Leafy galls (LG) were induced in four-week-old in vitro growing tobacco plants following infection with the virulent R. fascians strain D188. Figure 1 illustrates the phenotype of the non-infected (NI) tobacco plants 11 weeks post germination and a typical eight-week-old LG, a morphological structure with multiple dormant buds and malformed leaves characterized by increased trichomes formations.

As shown in Figure 2, visual inspection of the ¹H-NMR spectra of non-polar metabolites (chloroformic extracts) revealed roughly similar chemical shift patterns in the aromatic region (δ 7–8 ppm) but striking differences in both the olefinic (δ 4.9–6.5 ppm) and aliphatic (δ 0–3 ppm) regions. According to Qin [21], the methyl signals located at δ 0.7–1.3 ppm might originate from steroids or terpenoids and, at δ 1.2–1.4 ppm, from fatty components. Thus, the significant signals observed at δ 1.5–3.0 ppm and at δ 5.0–5.5 ppm might be assigned to methylene group from sesquiterpenes or diterpenes and to olefinic signals from fatty components, steroids, or terpenoids, respectively. These data suggest an alteration in biosynthetic pathways leading to the accumulation of terpenoids and/or fatty compounds to the detriment of the metabolites diversity in infected tobacco cells. In contrast, a visual observation of the ¹H-NMR spectra of polar metabolites (aqueous extracts) revealed no obvious differences in chemical shifts between R. fascians infected (LG) and non-infected (NI) tobacco plants (Figure 3).

As shown in the score plot (Figure 4a), the first two principal components explain 99.7% of the variability in the dataset (PC1: 78.6% and PC2: 21.1%) and clearly identify two different clusters of samples. These PCA scores were tested by a one-way-ANOVA (Table 1), indicating significant differences between R. fascians-infected and non-infected tobacco metabolites profiles (p < 0.001). According to the mapping metabolite approach described by Graham [22], the PC1 component allows an obvious discrimination between extracts of R. fascians-infected and non-infected tobacco plants. Examination of the loadings plot (Figure 4b) shows that the variance along the first principal component is mainly driven by modifications in the aliphatic (δ 0.7–3 ppm) and the olefinic (δ 5.0–5.5 ppm) regions. These PC1 discriminating signals confirm that low-polarity secondary metabolites, possibly steroids or terpenoids, could account for some major changes occurring in tobacco plants following R. fascians infection. To further support this assumption, qualitative GC-MS analysis of chloroformic extracts from both LG and non-infected plants was investigated. Figure 5 shows major differences in the GC chromatograms, particularly in compounds eluted from 45 to 60 min., which, according to Hamm [23], correspond to tobacco diterpenoids, such as cembrenoids, as illustrated by the retention time of commercially available cembrene (at 49.1 min) which are found to be accumulated more than two times in LG as compared to NI extracts. It is noteworthy to mention that cembrenoids and closely related metabolites are found to accumulate in tobacco trichomes [24], a characteristic anatomical feature of leafy galls (Figure 1). It is therefore tempting to postulate that, diterpene biosynthetic pathway and particularly cembrenoids family is stimulated in infected tobacco tissue.

The non-polar metabolites fingerprinting reveals that the R. fascians infection induces the expression of secondary metabolites with aliphatic signals consistent with the induction and accumulation of terpenic compounds. The qualitative GC analyses of the LG and NI tobacco extracts, according to a procedure described in the literature for the profiling of terpenoids [23], indicates that induced compounds could belong to the tobacco diterpene family. This confirms a previous observation by transcriptomic and BiNGO (Biological Network Gene Ontology) analysis of up-regulated genes in R. fascians infected-Arabidopsis that revealed an up-regulation of the plant host diterpenoid pathway [12]. To date, the role of terpenoid compounds in the R. fascians-plant interactions remains unknown, although some tobacco-pathogen interactions are known for leading to alterations in the terpenes pathway. The induction of the monoterpene E-(b)-ocimene and the sesquiterpenes caryophyllene, β-elemene and α-farnesene, have been shown during the interaction of Nicotiana tabacum with Golovinomyces cichoracearum [25] and Pseudomonas syringae [26], respectively. Seo [27] identified a diterpene, the (11E,13E)-labda-11,13-diene-8α,15-diol, as an endogenous signal for the activation of tobacco defense responses to wounding and to infection with the tobacco mosaic virus; many terpenoids are known for their function as phytoalexins, contributing to plant defenses against microbes [28].

As shown in Figure 6a, the first two principal components explain 81.5% of the variability in the dataset (PC1: 62.2% and PC2: 19.3%) and could not identify samples clusters. PCA scores were tested by a one-way-ANOVA (Table 1), indicating non-significant difference between R. fascians-infected and non-infected tobacco metabolites profiles (p > 0.05). Examination of the PCA loadings plot (Figure 6b) shows that the variance along the first principal component is mainly driven by modifications in the region δ 2.0–5.5 ppm. To avoid any biased interpretation of group discrimination based on PCA analysis, the dataset was reanalyzed using OPLS-DA, a supervised method based on an a priori knowledge of distinct samples.

