Metabolomics as a Potential Chemotaxonomical Tool: Application in the Genus Vernonia Schreb

The taxonomic classification of the genus Vernonia Schreb is complex and, as yet, unclear. We here report the use of untargeted metabolomics approaches, followed by multivariate analyses methods and a phytochemical characterization of ten Vernonia species. Metabolic fingerprints were obtained by accurate mass measurements and used to determine the phytochemical similarities and differences between species through multivariate analyses approaches. Principal component analysis based on the relative levels of 528 metabolites, indicated that the ten species could be clustered into four groups. Thereby, V. polyanthes was the only species with presence of flavones chrysoeriol-7-O-glycuronyl, acacetin-7-O-glycuronyl and sesquiterpenes lactones piptocarphin A and piptocarphin B, while glaucolide A was detected in both V. brasiliana and V. polyanthes, separating these species from the two other species of the Vernonanthura group. Species from the Lessingianthus group were unique in showing a positive response in the foam test, suggesting the presence of saponins, which could be confirmed by metabolite annotation. V. rufogrisea showed a great variety of sesquiterpene lactones, placing this species into a separate group. Species within the Chrysolaena group were unique in accumulating clovamide. Our results of LC-MS-based profiling combined with multivariate analyses suggest that metabolomics approaches, such as untargeted LC-MS, may be potentially used as a large-scale chemotaxonomical tool, in addition to classical morphological and cytotaxonomical approaches, in order to facilitate taxonomical classifications.


Introduction
The tribe Vernonieae has a Pantropical distribution, being widely present in the New and Old Worlds. In Brazil the tribe is represented by around 40 genera and 450 species [1][2][3][4]. The genus Vernonia Schreb, subtribe Vernoniinae, is one of the largest groups in the Asteraceae family and includes more than 1000 species [5,6]. In South America, there are around 350 species that mainly occur in Northern Argentina, Paraguay, Bolivia and Brazil, the later with approximately 200 species [1][2][3].
Despite the fact that the subtribe Vernoniinae is well established from a taxonomic point of view, there are several classification divergences concerning the generic limits of the Vernonia genus [3,7,8,9,10]. The species within this genus present a great variability in habit and morphology, leading to diverse criteria of taxonomic delimitation [11]. For example, Robinson (1999) suggested segregating several New World species into several new groups (genera), the most representative of them being Lessingianthus, Chrysolaena, Lepidaploa and Vernonanthura, thereby mainly restricting the genus Vernonia to those species growing in North America.
However, this reclassification of the New World species, all originally classified as Vernonia sensu Baker [10], into new genera has not generally been accepted, since the elevation of the different sections to generic level may be premature and does not resolve the taxonomical problem [8]. It can thus be stated that the taxonomical classification of the genus Vernonia is complex and needs further studies. A comprehensive phytochemical characterization of species within this genus may provide helpful chemotaxonomic information that can be used together with the classical morphological and cytotaxonomical data for a more proper and accurate classification of species within this genus [3,8,12]. Recent innovation in untargeted metabolomics approaches, aiming to analyze and compare samples for as many as possible of the detected compounds (both known compounds and yet unknowns) can provide a detailed insight into the differences and similarities in phytochemical composition resulting related from genetic background [13,14].
With regard to previous phytochemical analyses of leaves from plants of the genus Vernonia sensu Baker, in both North and South America this genus is characterized by the presence of both flavonoid classes, such as flavones and flavonols, and sesquiterpene lactones (SLs) [1] belonging to the classes of germacranolides [15], such as glaucolides [16], hirsutinolides [17], cadinanolides [18], and guaianolides [19]. Furthermore, several saponins, for instance vernonioside D 1 , D 2 and E, have been identified in Vernonia amygdalina [20,21]. In addition, there are reports on the presence of coumarins [1] while diterpenes and alkaloids seem to be absent [5,22].
We here report the use of untargeted metabolomics approaches, employing HPLC(DAD)-MS(ESI-QTOF), followed by multivariate analyses methods and a phytochemical characterization of ten Vernonia species (sensu Baker). For that we compare our results to the classification proposed by Robinson (1999), with the intention of evaluating if untargeted metabolomics could be employed as a chemotaxonomic tool in order to help taxonomical classifications.

