Phytochemical Profile, Chemotaxonomic Studies, and In Vitro Antioxidant Activities of Two Endemisms from Madeira Archipelago: Melanoselinum decipiens and Monizia edulis (Apiaceae)

Melanoselinum decipiens and Monizia edulis (Apiaceae) are two endemic plants from Madeira archipelago, phytochemical compositions of which remains little explored, despite their use in folk medicine. Using liquid chromatography with diode array and electrospray ionization/mass spectrometry analysis, their polyphenolic profile was established for the first time. Fifty‐six compounds were identified with 5‐O‐caffeoylquinic acid, quercetin‐O‐(malonyl)hexoside, luteolin diacetyl, and quercetin‐O‐hexoside being the major constituents in the leaves of both plant species (≥ 0.76 mg/g of dry extract). Principal component analysis provided a suitable tool to differentiate targeted plants. Naringenin‐6,8‐di‐C‐glucoside, quercetin 3‐O‐pentosylhexoside, and 1,5‐O‐dicaffeoylquinic acid can be used as discriminatory taxonomic/geographical markers for M. edulis subspecies from Madeira and Porto Santo populations. This methodology of using polyphenols as chemotaxonomic markers proved to be useful for identification of plant species since the results are consistent with previous taxonomical data. The free‐radical scavenging activities of the M. decipiens extracts proved to be higher than those of M. edulis, which correlated well with their phenolic content (R2 > 0.906).


Introduction
The study of the phytochemical chemical profile is rather important for herbal identification, clarification of their bioactivities and possible side effects, and enhancing product quality control. [1] High-performance liquid chromatography has been widely used for the chemical identification of plant secondary metabolites and discrimination of many closely related herbs, since it can provide accurate information on complex samples and mixtures. [2] Flavonoids and phenolic acids in particular are highly relevant as bioactive components of medicinal and edible plants. One important feature of polyphenols is their usage in chemotaxonomic studies, mainly due to their abundant occurrence in vascular plants and their structural variability and stability. [3 -5] These secondary constituents have been used to solve taxonomic problems that have arisen as a result of morphological classification. [1][2] [6] [7] Principle component analysis (PCA) has been successfully applied for distinguishing, for instance, Angelica sinensis from related Apiaceae herbs based on their chromatographic profiles and coniferyl ferulate amount. [8] Olivier et al. [9] used rosmarinic acid and its derivatives for the chemotaxonomical differentiation in Apiaceae subfamilies. Chemotaxonomic classification has also been applied for quality control of grapevines, tomatoes, flowers, and medicinal herbs. [1][7] [10 -12] Apiaceae (or Umbelliferae) is a large family represented by 2500 -3700 species, which includes known plants used in culinary like parsley, carrot, celery, coriander, fennel, and cumin. [13] Melanoselinum decipiens (SCHRAD. & J. C. WENDL.) HOFFM and Monizia edulis LOWE (Apiaceae) are two endemic plants from Madeira archipelago (Portugal), [14] [15] phytochemical compositions of which have been paid little attention.
Melanoselinum decipiens ('Madeira giant black pars-leyʼ or ʻcattle celeryʼ) inhabits shady rocks and banks in Laurissilva (Madeira Laurel Forest) and is an herbaceous monocarpic perennial shrub, resembling giant parsley, that can grow up to 3 m high. [15] It was once cultivated for cattle fodder and their leaf extracts were used for skin diseases. [16] Monizia edulis (ʻcarrot treeʼ) is a monocarpic, perennial shrub growing at clefts or hollows and hedges of the islands, that resembles a large arborescence carrot (about 1.2 m tall). [15] Formerly, it has been used as vegetable (stout roots) by fishermen and goat-herders in lack of other food supplies. This species was also gathered from the wild for medicinal digestive properties of the leaves. [17] Based on morphological traits, four subspecies of M. edulis have been recently designated from Madeira archipelago [18] : two from Madeira Island (M. edulis subsp. isambertoi and M. edulis subsp. giranus); one on Deserta Grande Island (M. edulis subsp. edulis) and another on Porto Santo Island (M. edulis subsp. santosii).
Previous studies have reported b-pinene as the predominant volatile from M. decipiens (essential oils and aerial parts) [19] [20] and c-terpinene and myristicin as main components from M. edulis essential oils. [21] The presence of some sesquiterpene lactones was also described in M. decipiens leaves. [22] [23] Despite the existence of the mentioned works, they addressed only essential oils components and overlooked phenolic characterization. Therefore, the aim of this study was to establish, for the first time, the phytochemical profile of MeOH leaf extracts from M. decipiens and M. edulis. Also, a comparison was made for M. edulis populations collected in two different geographical locations. PCA was used to investigate whether polyphenolic composition of M. edulis subspecies provide any correspondence with and support for the establishment of currently recognized described taxa. Additionally, correlation of the polyphenols with the in vitro antioxidant activity of these plants' extracts was determined, in order to evaluate their interest as possible novel foodstuffs.

