Variability of Polyphenol Compounds in Myrtus Communis L. (Myrtaceae) Berries from Corsica

Polyphenol compounds were extracted from Myrtus communis L. berries (Myrtaceae) by maceration in 70% ethanol and analysed by HPLC-DAD and electrospray mass spectrometry. The Myrtus berries were collected at maturity from seven localities on the island of Corsica (France) and the sampling was carried out during three years. The polyphenol composition of Corsican Myrtus berries was characterized by two phenolic acids, four flavanols, three flavonols and five flavonol glycosides. The major compounds were myricetin-3-O-arabinoside and myricetin-3-O-galactoside. Principal components analysis (PCA) is applied to study the chemical composition and variability of myrtle berries alcoholic extracts from the seven localities. Canonical analysis and PCA data distinguishes two groups of myrtle berries characterized by different concentrations of polyphenols according to soil and years of harvest. The variations in the polyphenol concentration were due to biotic and abiotic factors.


Introduction
Flavonol glycosides were the major compounds found in the myrtle extracts. They represented 58% of the total amount of phenolic compounds identified at 280 nm by HPLC-UV. Myricetin-3-Oarabinoside was the major constituent (106.6 to 1435.9 mg/100g dw ). These flavonols compounds are present in most fruits and plants, and are also the most studied [21]. Their anti-cancer action is also acknowledged [22][23][24].
The major component was myricetin (207.8 to 1053.6 mg/100g dw ). (-) Epigallocatechin, with a concentration range between 124.0 to 952.9 mg/100g dw , was the main constituent from the flavanol family. Phenolic acids were present at low concentrations. The major compounds were previously identified in the literature [12,17,18,25]. Romani et al. [12] identified caffeic acid, in addition to the compounds identified in our samples. Montoro et al. [17,18] showed the presence of six flavonoids. In the berry extracts from Myrtus we identified quercetin and kaempferol that aren't present in other berry extracts.

Effects of localities of origin of M. Communis berries on polyphenol compound compositions
The localities of origin have a significant effect on the polyphenol composition of myrtle extracts ( Table 2). The total concentration was not significantly different between localities 1 and 2 (ANOVA, p > 0.05). Their concentrations, measured at between 6,400.8 ± 281.4 to 6,743.3 ± 241.8 mg/100g dw , were the highest observed among all the sites. Similarly, localities 5 and 6 have total concentrations of polyphenols which were not significantly different (ANOVA, p > 0.05). The other localities were significantly different (ANOVA, p < 0.05). To synthesize the polyphenolic compositions data, Canonical Analysis (CA) was applied to examine the relative distribution of localities according to polyphenol compounds ( Figure 2).  The cluster analysis suggested the existence of groups based on the polyphenol concentrations. The first group was constituted by localities 1 and 2, and the second group by the other localities. The general structure of the dendrogram confirms the ANOVA results and suggests the existence of two main clusters. The same result was reported by Piras et al [5] who showed that the PCA of ToF-SIMS data from alcoholic extracts of myrtle berries distinguishes two groups characterized by a different concentration of anthocyanins, flavonols and α-tocopherol.
The two PCA axes explained 84.0% and 14.9% of the variance respectively. The first PCA axis was positively correlated with all polyphenol compounds. The second PCA axis was positively correlated with phenolic acids and negatively with flavanols, flavonol glycosides and flavonols. PCA offered a representation of the distribution of polyphenol compounds families from different localities. As shown in Figure 3, the first axis explained 84.0% of the variance and showed that the polyphenols concentration was more important from localities 1, 2 and 3, corresponding to shaly soil. In the literature data, a variation of anthocyanin derivatives in myrtle berries collected in different geographical areas in Italy [18] and Sardinia [5] were observed. Moreover, localities 4-7 showed a lower concentration of polyphenols in alluvia and limestone soil as compared to shaly soil. The total amount of polyphenols was thus more important in the shaly soil sites.

Effect of the meteorological conditions on the polyphenol compositions of M. communis berries
An important variability between polyphenols concentration and year of harvest was observed ( Table 3). The polyphenol concentration was significantly different (ANOVA, p < 0.05) from one year to another during the three years of harvest examined. This variability is not related to the soil and origin of the M. communis berries. We noticed however an exception to this general behaviour for flavanols and flavonol glycosides, whose concentration was not significantly different (ANOVA, shaly Alluvia and limestone p > 0.05) for localities 5 and 2, respectively. The two PCA axes explained 58.0% and 23.2% of the variance, respectively (Figure 4). The first PCA axis was positively correlated with all polyphenol compounds and shaly soil and negatively correlated with limestone and alluvia.  In the literature, Piras et al. [5] showed that the differences were not necessarily ascribable to geoclimatic factors, and genetic and/or environmental factors could be considered to affect the quantitative composition of myrtle berry extracts and irrigation affected the concentration of many compounds. As shown in Figure 4, the year 2005 presented the highest polyphenol concentration, while years 2003 and 2004 presented lower concentrations. We noticed however that along the three years of harvest, the concentrations were more important when berries were grown on shaly soils. Moreover, the biotic and abiotic factors induced the same differences in polyphenol concentrations (Figure 4). These variations can impact the quality of myrtle-derived products (myrtle liquor).