Results and Discussion
LC-MS-based metabolic fingerprinting of crude aqueousmethanol extracts prepared from dried leaves was performed for all species, in both positive and negative electrospray ionization (ESI) modes. The data obtained in positive mode were automatically processed by MetAlign and redundant peaks removed using MSClust software, then these final reconstructed metabolite features were submitted to multivariate analysis. The resulting principal component analysis (PCA) and hierarchical cluster analysis (HCA) are shown in Fig. 1 and 2, respectively. Concomitantly, the main chromatographic peaks were identified.
For compound identification, UV spectra were used to infer the secondary metabolite class of the major chromatographic peaks, while the accurate mass obtained for the molecular ions ([M+H] + and/or [M2H] 2 ) were used to calculate possible molecular formula, considering a maximum deviation between observed and calculated mass of 5 ppm. The molecular formula and accurate mass were used as entry for Scifinder and Dictionary of Natural Products databases.
Product ion spectra (MS/MS) of selected precursor ions formed by collision-induced dissociation (CID) fragmentation were performed for determination/confirmation of compound annotation. The MS/MS spectra were compared with literature reports for chlorogenic acids [23,24], flavonoids [25,26], and sesquiterpene lactones [24]. The annotation of chromatographic peaks was performed partially based on previous phytochemical studies of Vernonia species. Whenever possible, retention time comparisons with authentic standards were carried out.
The metabolic profiling of plants analyzed led to the putative identification of 81 compounds, comprising chlorogenic acids, flavonoids, SLs and saponins, several of which are described here for the first time in the genus Vernonia Schreb. Table 1 shows the compound identities and the differentially annotated compounds accumulating in the various species analyzed. Detailed mass spectrometry information for identification of all 81 compounds is provided as Supporting Information S1.
When comparing the metabolite profiling and multivariate analysis results with the classification proposed by Robinson (1999) for the plants from Vernonia genus, i.e. the segregation of these species into new genera, one can note that our results are in agreement with this classification. As proposed by Robinson (1999), V. brasiliana, V. discolor, V. ferruginea and V. polyanthes should be considered as belonging to the Vernonanthura group, while V. linearifolia, V. glabrata and V. onopordioides belong to the Lessingianthus group, V. herbacea and V. platensis to the Chrysolaena group, and V. rufogrisea to the Lepidaploa group.
SLs were only found in V. brasiliana, V. platensis, V. polyanthes and V. rufogrisea. V. rufogrisea was the only species accumulating 8aacetoxy-10a-hydroxy-13-O-methylhirsutinolide; acetoxy-hydroxymethylhirsutinolide; 1,4-epoxy-1-methoxy-8,13-diacetoxy-10-hydroxygermacra-5(E),7(11)-dien-6,12-olide; 8b-propioniloxy-10bhidroxyhirsutinolide-13-O-acetate; 8b-acetoxy-10b-hidroxyhirsuti-nolide-1,13-O-diacetate and glaucolide B. The presence of this great and unique variety of SLs in V. rufogrisea is in agreement with the classification proposed by Robinson (1999), placing V. rufogrisea apart from the other species studied, since this species is unique within the group of Lepidaploa (sensu Robinson). This separation of V. rufogrisea could be clearly observed in the PCA (Fig. 1), which confirms that differentiation also occurs on metabolic level. The SL 8a,13-diacetoxy-10a-hydroxyhirsutinolide was found in both V. rufogrisea and V. platensis. V. polyanthes was the only species that showed accumulation of the SLs piptocarphin A and piptocarphin B, while glaucolide A was also found in V. brasiliana, separating these two species from two other species from Vernonanthura group, V. discolor and V. ferruginea. Also, HCA (Fig. 2) clearly shows that V. discolor and V. ferruginea belong to Vernonanthura group. In spite of PCA (Fig. 1) showing both species very close to Chrysolaena group, y-axis explains Vernonanthura group separation. The great difference observed in SLs profiles from the species studied suggests that this class may not be the best chemical markers for the genus Vernonia as they are being used for Asteraceae in general [15,27,28].
Regarding the contribution of flavonoids in the classification of some species, it should be noted that isoquercetrin was found in all species of the Vernonanthura group, while rutin was only found in V. discolor and V. ferruginea. On the other hand, the flavones chrysoeriol-7-O-glycuronyl and acacetin-7-O-glycuronyl were found in V. polyanthes only, while the flavanone hesperetin-7-Orhamnoglucoside was only present in V. rufogrisea. In addition, it was observed that V. linearifolia is chemically apart from the other species of the Lessingianthus group, being the unique species within this group accumulating isoorientin-30-O-glucupyranoside, quercetin-3-O-di-hexose-O-pentose and kaempferol-3-O-di-hexose-Opentose. This difference can be observed in the multivariate analysis, in which V. linearifolia was separated from the other species of this group ( Fig. 1 and 2).
Saponins appeared to contribute to the segregation of V. linearifolia, V. glabrata and V. onopordioides within the Lessingianthus group. These three species were unique in showing a positive response to the foam test, indicative of the presence of saponins, while LC-MS analysis also showed the presence of putative saponins only in the Lessingianthus group. Thus, saponins help to explain the differentiation of the Lessingianthus group from the others and could be an auxiliary tool for a rapid identification of species belonging to this group.
Finally, it is important to highlight the specific presence of clovamide, a N-coumaroyl-3-hydroxytyrosine [29], in both V. herbacea and V. platensis, placing these species away from the other species studied. This segregation is also in accordance with the classification proposed by Robinson (1999) of these two species into the Chrysolaena group.

