Multivariate statistical analyses on NMR data
Typical 600 MHz 1H NMR spectra for treated and untreated aqueous extracts obtained from olive leaves are reported in Fig. 1. Sugars and organic acids characterized the alkyl and hydroxyl-alkyl region (middle and low frequencies, from 5.5 to 0.5 ppm), whereas phenolic compounds are typical for the aromatic region (high frequencies, 9.0–6.0 ppm). Relevant 1H NMR data are reported in Table 2. The metabolites were assigned on the basis of 2D NMR spectra analysis (2D 1H Jres, 1H COSY, 1H–13C HSQC, and HMBC) and by comparison with published data [20,21,22].
In order to reveal a possible general data grouping of the samples, an unsupervised PCA analysis was applied considering separately the untreated and the DENTAMET® treated samples of the two classes, Ogliarola salentina and Cellina di Nardò cultivars (Fig. 2). The first PCA model is built with all the untreated Ogliarola salentina and Cellina di Nardò samples (Fig. 2a). In this model, the first two components give R
2 = 0.71 and Q
2 = 0.55, t and t accounting for 56 and 15% of the explained variance, respectively. The PCA model of Fig. 2b is built with all the treated Ogliarola salentina and Cellina di Nardò samples, and the first two components give R
2 = 0.71 and Q
2 = 0.59, describing the samples distribution in the bidimensional space defined by t and t (in this case accounting for 53 and 18% of the explained variance, respectively). Analysis of the PCA (t/t) score plots showed that only in the case of samples treated with DENTAMET®, a clear partition of data was observed with a good grouping according to the original cultivar (Ogliarola salentina and Cellina di Nardò). On the other hand, when all the CoDiRO-exhibiting plants (the untreated samples) were submitted to PCA analysis, no differences of metabolomic profiles appeared among samples. These results suggest that the effect of the presence of a pathogen, such as X. fastidiosa on the metabolic profile of CoDiRO symptomatology exhibiting plants samples, could be predominant with respect to the differences normally observed among the olive cultivars. Nevertheless, the presence of other external factors, abiotic or biotic, responsible for the lack of discrimination observed in PCA score plot (Fig. 2a) could not be excluded. On the other hand, the further observed discrimination after the treatment strongly suggests that CoDiRO complex could be responsible for the metabolic uniformity observed in Fig. 2a.
All the CoDiRO-exhibiting plants (considering at the same time treated and untreated samples and the two cultivars, Ogliarola salentina and Cellina di Nardò) have been studied by unsupervised PCA and supervised OPLS-DA analyses. The explorative unsupervised method (PCA) used for the whole data did not give clear group separation (data not shown), while the OPLS-DA model based on treated vs. untreated category as discriminating class (Fig. 3) produced a good descriptive but weak predictive model [one predictive and four orthogonal components give R
X (cum) = 0.82, R
Y (cum) = 0.60, and Q
2 (cum) = 0.24]. A first level of discrimination was also observed on the basis of the treatment applied to the samples (DENTAMET® treated vs untreated). The study of the variables responsible for the class separation observed in Fig. 3a could be determined by the analysis of the p(corr) in the S-line plot (Fig. 3b). Interestingly, by examining the loadings of the original variables a higher relative content of polyphenols, such as oleuropein and ligstroside and their derivatives (tyrosol and hydroxytyrosol) was observed for the treated sample. This resulted from the presence of signals in the aromatic region, at frequencies corresponding to tyrosol and hydroxytyrosol (6.84, 6.70, 3.78, 2.76 ppm and 6.90, 6.81, 6.70, 3.78, 2.76 ppm) and oleuropein and its aldehydic derivatives (6.04, 5.78, 3.89, 2.68, 2.5, 1.55 ppm and 9.22, 9.18, 6.04, 5.74 ppm). On the other hand, a higher relative content of sugars was observed for the untreated samples. This resulted from the presence of signals of anomeric protons of α- and β-glucose (doublets at 5.22 and 4.62 ppm, respectively) [20,21,22].