Qualitative Phytochemical Analysis
In this study, MeOH extracts from two plant species were submitted to liquid chromatography with diode array and electrospray ionization/mass spectrometry (HPLC-DAD-ESI À /MS n ) analysis in order to establish their phytochemical profile and identify potential chemotaxonomic markers. Representative chromatograms of the MeOH extracts are shown in Fig. 1.
In general, in the negative ionization mode (ESI À ) MS 1 spectrum, the most intense peak corresponded to the deprotonated molecular ion [M À H] À ; this permitted to perform MS n analysis. The mass spectra of the conjugated phenolic compounds showed the aglycone ion as a result of the loss of sugar moieties like hexosyl, caffeoyl, and pentosyl (À162, À162, À132 Da, resp.). Compounds were numbered by their order of elution and this numeration was kept identical for both species (Table 1).
Among the 56 identified compounds, there were 23 phenolic acids (hydroxycinnamic and hydroxybenzoic acids), 18 O-flavonoids (flavones, flavonols, and flavanones type), and 15 other compounds (terpenoid, lignan, coumarin, organic acids, and saccharides). As expected, different species exhibited different phytochemical profiles. Within the same species, quantitative variations were more relevant than qualitative ones, but still there are statistically significant differences. Nevertheless, most of the identified compounds were common to both species.

Phenolic Acids
The presence of mono-and dicaffeoylquinic acids isomers was confirmed by comparison of their MS n spectra with standards and information from previous reports. [24] [25] Compounds 9 and 17 gave [M À H] À ions at m/z 353 and showed fragment ion at m/z 191 as MS 2 base peak. Compound 9 showed an intense fragment ion at m/z 179 (> 40% of base peak) and was characterized as 3-O-caffeoylquinic acid. The isomer 5-O-caffeoylquinic acid, with higher retention time, was assigned to compound 17.
Compound 10 displayed [M À H] À ion at m/z 707 and the presence of fragment ions at m/z 515 and 353, characteristic of caffeoylquinic acid (CQA) derivatives, was observed at MS 2 . However, 10 was not completely characterized being assigned as a dicaffeoylquinic acid derivative.
Compound 6 exhibited [M À H] À ion at m/z 349 and produced MS 2 base peak at m/z 313 (by loss of 36 Da). After loss of a sugar moiety, it gave origin to fragment ions at m/z 177 and 151 typical of vanillin. [28] Since additional data were not available, 6 was characterized as a derivative of vanillin-O-hexoside.
Malic acid (compound 3) was characterized based on previous publication. [28] Compound 14 with [M + HCOOH À H] À at m/z 399 suffered a loss of 208 Da (46 + 162 Da) at MS 2 . Based on further fragmentation data (shown in Table 1), 14 was characterized as scopoletin-O-hexoside. [40] Drovomifoliol-O-glucoside (roseoside) with [M + HCOOH À H] À ion at m/z 431 was assigned to compound 19. [41] Furthermore, compounds 21 and 23, with [M + HCOOH À H] À at m/z 433, were characterized as isomers of dihydrovomifoliol-O-glucoside. Compound 36, with [M À H] À ion at m/z 523, gave origin to a fragment ion at m/z 361 (by loss of 162 Da) and was characterized as secoisolariciresinol-O-hexoside. [42] Compound 60 displayed [M À H] À ion at m/z 519 and gave origin to MS 2 base peak at 315 by loss of 204 Da. By loss of 15 Da, it displayed typical ellagic acid fragment ions. This fragmentation behavior was consistent with that of methyl-ellagic acid-O-acetylhexoside. [31] Quantitative Analysis of Phenolic Compounds In total, 15 main polyphenols were quantified by the HPLC-DAD method ( Table 2). It was not possible to quantify all identified compounds due to their low UV-absorption and because some of them were present in trace amounts.
Quantification of individual phenolic compounds varied among the analyzed samples, with levels ranging between 15.76 and 19.52 mg/g ( Table 2) with M. decipiens extract being the richest one. Statistical significant differences (P < 0.05) were found between total individual phenolic content (TIPC) of M. edulis populations. Hydroxycinnamic acids were the dominant phenolic group (> 76% of TIPC), followed by A previous chemical characterization of Apiaceae species, [27] [43] found TIPC of 42 mg/g dry extracts (DE) and 2.2 mg/g DE for anise and coriander seeds, respectively; although 5-O-CQA (17) and luteolin-O-hexoside (40) were present in lower quantities than in present work.