Conclusions
It was observed that the segregations of species based on PCA (Fig. 1) and HCA (Fig. 2) of their metabolic profiles are well in agreement with the latest classification proposed by Robinson (1999). For example, the species V. brasiliana, V. discolor, V. ferruginea and V. polyanthes clustered together, coinciding with the group Vernonanthura, while V. glabrata and V. onopordioides clustered according to the Lessingianthus group, V. herbacea and V. platensis clustered into the Chrysolaena group, while V. rufogrisea was separated corresponding to the Lepidaploa group.
Our results suggest that metabolic profiling and multivariate analysis might be a fast and more comprehensive tool for chemotaxonomic purposes than the classical and laborious phytochemical investigation. The results obtained using comprehensive metabolomics approaches, such as untargeted LC-MS, PCA and HCA, may be applied in chemotaxonomic studies with the aim to help taxonomical classifications, in a similar way as applied to filamentous fungi classification [30].    Table 1. doi:10.1371/journal.pone.0093149.g002 Table 1. HPLC chromatographic peaks identified in species from Vernonia Schreb. obtained by the HPLC-MS analysis were used as input for CID fragmentation in HPLC-MS/MS. N 2 was used as drying, nebulizer and fragmentation gas.

Sample preparation
The leaves of each single plant were dried under air circulation (356C, 24 h) and powdered using an analytical knife mill. The samples were prepared using 20.0 mg of dry wt weighed in a glass vial and extracted with 3.0 ml of a solution of MeOH-H 2 O (7:3, v/v) in an ultrasonic bath for 10 min. Finally, an aliquot of 1.0 ml was taken from the extract, filtered in a 0.45 mm PTFE membrane and 20 ml were analyzed as described in item 4.3.

Chromatographic peak identification
Compounds were tentatively identified relied on UV spectrum and molecular formulae calculated from accurate mass measurements, both obtained from HPLC-UV-MS analyses. Such data were used to suggest secondary metabolites for each peak and were screened against the molecular formulae in the Scifinder and Dictionary of Natural Products databases. Hence, obtained informations were compared with the secondary chemistry previously reported for the Vernonia Schreb genus.

Detection of saponins by foam test
The dried plants (500 mg) were put in a graduated cylinder with 2 ml of distilled water. The suspension was shaken for 15 s and a two cm layer of foam indicated the presence of saponins.

Untargeted data processing and multivariate analysis
Mass signals from the raw data files were automatically extracted and aligned by MetAlign software [31], resulting in 4438 mass signals (determined as peak height) at a signal to noise ratio higher than 4000. Mass signals belonging to the same molecule, like isotopes, fragments and adducts, were subsequently re-grouped using MSClust software [32], resulting in 528 reconstructed metabolites and their relative intensity in each sample. Multivariate analysis was performed using GeneMaths XT software (version 2.11 -AppliedMaths), after 2 log transformation of metabolite signal intensities. Metabolite intensity signals (variables) were normalized by dividing the mean of each variable. HCA was performed by using Neighbor joining and Pearson's coefficient matrix.

Ethics statement
The use of these species was allowed by CNPq at the Authorization for access of samples from the Brazilian Genetic Heritage (010091/2011-4).