HPTLC-PCA Complementary to HRMS-PCA in the Case Study of Arbutus unedo Antioxidant Phenolic Profiling

A comparison between High-Performance Thin-Layer Chromatography (HPTLC) analysis and Liquid Chromatography High Resolution Mass Spectrometry (LC–HRMS), coupled with Principal Component Analysis (PCA) was carried out by performing a combined metabolomics study to discriminate Arbutus unedo (A. unedo) plants. For a rapid digital record of A. unedo extracts (leaves, yellow fruit, and red fruit collected in La Maddalena and Sassari, Sardinia), HPTLC was used. Data were then analysed by PCA with the results of the ability of this technique to discriminate samples. Similarly, extracts were acquired by non-targeted LC–HRMS followed by unsupervised PCA, and then by LC–HRMS (MS) to identify secondary metabolites involved in the differentiation of the samples. As a result, we demonstrated that HPTLC may be applied as a simple and reliable untargeted approach to rapidly discriminate extracts based on tissues and/or geographical origins, while LC–HRMS could be used to identify which metabolites are able to discriminate samples.


Introduction
Plant metabolic profiling is a very useful strategy to study the complexity and the large variety of compounds belonging to different chemical classes [1] and is ideally suited to comparing many samples in order to classify them according to botanical, geographical origins and chemotypes [2][3][4][5].
As a consequence, one of the many aims of research in the field of metabolomics is to analyze a large number of samples and obtain information in the shortest times and with a little or no sample preparation time [6,7]. Creation of rapid and convenient methods for simultaneous metabolites fingerprinting and their quantification requires the use of peculiar analytical techniques.
In the last few years, the progress and developments of analysis techniques, including advanced hyphenated techniques (like Liquid Chromatography-Mass Spectrometry (LC-MS) and Gas Chromatography-Mass Spectrometry (GC-MS)), can effectively satisfy this demand [8][9][10].
Until now, most metabolomics studies were performed using the most popular analytical technologies, such as Nuclear Magnetic Resonance, GC, LC, and MS [11]. The new developments reached, employing high chromatographic resolution/separation interfaced with high resolution mass spectrometry, showed the power of this coupled technique in metabolomics [12][13][14][15]. High Performance Thin Layer Chromatography (HPTLC) is gaining more attraction in the field of metabolomics. In fact,

Reagents and Chemicals
HPLC-MS grade methanol, acetonitrile, and formic acid were purchased from Sigma-Aldrich Chemical Company (St Louis, MO, USA). HPLC grade water (18 mΩ) was obtained by using a Milli-Q purification system, Millipore (Bedford, MA, USA). Chemicals and reagents necessary for antioxidant activity assays were supplied by Sigma (Dorset, UK).

Sampling Sites and Extraction
Wild plants of A. unedo were collected in October 2015 in two selected geographical areas of Sardinia: Sassari and La Maddalena. Botanical identity of the plants was assigned by Doctor M. Chessa. Voucher specimens were left at the Erbarium SASSA of Sassari University (number 514).
Leaves and yellow and red fruits of A. unedo from Sassari and La Maddalena were extracted under ultrasound agitation for 1 h, with ethanol/water (70:30 v/v using a sample to solvent ratio 1:10 w/v); then samples were stored in the dark overnight. Samples were then filtered and dried using a rotary evaporator under a vacuum and temperature of 30 • C, working in the dark. For qualitative analysis, dried samples were dissolved again in methanol to generate a solution of 1 mg/mL. The solutions were filtered through 0.20 µm syringe PVDF filters (Whatmann International Ltd., Maidstone, UK).

HPTLC Analyses
The HPTLC analyses were performed following the protocol described by Maldini et al., 2016 [16], with slight modification. Extracts were reconstituted at a concentration of 6 mg/mL and a volume of 6 µL was loaded. The developing solution for HPTLC plates was a mixture of ethyl acetate/dichloromethane/acetic acid/formic acid/water (100:25:10:10:11; v/v/v/v/v). The length of the chromatogram run was 70 mm from the point of application.
For the densitometric analysis, a CAMAG TLC scanner 3 (CAMAG, Muttenz, Switzerland) linked to winCATS software (version 1.2.1, CAMAG, Muttenz, Switzerland). was set at 254 nm and 366 nm, after an optimization performed by a multi-wavelength mode from 220 to 700 nm. A minimum background compensation was performed on the x-axis during the scanning. Deuterium and tungsten lamps were used as sources of radiation. The slit dimension was kept at 6.00 × 0.45 mm and the scanning speed was 100 mm/s.