Principal Component Analysis
PCA of 15 compounds in two Apiaceae species was performed, and as shown in Fig. 2, the distribution plots were divided into three groups.
The PCA score scatter plot of the two-first principal components (which explains 100% of the total variability) is shown in Fig. 2a. The loadings of each compound (variable) that contribute to explain the differentiation between the plant species and collection area is shown in Fig. 2b.
PC1 that explained 69% of the total variability shows target plants discrimination based on their species, where M. edulis (Madeira and Porto Santo) are projected in PC1 negative and M. decipiens is on PC1 positive. Taking in account the loading plot (Fig. 2b), the compounds that contribute to these results were: sinapic acid-O-hexoside (12), luteolin-O-hexoside (40), and luteolin-O-diacetylhexoside (56). However, PC2 (that explained 31% of the total variability) separated M. edulis samples based on subspecies (or collection area): subsp. isambertoi (Madeira) samples are below PC2 axis while subsp. santosii (Porto Santo) is positioned in PC2 positive. In case of M. edulis samples, the obtained results support their taxonomical separation into two distinct taxonomic groups as suggested by Fernandes and Carvalho. [18] According to Fig. 2b, polyphenols responsible for the obtained results are naringenin-6,8-di-C-glucoside (13), quercetin-O-pentosylhexoside (32), and 1,5-O-dicaffeoylquinic acid (48). Based on the statistical analysis, these compounds can be used as potential geographic markers. Previously, flavonoids and phenolic acids have been used as chemotaxonomic markers for other Apiaceae species. For example, chrysoeriol-O-(pentosyl) hexoside and luteolin-O-pentoside are markers for different Torilis species. [44] (R)-3 0 -O-b-D-Glucopyranosylrosmarinic acid is used as chemotaxonomic marker for the subfamily Saniculoideae, [9] while luteolin-7-O-glycosides are useful to discriminate Soranthus and Ferula species. [45] Compounds with flavonol-like structures (rutin, isoquercitrin, isorhamnetin-3-glycoside, isorhamnetin-3-rutinoside, and quercetin) are chemotaxonomically important in plants of the genus Peucedanum (Apiaceae). [46] Total Phenolic and Flavonoid Contents and In Vitro Antioxidant Activity Assays The amounts of total phenolics and total flavonoids varied in the different analyzed extracts and ranged from 34.10 to 42.63 mg GAE/g DE and from 10.33 to 19.66 RUE mg/g DE, respectively (Fig. 3a). M. decipiens showed the highest contents, followed by M. edulis Porto Santo > M. edulis Madeira. Only for total flavonoid content (TFC) there were significant differences between M. edulis samples (P < 0.05). This variation can be expected for plant extracts due to the presence of other constituents and/or the presence of different types of polyphenols. Also, the absolute numerical value of TIPC was, as it generally is, lower than those determined by colorimetric methods (Fig. 3a). This difference shows well that colorimetric assays are not specific to polyphenols. Despite their shortcomings, colorimetric assays for measurement of phenolic and flavonoids contents are still present in many publications and are useful to establish comparison with other available data.
Recently, [47] TPC and TFC of seven Indian Apiaceae spices were determined. Using those for comparison, M. decipiens showed higher TPC than tested plants (< 38.83 mg GAE/g DE); while caraway and coriander had higher amounts than M. edulis. In case of TFC results, coriander, cumin, and carom presented higher contents (> 27.45 mg RU/g DE) than our targeted species. In another study, [48] coriander and parsley (leaves and seeds) showed lower TPC (6.2 -9.2 mg GAE/g DE) than reported in present work. Hinneburg et al. [49] also found lower amounts of TPC in parsley, aniseed, and fennel (20.8 -30.3 mg GAE/g DE); only cumin was richer than M. edulis extracts (37.4 mg GAE/g DE). TPC varied in different parts of fennel (8.61 -65.85 mg GAE/g DE). [50] Only shoots had higher values than our targeted species, while leaves and inflorescences showed comparable contents with M. edulis.