As shown in Table 2, the comparisons between samples in models gave better predictions in OPLS-DA compared to PCA analysis (predictability of the total model: Q² OPLS-DA > Q² PCA; 0.93 and 0.65, respectively). OPLS-DA facilitates the identification of the different sources of variation that contribute to the differences between the extracts, by separating inter-group variation (i.e., variation that is predictive of differences between LG and NI tissues) from intra-group variation (i.e., variation that is unrelated to group separation). As shown in Figure 7a, a part of the total variance (predictive component, Cp 46.1%) is predictive, being specifically related to the differences induced by R. fascians infection. In contrast, the first orthogonal component (Co) describing 19.4% of the total variation in the dataset is responsible of the high variability within the dataset. The discriminating signals were interpreted through a loading S-plot (Figure 7b) that allows distinguishing of both the covariance (contribution of the effect) and the correlation (reliability of the predicted variables) for the identification of the major contributors in samples discrimination. In this OPLS model, increased metabolites correspond to those that show a high (i) correlation with the predictive scores, (p(corr) > 0.5) and (ii) co-variation with the predictive scores (p > 0.05), with respect to the component scores model. At these cut-off levels, only the metabolites identified by the OPLS-DA model were considered as significant contributors [29]. In OPLS-based regression, VIP values (variable importance in the projections) are directly proportional to the influence of variables on the separation on score plots. Variables with higher VIP values are more important for the samples discrimination. In the OPLS-DA analysis of LG and NI tobacco plants extracts, VIP values for the main metabolites, responsible for separation on the score plot, are represented in Figure 7c and Table 3. The high VIP scores identified for proline, inositol, malic acid, and caffeoylquinic acid, legitimate their occurrence in the well-developed LG, as compared to NI tissues.

The polar metabolites fingerprinting shows that metabolites discriminating LG from NI tobacco plants are likely primary metabolites, such as sugars (raffinose, galactitol), amino acids (proline, glutamine), and some organic and phenolic acids (malic acid, caffeoylquinic acid derivatives), suggesting a role of these compounds in the established ecological niche of R. fascians. Synthesis of primary metabolites seems to be a quite general mechanism, observed during plant-pathogen interactions [31,32], and the response of N. tabacum is probably not specific to this interaction. The increase in sugar contents in LG may reflect the energy costs linked to the activation of the plant defenses [33] or a redirection of the host plant metabolism, presumably to the R. fascians own advantage [12]. The increase in some amino acids, such as proline and glutamine, may contribute to the pathogenicity of R. fascians by providing appropriate conditions for the expression of virulence factors [12]. Interestingly though, the third main discriminatory class of molecules emerging from our study are organic acids and phenols (malic acid, caffeoylquinic acid, and kaempferol), pointing to a possible defense reaction occurring in LG; this is consistent with the induction of defense-related genes observed in developed LG of Atropa belladonna, tobacco and Arabidopsis [12,34,35]. Indeed, the phenylpropanoid pathway is known to be induced in response to a pathogen attack [36], and several phenolics are known for their potential anti-bacterial activities. For instance, caffeoylquinic acid derivatives were recently identified as antimicrobials and inhibitors of efflux pump of Gram-positive pathogenic bacteria [37].

The occurrence of antiproliferative activity in LGs and NI was assessed in crude water, hexane, methanol, and chloroform extracts by determining the concentration of extract required to obtain a 50% inhibition (IC₅₀) of the proliferation of four different cancer cell lines, U373, A549, MCF-7, and PC-3 using MTT assays. The hexane, methanol, and water extracts of the LGs and the NI extracts did not affect any of the cell lines (Table 4). In contrast, the chloroformic extract of the LGs, but not of NIs, exhibited a distinct antiproliferative activity against all four tested cell lines, albeit to different extents.

The proliferation of line U373 was affected most strongly (IC₅₀ = 58.9 ± 2.0 μg/mL), followed by that of line A549 (IC₅₀ = 69.8 ± 0.9 μg/mL), whereas lines MCF-7 and PC-3 had a lower but comparable sensitivity for the crude extract (IC₅₀ = 84.5 ± 4.3 μg/mL and 85.6 ± 1.0 μg/mL, respectively).

Because the glioblastoma cell line U373 exhibited the strongest sensitivity toward the LG extract, this line was chosen for all further experiments. As a first step in the chemical characterization of the bioactive compound(s) present in the LGs, the crude chloroformic extract was fractionated and the IC₅₀ was determined for each fraction. As shown in Figure 8, a first chromatographic fractionation yielded two out of seven fractions with IC₅₀ values below the value obtained for the crude extract (F2 and F3). These fractions were subjected to a second chromatography fractionation allowing the isolation of three active sub-fractions (F2.2, F2.3 from F2, and F3.1 from F3). These preliminary results on the inhibition of human tumor cell growth indicate that fully developed leafy gall, could represent a potential source of new bioactive compounds.