LC-ESI-Orbitrap-MS Analysis
To characterize the main metabolites representative of each sample, the LC-ESI-Orbitrap-MS method was developed. Each extract was dissolved 1:100 with methanol and a 10 µL aliquot injected into the analytical system. A duplicate of each sample was carried out, obtaining a total of 36 analyzed samples. Experiments were run using a Thermo Scientific liquid chromatography system, equipped with a quaternary Accela 600 pump and an Accela auto sampler, in conjunction with a linear Trap-Orbitrap hybrid mass spectrometer (LTQ-Orbitrap XL, Thermo Fisher Scientific, Bremen, Germany), combining a linear trap quadrupole (LTQ) and an Orbitrap mass analyzer with an electrospray ionization (ESI) source. Chromatographic separation was obtained using an X-Select T3 C18 reversed phase column (2.1 × 150 mm, 3.5 µm particle size) (Waters, Milford, Massachusetts). The mobile phase consisted of solvent A (water + 0.1% formic acid) and solvent B (acetonitrile + 0.1% formic acid). A linear gradient program at a flow rate of 0.200 mL/min was used: 0-3 min, from 0 to 10% (B); 3-25 min, from 10 to 20% (B); 25-35 min, from 20 to 30% (B); 35-40 min, from 30 to 60% (B); 40-45 min, from 60 to 100% (B); then 0% (B) for 5 min. The mass spectrometer was operated in negative ion mode. The ESI source parameters were the following: The capillary voltage −48 V; tube lens voltage −176.47 V; capillary temperature 300 • C; Sheath and Auxiliary Gas flow (N2) 15 and 5 (arbitrary units); Sweep gas 0 (arbitrary units); Spray voltage 3.50 V. MS spectra were acquired by full range acquisition covering m/z 200-1200 (Resolution: 30,000). For MS/MS experiments, a data-dependent scan experiment was established, with the selection of precursor ions corresponding to the most intense peaks observed in the previous LC-MS analysis (threshold value 300).
Compounds were identified on the basis of their spectral characteristic fragmentation and retention time, with comparison with data reported in literature and databases. Data acquisition, data analysis, and instrument control were performed by Xcalibur software version 2.1 (Thermo Scientific™, Waltham, MA, USA)

PCA
Principal component analysis (PCA) is a multivariate data analysis technique used to reveal important patterns correlating to physiological, genetic, and environmental issues and has been used widely in assessing the differences between plant varieties at a metabolomics level [25]. An m × n matrix (where m is the number of samples and n is the number of variables) was used in PCA analysis of data obtained from HPTLC. For matrix building, the variables were taken from the pseudo-chromatogram and reported as the area % corresponding to intensities of the individual retention time factors of the most intense spots in the fingerprint of each sample. PCA was performed on the dataset scaled by unit Similarly, an m × n matrix (where m is the number of samples and n is the number of variables) was used in PCA analysis of data obtained from LC-ESI-Orbitrap-MS. The untargeted approach was obtained working with the base peak chromatograms derived from LC-MS (negative ion mode), which were evaluated using a platform independent open source software package called MZmine (http://mzmine.sourceforge.net/). By means of this toolbox normalization by total raw signal and excluding noise from LC-MS profiles (Noise level 5.0 E3-all data points below this intensity level were ignored), 280 peaks were detected and the peak area was determined [26].
After transferring the processed data in tabular format (cvs file), further analysis of the data matrix (36 observation and 280 variables) were made by SIMCA (+) software 12.0 (Umetrix AB, Umea, Sweden) by PCA. PCA was achieved by measuring the peak areas obtained from LC/MS analysis [27]. Pareto scaling was applied to data before multivariate data analysis.

Antiradical Activity by Diphenyl-1-Picarylhydrazyl (DPPH) and TEAC Assays
The radical scavenging activity assay was performed according to the method proposed by Brand-Williams (1995) [28] with some modifications.
The ABTS free radical scavenging activity of each sample was determined according to the method described by Petretto et al. (2015) [29].

Statistical Analysis
All experiments were carried out in triplicate. Statistical analyses were performed by comparison of ethanolic extracts from red fruits, yellow fruits, and leaves of Arbutus unedo collected in different areas of Sardinia, with unpaired Student's t-test using Sigma-Stat v. 3.5 software (Systat Software GmbH, Erkrath, Germany). The distribution of samples was performed by the Kolmogorov-Smirnov and Shapiro tests. The strength of association between variables was investigated with the Pearson product moment correlation coefficient (data normally distributed). A p ≤ 0.05 was considered statistically significant.