Conclusions
The major constituents of Myrtus communis. L. leaf extracts were myricetin-3-O-arabinoside and myricetin-3-O-galactoside. We showed that there are variations in the concentrations of polyphenols according to the sampling sites and years of harvest. The polyphenol concentrations were more important on sites characterized by shaly soils. These concentrations were low when the soil was constituted by alluvia. These levels vary according to biotic and abiotic factors and consequently year to year, in line with the notion of "vintage" as it refers to myrtle liquor and wine products.

Sampling and reagents
Samples were collected in seven stations spread over the whole island, under various environmental conditions (pedoclimatic conditions; Table 4). We collected 400 grams of berries M. communis for each locality on the same individual tree during three years in the same vegetative stage (November, 15 th ). November corresponds to the maturity of the berries and thus their harvest for the preparation of wines and liqueurs. One hundred grams were used for the extraction of polyphenol compounds. All extractions were made in triplicate (3 × 100 g). Sampling should allow the detection of a specific intravariability that can lead to a possible chemical polymorphism. The berries were lyophilised and extracted with a solution of ethanol (500 mL at 70%) during one day at room temperature. After filtration on a Buchner funnel and extraction with ethyl acetate (3 × 50 mL; Aldrich > 99.9%), the solvent was evaporated with a rotary evaporator at 30 °C under vacuum. The extracts were combined and evaporated. The residue obtained is dissolved in 10 mL of methanol (Aldrich > 99.9%), and then filtered through a 0.45 µm membrane (Phenomenex, France).

Quantification of phenolic compounds
The polyphenols were quantified using external standards which were purchased from Extrasynthese (Geney, France). Five injections were made for each level, and a weighed linear regression was generated. The calibration curve with the external standard was obtained using concentrations with respect to the area obtained from the integration of peaks. The relation between variables was analysed using linear simple correlation.

MS n standard analysis
MS and MSn analysis were carried out on a Finnigan LCQ Classic ion trap mass spectrometer (Thermo, formerly Finnigan-MAT, San Jose, CA, USA) equipped with an APCI interface and an ion trap mass analyzer. The software used is Xcalibur. A syringe pump was used for the direct infusions of reference compound solutions, at a flow rate set at 5 µL/min in an HPLC flow (500 µL/min) composed of acetonitrile and MilliQ (18.2 MΩ) water with 0.1% of acetic acid through an HPLC tee. Solutions of standards were prepared at a concentration of 0.1 mg/mL. In other cases, the reference compounds were directly diluted in acetonitrile (LC/MS grade ROTISOLV 99.95% purchased from Carl Roth, Karlsruhe) The operating parameters were as follows: damping gas, helium (He); nebulizing gas, nitrogen (N 2 ); maximum injection time was set at 50 ms; the number of microscans was set at 3. For APCI in negative and positive mode, the source parameters were set as follow: discharge current, 5 µA; capillary temperature 200 °C; sheath gas (N 2 ), 90 (arbitrary units); auxiliary gas, 25 (arbitrary units). In order to maximize the response for each reference compound, the voltages on the lenses (capillary voltage, tube lens offset, second and first octapole offset, interoctapole lense) were optimized using the automated LCQ TunePlus function of Xcalibur software. For MS/MS experiments, CID was carried out using Helium as collision gas. The collision energy are reported as a percentage of the maximum 5 V p-p normalized for the parent ion m/z (NCE: Normalized Collision Energy) [26]. Collision energy for CID was optimized between 25% and 46% of maximum, and the isolation with of precursor ions was 1.5 amu.

Data analysis
The significant difference for statistical analysis was determined by one-way analysis of variance (ANOVA). ANOVA was used when the application conditions were satisfied, i.e. normal distribution of treatment group means and homogeneity of variances between means, using the Shapiro-Wilk test and Bartlett test, respectively. Differences were considered to be significant when p < 0.05. When a difference was found, Tukey's HSD post-hoc comparison technique was used to determine which plots were different. The correlations between the polyphenols concentration from localities and extracted berries of M. communis were established using PCA on a matrix formed by the association of different classes of polyphenols compounds. CA produced a dendrogram (tree) using Ward's method of hierarchical clustering, based on the Euclidean distance between pairs of sample localities. The Statistical Graphics Corporation's® "Statgraphics for Windows" software package was used for these various tests, together with Stat R version 2.6.1. PCA and CA are data mining tools that are useful for providing unsupervised visual classification of multivariate data like GC data [27].