et al., 2019).This is in agreement with the results obtained in this research, where mango showed a higher phenolic content compared to passion fruit.Rotta et al. (2019) determine the phenolic composition in the pulp of three different Passiflora species using a QuEChERS-dSPE/UHPLC-MS/MS method.Differences phenolic profile were observed among the three species, with quercetin, vanillic aid and rutin the most abundant.The total concentration of phenolic compounds in these three species was lower than that determined in the current study (17.5 mg/100 g of fruit).On the other hand, the passion fruit analyzed in this study showed lower content of phenolic compounds compared to that reported by Reis et al. (2018) using an exhaustive extraction with 20 mL of ethanol.
For passion fruit, the most abundant phenolic compounds were myricetin arabinopyranoside and quercetin glucoside (representing 54.4% of the total phenolic composition), whereas for mango it was myricetin arabinopyranoside and quercetin pentoside representing 51.1% of the total phenolic composition.Nevertheless, it should be highlighted that stilbenes such as piceatannol glucoside (< LOQ), piceatannol (0.17 mg/100 g of fruit) and trans-resveratrol (0.06 mg/100 g of fruit) were only detected in passion fruits.Piceatannol similar to its precursor resveratrol showed health-promoting properties such as antiaging, anticarcinogenic, anti-diabetic, anti-inflammatory, antiobesity properties, as well as cardio-, hepato-and neuro-protection, in several pre-clinical studies (Dai et al., 2020;Kershaw and Kim, 2017;Wen et al., 2018;Zhang et al., 2018).
These outcomes endorse that tropical fruits are a dietary source of phenolic compounds, mainly quercetin glucoside and piceatannol, consequently its ingesting can result in health-promoting benefits.

Conclusions
The QuEChERS-dSPE/HPLC-DAD-ESI-MS n method was successfully validated and applied to establish the phenolic profile of tropical fruits from Madeira Island.A total of 34 phenolic compounds were identified in the investigated tropical fruits, including 5 hydroxybenzoic acids, 7
1 was identified as hydroxybenzoyl glucose based on fragmentation pattern, with [M− H] − at m/z 299 showing product ions at m/z 137 (loss of glucose moiety, 162 Da) and m/z 93 (loss of CO 2 ([4hydroxybenzoic acid − CO 2 − H + ] − ).Compound 2 was assigned as galloyl glucose presenting a pseudomolecular ion [M− H] − at m/z 331 that released an MS 2 fragment at m/z 169 ([M − 162] − , loss of a glucose moiety) corresponding to gallic acid (Fig. 3).Compound 3 showed the characteristic fragmentation pathway of an galloyl− HHDP-glucoside with a [M− H] − at 633 and the MS 2 spectrum showed ions at m/z 463 by loss of galloyl + CO 2 unit (170 Da) and m/z301 as main fragment (loss of galloylglucose, 332 Da).Orsellinic acid glucoside (compound 4) exhibited a pseudomolecular [M− H] − ion at m/z 329, yielding MS fragments at m/z 167 corresponding to [orsellinic acid− H] − due to a loss of glucose (− 162 Da).Compound 8 with a pseudomolecular ion [M− H] − at m/z 635 and MS 2 fragment ions at m/z 465 (loss of gallic acid moiety), m/z 313 (loss of a galloyl moiety) and m/z 169 ([gallic acid-− H] − ) was identified as trigalloyl glucose3.1.2.Hydroxycinnamic acidMass spectra of compound 5 displayed a parent ion at m/z 341 and three fragment ions with one at m/z 179 and 135 for caffeic acid through the loss of a glucose and CO 2 moiety.Compound 9 showed a pseudomolecular ion [M− H] − at m/z 325 and MS 2 fragment ions at m/z ([coumaric acid− H] − ) corresponding to a loss of glucose moiety, which allow identification of this compounds as p-coumaric acid glucoside.Compound 10 presented a pseudomolecular ion [M− H] − at m/z 385, yielding MS 2 fragments at m/z 223 (loss of a glucose moiety; [sinapic acid− H] − ), suggesting that it could be a sinapic acid glucoside.Compound 17 was tentatively identified as coumaric acid derivate due to its typical MS 2 fragments (m/z 163 and 119), whereas compounds 6 and 14) was identified based on the precursor [M− H] − ion at m/z 405, and released an MS 2 fragment at m/z 243 corresponds to the deprotonated [piceatannol− H] − by a neutral loss of a glucose moiety.Compound 25 predominated in the phenolic profile and exhibited a pseudomolecular ion [M− H] − at m/z 243 and fragment ions at m/z 225, 201, 175 and 159.This fragmentation pattern is typical for the stilbene piceatannol.Trans-resveratrol (compound 32) identified as resveratrol showed a pseudomolecular ion [M− H] − at m/z 227 and MS 2 fragment ions at m/z 185 (loss of C 2 H 2 O moiety, − 42 Da).This

Table 1
Validation data used for the HPLC-DAD semi-quantification of phenolic compounds in tropical fruits.Maximum wavelength; R 2 -Correlation coefficient; LOD -Limit of detection; LOQ -Limit of quantification.
a Ions in boldface indicate the more abundant m/z ratio.