HPTLC-PCA Analysis
The power of HPTLC analysis is to get characteristic fingerprints of secondary metabolites occurring in numerous biological samples (~20) in a single run. Thus, it was here employed to quickly compare A. unedo samples (red fruit, yellow fruit, and leaves) from different places of collection.
By using an optimized separation method, the different pigments are highlighted, using UV light at 254 or 366 nm or under reflectance and transmission white light (WRT Figure 1A). HPTLC metabolomics patterns didn't show qualitative differences among samples collected in Sassari or in La Maddalena ( Figure 1A). Successively, densitometric analysis was carried out at 254 nm and 366 nm. The HPTLC 3D chromatogram recorded at 254 nm in Figure 1B shows the characteristic traces of A. unedo red fruit, yellow fruit, and leaves from La Maddalena (1-9) and Sassari (10)(11)(12)(13)(14)(15)(16)(17)(18). Densitometer scanning at 254 nm wavelength was the most useful to characterize and differentiate extracts; thus, Rf values with a corresponding area for each spot were obtained and designated to produce an m × n matrix made up by 18 observations and 16 variables (metabolites detected). A PCA was conducted on the dataset to provide an overview to discriminate samples on the basis of the tissue and the geographical origin. Figure 2 shows the Score Plot obtained for a studied data set. The first principal

LC-ESI-Orbitrap-MS Analysis and PCA
With the aim to identify which secondary metabolites are responsible with a major impact for the variation of the analyzed samples, metabolites profiling of A. unedo leaves and red and yellow fruit ethanolic extracts from La Maddalena and Sassari were analyzed by liquid chromatography, coupled with high resolution mass spectrometry. Nineteen metabolites were detected and putatively identified (Table 1) according to the information obtained from accurate mass fragmentation spectra and then the obtained results were confirmed with those present in the literature and database. All the identified compounds were already reported in A. unedo. The main detected metabolites belong to flavonoids, principally quercetin, kaempferol, and myricetin derivatives (9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) [31][32][33][34][35]. Quercetin and kaempferol derivatives were both similarly detected in leaves and fruit of A. unedo from Sassari and La Maddalena area, while myricetin derivatives are differently distributed in the samples; in fact, myricetin hexoside (9) was detected only in leaves. Other identified compounds belong to catechin

LC-ESI-Orbitrap-MS Analysis and PCA
With the aim to identify which secondary metabolites are responsible with a major impact for the variation of the analyzed samples, metabolites profiling of A. unedo leaves and red and yellow fruit ethanolic extracts from La Maddalena and Sassari were analyzed by liquid chromatography, coupled with high resolution mass spectrometry. Nineteen metabolites were detected and putatively identified ( Table 1) according to the information obtained from accurate mass fragmentation spectra and then the obtained results were confirmed with those present in the literature and database. All the identified compounds were already reported in A. unedo. The main detected metabolites belong to flavonoids, principally quercetin, kaempferol, and myricetin derivatives (9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) [31][32][33][34][35]. Quercetin and kaempferol derivatives were both similarly detected in leaves and fruit of A. unedo from Sassari and La Maddalena area, while myricetin derivatives are differently distributed in the samples; in fact, myricetin hexoside (9) was detected only in leaves. Other identified compounds belong to catechin and epicatechins (3 and 5) and galloyl organic acid derivative groups (2 and 8) [31] that, in the present study, were found both in fruit and leaves of the analyzed samples. In addition, other secondary metabolites belonging to ellagitannins class (4 and 7) were detected. These were found in all analysed samples, with the exception of yellow fruit of La Maddalena (LMFG) and red fruit of Sassari (SSFR). Finally, chlorogenic acid (6), present in all analysed samples, and arbutin (1), detected in all the samples except in yellow fruit from Sassari, were also identified. Figure 3 shows LC-HRMS and extracted ion chromatograms of compounds 1-19 identified in La Maddalena leaves extract. samples, with the exception of yellow fruit of La Maddalena (LMFG) and red fruit of Sassari (SSFR). Finally, chlorogenic acid (6), present in all analysed samples, and arbutin (1), detected in all the samples except in yellow fruit from Sassari, were also identified. Figure 3 shows LC-HRMS and extracted ion chromatograms of compounds 1-19 identified in La Maddalena leaves extract.  Table 1.
The collected LC-ESI-Orbitrap-MS data were then processed by multivariate data analysis. With the aim to perform a comparative study of Arbutus extracts, the PCA approach was applied in an untargeted fashion to acquired raw data. LC/MS chromatograms were then pre-processed with the software MZmine to compensate for differences in retention time and m/z between the chromatographic runs. The pre-processed chromatograms were transferred as a peak list table, with rows corresponding to the individual samples, and columns corresponding to the integrated and normalized peak areas.
PCA was developed by working with the peak areas of the total peaks present in the LC/MS dataset (excluding the noisy) in such a way that a matrix was obtained considering these areas, each corresponding to specific m/z values (variables), and the column of the matrix was realized by the different analyzed samples. Score scatter and loading plots are reported in Figure 4A Table 1.
The collected LC-ESI-Orbitrap-MS data were then processed by multivariate data analysis. With the aim to perform a comparative study of Arbutus extracts, the PCA approach was applied in an untargeted fashion to acquired raw data. LC/MS chromatograms were then pre-processed with the software MZmine to compensate for differences in retention time and m/z between the chromatographic runs. The pre-processed chromatograms were transferred as a peak list table, with rows corresponding to the individual samples, and columns corresponding to the integrated and normalized peak areas.
PCA was developed by working with the peak areas of the total peaks present in the LC/MS dataset (excluding the noisy) in such a way that a matrix was obtained considering these areas, each corresponding to specific m/z values (variables), and the column of the matrix was realized by the different analyzed samples. Score scatter and loading plots are reported in Figure 4A