In all antioxidant assays, a good scavenging activity was shown for all species with a wide range from 0.035 to 1.09 mmol TE/g DE (Fig. 3b). The higher values obtained by ABTS method could not only be related to differences in the sensitivity of these methods but also it measures both hydrophilic and lipophilic antioxidants. Based on the obtained results, target species may also prevent the formation of other biologically important oxidative species resultant from the reaction of NO and SO, like peroxynitrite and hydroxyl radical. [51] In general, M. decipiens was found to be the most potent radical scavenger toward DPPH, NO, and SO, followed by M. edulis Porto Santo > M. edulis Madeira. Only in NO assay, there were significant differences (P < 0.05) between M. edulis populations. These results were also supported by quantitative determination of antioxidant compounds using the HPLC-DAD quantitative analysis. For ABTS, a different trend was observed (M. decipiens > M. edulis Madeira > M. edulis Porto Santo). This variation in the observed antioxidant effects may be related with the distribution of polyphenols on different samples.
An explanation for reported differences of M. edulis samples may be because these samples were collected in different geographical locations (Madeira and Porto Santo Islands), which according to Fernandes and Carvalho [18] correspond to different subspecies. Porto Santo population (subsp. santosii) were grown on a wild environment (at sea level) and was subjected to harsh environmental conditions, such as high UV levels from the sunlight and dryness. In the wild environment, plant species are more subjected to stress factors which induce intense synthesis of phenolic compounds as a response to abiotic stress in order to prevent oxidative damage of the plant cellular structures. Madeira counterpart (subsp. isambertoi) was collected in Madeira Botanical Garden where the growth conditions are ʻcontrolledʼ, with less sun exposure, higher altitude (ca. 300 m), regular watering, etc.
These environmental factors are well-known to affect the phytochemical composition of plants. [52] Nevertheless, the observed trend for the antioxidant activities of the different extracts correlated well with the TIPC values (R 2 ≥ 0.906) ( Table 3). In general, hydroxycinnamic acids were better correlated than flavonoids in all assays. Individual components correlation with antioxidant activities was also determined ( Table 3): coumaroylquinic acid (26), quercetin-O-(pentosyl)hexoside (32), and quercetin-O-(malonyl)hexoside (47) are the most important contributors to the obtained data. Antioxidant activity is directly related to the particular structure of polyphenols. In fact, alterations in the arrangement of the OH groups and degree of substitution by glycosylation decrease the antioxidant activity. Moreover, interactions established between components of a matrix can lead to synergistic/antagonist effects. [53] Conclusions In this work, the phytochemical profile of two endemic plants from Madeira archipelago was established for the first time   Table 1). species. A comparison study was made for M. edulis populations collected in two different geographical locations. Samples from Porto Santo showed a higher phenolic content and slightly different phytochemical composition than Madeira counterpart; naringenin-6,8di-C-glucoside, quercetin 3-O-pentosylhexoside, and 1,5-O-dicaffeoylquinic acid can be used as geographical markers or, if not most importantly, as taxonomic markers. A natural follow-up of this work will be the cultivation of M. edulis subsp. santosii in the control conditions of a Botanical Garden, from seeds collected in the wild. Finally, evaluation of antioxidant activity revealed that M. decipiens was the most active compared to M. edulis samples, which is in agreement with the higher phenolic composition. Taking into account the obtained data and relative abundance, M. decipiens could be a good candidate as novel spice/additive for seasoning foods.