In order to deeply analyze the response of the CoDiRO-exhibiting plants to the treatment, the metabolic profile of treated and untreated plants was better characterized for each cultivar. In the first case, the unsupervised PCA analysis, applied to Cellina di Nardò samples resulted in no data clustering observation for the first two components, PC1 and PC2. Indeed, inspection of further components other than the first two was required (PC2 vs. PC4), in order to observe in the scoreplot a certain degree of samples clustering (see Additional file 1: Figure S2). Therefore, the supervised OPLS-DA analysis gave a good model [1 + 2 + 0, R
X (cum) = 0.725, R
Y (cum) = 0.728 and Q
2 (cum) = 0.53] with a clear partition between DENTAMET® treated and untreated samples (Fig. 4a). By examining the loadings of the original variables, the molecular components distinctive for each class could be determined. CoDiRO-exhibiting samples showed a lower polyphenol content for untreated with respect to treated samples. Interestingly, a relatively higher polyphenol content (with respect to other Salento cultivars) was observed for the for Cellina di Nardò EVOOs samples originating from healthy trees . In the case of Ogliarola salentina samples, the chemometric analysis of a matrix composed by a reduced number of 1H spectra showed a clear partition between DENTAMET® treated and untreated samples, as reported in the OPLS-DA score plot (Fig. 5a). The unsupervised method (PCA) gave unclear results (see Additional file 1: Figure S3), while the OPLS-DA analysis gave a good model [1 + 2 + 0, R
X (cum) = 0.786, R
Y (cum) = 0.837 and Q
2 (cum) = 0.489], showing a clear partition between DENTAMET® treated and untreated samples along the first predictive component (Fig. 5a). By examining the loadings of the original variables in the S-line plot (Fig. 5b), the molecular components distinctive for each class could be defined. In particular, infected Ogliarola salentina untreated plants showed a metabolic profile characterized by a higher content of polyphenol molecules. On the other hand, the DENTAMET®-treated infected plants were characterized by a higher sugar content. In this case, the observed polyphenols decrease, in treated with respect to control trees, is in accord with the polyphenols production associated to drought stress [24, 25], notwithstanding the levels of phenols are characteristic for each cultivar such as abiotic stress responses [26,27,28]. Interestingly, when unsupervised exploratory method (PCA) was applied a good grouping according to the original cultivar (Ogliarola salentina and Cellina di Nardò) resulted only in the case of DENTAMET®-treated samples while the differences normally observed according to the olive cultivars were not predominant in the case of untreated CoDiRO-exhibiting plants. Considering all the CoDiRO-exhibiting samples (obtained from both the Ogliarola salentina and Cellina di Nardò cultivars), supervised methods (in particular OPLS-DA analysis) were required to obtain the discrimination between untreated and DENTAMET®-treated samples. The different olive cultivars would seem to differently respond to the DENTAMET® treatments by altering their metabolic profiles in the sugars and polyphenols content. In particular, OPLS-DA analyses revealed that Cellina di Nardò CoDiRO samples showed a lower polyphenols and a higher sugar content for the untreated with respect to the treated ones. In contrast, in the case of Ogliarola salentina CoDiRO-exhibiting trees, DENTAMET®-untreated samples showed a higher content of polyphenol molecules while samples from treated infected plants were characterized by a higher sugar content. As already reported in literature [24, 25], physiological and biochemical responses to stress are closely cultivar-dependent. In the present case, since studied Ogliarola salentina and Cellina di Nardò cultivar trees were also characterized by a different level of pathogen attack, the observed changes in metabolic profiles due to DENTAMET® treatment could be also related to such a specific factor. Moreover in plants, both polyphenols and carbohydrates can play a relevant role during the pathogen infection. In fact, the downregulation of polyphenol-oxidase expression dramatically increased the susceptibility of tomato plants to Pseudomonas syringae pv. tomato . In olive, polyphenols could also play a direct and significant role in protecting the tree towards pathogen infections. In particular, oleuropein, as extracted from olive waste water, has been shown to be effective towards Pseudomonas savastanoi pv. savastanoi, the causal agent of olive knot disease, by inhibiting its growth . Also carbohydrate activation can be related to a number of stress that can disturb or subvert to normal plant metabolism such as a wound or a pathogen attack [31, 32]. It should be said, however, that the interplay occurring between polyphenols and carbohydrates during a pathogen infection has not been studied in detail. The metabolomic approach here preliminary applied to olive trees showing the CoDiRO symptoms could shed light into their interrelationships upon a pathogen attack.