Antioxidant Activity
Two different assays were employed to assess the antioxidant activities of different A. unedo extracts: (a) DPPH assay, a direct method based on the scavenging of the radical; and (b) ABTS assay, based on the inhibition of cation radical by antioxidants. IC50 value (the quantity of an extract able to neutralize the 50% of the radical) was taken into account to evaluate the antioxidant activity because the measurement of the absolute value for the antioxidant activity of an extract is contingent

Antioxidant Activity
Two different assays were employed to assess the antioxidant activities of different A. unedo extracts: (a) DPPH assay, a direct method based on the scavenging of the radical; and (b) ABTS assay, based on the inhibition of cation radical by antioxidants. IC50 value (the quantity of an extract able to neutralize the 50% of the radical) was taken into account to evaluate the antioxidant activity because the measurement of the absolute value for the antioxidant activity of an extract is contingent on many variables, comprising degradation during the analysis and matrix interference, and these contribute to the possibility of a wrong value. Table 2 reports the results obtained for both DPPH and ABTS assays. Regarding the DPPH assay, leaves extracts of both Sardinia's areas (Sassari and La Maddalena) showed an antioxidant activity significantly higher than red fruits and yellow fruits extracts, respectively, at time zero and after 30 min, a time when the reaction was stable and complete.  La Maddalena leaves extracts presented a higher free radical scavenging effect than those from Sassari, even if not statistically significantly different. Similarity, the ABTS results were in agreement with those obtained for DPPH, even if ABTS gave the best results, showing for Sassari and La Maddalena leaves extracts IC50 values of 2.35 ± 0.06 and 2.72 ± 1.68 µg/mL, respectively, lower than those obtained for Trolox, reference compound.
The best results showed by ABTS assay over DPPH depend on the fact that ABTS can be used over a wider range of pH and is able to measure the antioxidant capacity of both water-soluble and lipid-soluble metabolites, since it can be dissolved in both aqueous and organic media, unlike DPPH, which can only be solubilized in alcoholic media [36].
Determination of total phenolic content was carried out for all extracts using Folin-Ciocalteu assay (Table 3). Obtained results were in agreement with the demonstrated antioxidant activity of the three ethanolic extracts obtained from samples from different areas of Sardinia. Data were expressed as means ± SD of three independent experiments. Each result of red and yellow fruits showed a positive correlation with DPPH (* p < 0.01) and ABTS results ( † p < 0.05).
However, a positive Pearson correlation between antioxidant activity measured with DPPH and ABTS and total phenolic contents was detected only in red and yellow fruits, probably because in leaves, antioxidant activity measured with DPPH and ABTS reached saturation early as low concentrations, due to the occurrence of different compounds in leaves extracts. Regardless, the higher amount of total phenols of leaves as well as red fruits and yellow fruits demonstrated that antioxidant activity is governed by phenolic amounts, confirming previous data [37].

Conclusions
In conclusion, in a totally untargeted approach, the power of HPTLC in discriminating origin and organs of samples was evaluated. The obtained results were compared with a more performant as well as more complex technique, mainly HRMS. Comparing the results obtained for both the techniques and visualized by PCA, the discrimination results are comparable. Then, for a preliminary untargeted approach, by integrating HPTLC with PCA, it's possible to discriminate samples with good confidence and in a short time.
The second aim of the paper was to carry out a qualitative analysis of the main compounds occurring in the samples from different origins and organs. This objective was reached by data collected by LC-HRMS, which permitted it to assign an elemental formula for the main m/z value and LC-HR-MS (MS) to enrich information with a fragmentation pattern useful to complete identification of compounds.
Finally, we also confirm that the high antioxidant activity is related to the amount of total phenolic content.
The present study highlighted that the qualitative and quantitative chemical diversity detected by analytical techniques, related to the slightly different habitats in which Arbutus unedo grow, give rise, after multivariate analysis, to a discrimination of the studied samples. HPTLC and LC-ESI-Orbitrap-MS provide different and complementary data and together these may be employed to really differentiate between a wide variety of crude drug powders and herbal medicinal products. Funding: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.