Chemicals and Standards
The following reagents were purchased from Panreac

Sample Preparation
For analysis, leaves were lyophilized to dryness (Alpha 1-2 LD plus; Christ, Germany), ground to powder, and stored at À20°C. Dried material (1 g) was extracted with 25 ml of methanol (25 ml) using a sonicator Bandelin Sonorex (Germany) at 35 Hz and 200 W for 60 min (r.t.). Then, chlorophylls were removed by adsorption on activated charcoal, and extracts were filtered and concentrated to dryness under reduced pressure in a rotary evaporator (B€ uchi Rotavapor R-114; USA) at 40°C. The resulting dry extracts (DE) were stored at 4°C until further analysis.

Chromatographic Conditions
The HPLC analysis was carried out on a Dionex ultimate 3000 series instrument coupled to a binary pump, a diode-array detector (DAD), an autosampler, and a column compartment (kept at 20°C). Separation was performed on a Phenomenex Gemini C 18 column (5 lm, 250 9 3.0 mm i.d.) using a mobile phase composed by MeCN (A) and HCOOH/H 2 O (0.1%, v/v) at a flow rate of 0.4 ml/min. The following gradient program was used: 25% A (10 min), 25% A (20 min), 50% A (40 min), 100% A (42 -47 min), and 20% A (49 -55 min). Spectral data for all peaks were accumulated in the range of 210 -400 nm. Plant extracts were filtered (0.45 lm) and injected (5 ll). For HPLC-DAD/ESI-MS n analysis, a Bruker Esquire model 6000 ion trap mass spectrometer (Bremen, Germany) with an ESI source was used. The MS n analysis worked in negative mode and scan range was set at m/z 100 -1000 with speed of 13 000 Da/s. The conditions of ESI were as follows: drying and nebulizer gas (N 2 ) flow rate and pressure, 10 ml/min and 50 psi; capillary temp., 325°C; capillary voltage, 4.5 keV; collision gas (He) pressure and energy, 1 9 10 À5 mbar and 40 eV; and fragmenter, 1.0 eV. Esquire control software was used for the data acquisition and data Analysis for processing.

DPPH Radical Scavenging Activity
The DPPH assay followed a previously reported method [54] : aliquot (100 ll, 5 mg/ml) was added to DPPH radical soln. (3.5 ml, 0.06 mmol/l). Absorbance at 516 nm was measured after 30 min of reaction. The results were expressed as mmol TE/g DE.

Superoxide Radical (SO) Scavenging Activity
Superoxide radicals were generated by the NADH/PMS system as described previously [56] : sample (25 ll, 5 mg/ml) was mixed with soln. (200 ll, 0.1 mM EDTA, 62 lM NBT, and 98 lM NADH). The reaction was initiated with the addition of PMS (25 ll, 33 lM containing 0.1 mM EDTA) to each well. All solutions were prepared in 0.1M phosphate buffer (pH 7.4). The absorbance was read at 550 nm (Victor 3 microtiter reader; Perkin-Elmer, Germany) and results were expressed as mmol TE/g DE.

Statistical Analysis
All samples were assayed in triplicate and the results were given as the means AE standard deviations. Data was analyzed by a one-way ANOVA using SPSS for Windows, and IBM SPSS Statistics 20 (SPSS, Inc., USA). A value of P < 0.05 was considered statistically significant. PCA was applied to the concentration of individual polyphenols, determined by the HPLC-DAD method.