Because none of the tested extracts from NI exhibited antiproliferative activity, the biological activity observed in the chloroformic extract of LGs has to be associated with the reprogramming of the development and metabolism of the host, essentially the diterpenoids family (Figure 5) since such compounds have already been reported to inhibit cell proliferation [38].

Analytical grade chloroform was purchased from Chemlab (Redu, Belgium). CDCl₃ (99.96%) and D₂O (99.0%) with buffering agent (pH 6.5) were purchased from Sigma Aldrich (St. Louis, MO, USA).

Tobacco plants (Nicotiana tabacum L. cv. Petit Havana) were aseptically grown from sterile seeds (obtained at the Jean Massart Botanical Garden from the Université Libre de Bruxelles) on solid half-strength MS medium (Duchefa), supplemented with 3% sucrose, in a growth chamber at 23 °C in a 16-h light photoperiod (70 μmol photonsm⁻²s⁻¹). Four week-old plants were injured and infected at the apical bud with 10 μL of R. fascians virulent cultures strain D188 according to Rajaonson et al [39]. The leafy galls (LG) that formed were harvested eight weeks after infection. Non-infected (NI) tobacco plants were harvested after twelve weeks of growth (at the same developmental stage as the infected plants).

The human U373 (ATCC code HTB-17) glioblastoma, A549 (DSMZ code ACC107) non-small cells lung carcinoma, MCF-7 (DSMZ code ACC115) breast cancer and PC-3 (DSMZ code ACC465) prostate cancer cell lines were obtained from the American Type Culture Collection (ATCC; Manassas, VA, USA) and maintained in our laboratory as described previously [40].

All spectra were recorded on a Brüker AV-300 NMR spectrometer operating at a proton NMR frequency of 300.13 MHz. For each sample, 128 scans were recorded following the conditions described by Choi [30]. The resulting spectra were automatically phased, baseline-corrected and manually calibrated at 0.0 ppm using residual solvent signals in MestReNova v8.0.2 (Mestrelab Research, Santiago de Compostela University, Santiago de Compostela, Spain).

The ¹H-NMR spectra were aligned and automatically reduced to ASCII files using MestReNova. The data were reduced to integrated regions of equal width (0.04 ppm) over the region δ 0.0 to 12.0 ppm and normalized to the total spectral area in order to suppress trivial separation based on variations in amount of sample. The regions δ 7.26 and δ 4.1–5.3 ppm that correspond to solvents residual signals were excluded from all spectra used for analysis of chloroformic and aqueous extracts, respectively, in order to not compromise the analysis. Principal component analysis (PCA) and orthogonal projections to latent structures discriminant analysis (OPLS-DA) were performed with the SIMCA-P+ software, version 12.0.1 (Umetrics, Umeå/Malmö, Sweden). The degree of separation for each group of data was determined from the scores plot, and the signals that contribute to the composition were obtained from the factor loadings plot. A covariance matrix, generated by the mean-centered method without scaling (subtraction of the column mean from each data point), was used for analysis. OPLS-DA was carried out between R. fascians-infected and non-infected tobacco plants to identify metabolites affected by infection. To facilitate interpretation of OPLS-DA a predictive component was calculated after removing the variation explained by orthogonal components and, correlation coefficient [p(corr)] were computed for each bin and represented by S-plot. Metabolites (bins) showing p(corr) > 0.5 were then considered significant. For all plots, a hostelling 95% confidence ellipse was drawn. Metabolites indicated in Table 3 and Figure 7b were assigned on the basis of chemical shifts and identified from spectral databases and bibliographic search for the chemical shifts of pure compounds putatively present in the ¹H-NMR spectra of the extracts [21,30].

GC-MS analyses were performed on a Thermo ITQ 900 system (Thermo Fisher Scientific Inc., Hudson, NH, USA) equipped with an Agilent DB5 capillary column of 30 m × 0.25 mm with a phase thickness of 0.25 μm (Agilent Technologies, Santa Clara, CA, USA). The injector temperature was set to 250 °C, the transfer line to 250 °C and the source to 150 °C; the temperature program was 40 °C for 1 min, 9 °C/min increase rate up to 130 °C, followed by a 2 °C/min increase rate to 230 °C as described by Hamm [23]. The carrier gas was helium, with a column head pressure of 10 psi. The chloroformic and commercially available cembrene (Accros Organics, Geel, Belgium) samples (50 μg/mL) were injected in the splitless mode (2 μL injected, 1 min). The mass spectrometer was operated in electron-impact mode (EI) at 70 eV, in the scan range m/z 29–400.

The colorimetric MTT viability assay (3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide; Sigma, Diegem, Belgium) was used to determine the overall growth level of each cell line at 72 h as described previously [41] (n = 3 with 6 replicates for each concentration).