Variation in the chemical profiles of three foxglove species in the central Balkans

Abstract

The aim of this study was to determine intra- and interspecies variation in the qualitative and quantitative composition of methanol-soluble metabolites in the leaves of three Digitalis species (D. lanata, D. ferruginea, and D. grandiflora) from the central Balkans. Despite the steady use of foxglove constituents for human health as valuable medicinal products, populations of the genus Digitalis (Plantaginaceae) have been poorly investigated to describe their genetic and phenetic variation. Following untargeted profiling using UHPLC-LTQ Orbitrap MS, by which we identified a total of 115 compounds, 16 compounds were quantified using the UHPLC(-)HESI-QqQ-MS/MS approach. In total, 55 steroid compounds, 15 phenylethanoid glycosides, 27 flavonoids, and 14 phenolic acid derivatives were identified across the samples with D. lanata and D. ferruginea showing a great similarity, while 15 compounds were characteristic only for D. grandiflora. The phytochemical composition of methanol extracts, considered here as complex phenotypes, are further examined along multiple levels of biological organization (intra- and interpopulation) and subsequently subjected to chemometric data analysis. The quantitative composition of the selected set of 16 chemomarkers belonging to the classes of cardenolides (3 compounds) and phenolics (13 compounds) pointed to considerable differences between the taxa studied. D. grandiflora and D. ferruginea were found to be richer in phenolics as compared to cardenolides, which otherwise predominate in D. lanata over other compounds. PCA revealed lanatoside C, deslanoside, hispidulin, and p-coumaric acid to be the main compounds contributing to the differences between D. lanata on one side and D. grandiflora and D. ferruginea on the other, while p-coumaric acid, hispidulin, and digoxin contribute to the diversification between D. grandiflora and D. ferruginea. However, quantitative variation in the metabolite content within species was faint with mild population diversification visible in D. grandiflora and particularly in D. ferruginea. This pointed to the highly conserved content and ratio of targeted compounds within the analyzed species, which was not severely influenced by the geographic origin or environmental conditions. The presented metabolomics approach might have, along with morphometrics and molecular genetics studies, a high information value for further elucidation of the relationships among taxa within the genus Digitalis.

Keywords: D. ferruginea; D. lanata; Digitalis grandiflora; UHPLC-LTQ OrbiTrap MS; cardiac glycosides; flavonoids; phenylethanoids; steroids.

Figure 1

Map presenting the populations of Digitalis grandifloraa, D. lanata, and D. ferruginea originating from the central Balkan Peninsula analyzed within the present study. For the population labels, please refer to Table 1 .

Figure 2

A simplified scheme of the lanatoside C hydrolysis that results in the formation of digoxin in two stages and involves the enzyme digilanidase and one deacetylation step (A). The conversion goes via either deslanoside (desacetyllanatoside C) or acetyldigoxin. D. lanata (a) is a rich source of lanatoside C, D. ferruginea (b) contains predominantly deslanoside and digoxin, while D. grandiflora (c) is abundant in digoxin. Correlation analysis of the compounds’ quantitative data was performed (B), and the correlation matrix was constructed using Pearson’s correlation algorithm. Color scale indicates a positive (blue) or a negative (red) correlation.

Figure 3

Principal component analysis (PCA) biplot with the first two PCs explaining 83.56% of total variance (A). For each species 95% confidence ellipses are presented: black − Digitalis grandiflora, red − D. lanata, green − D. ferruginea. Different symbols indicate affiliation of populations. Participation of the variables in the first two PCs is indicated by the corresponding vectors and by loading plots presented separately for PC1 (B) and PC2 (C). Variables protocatechuic acid , A; syringic acid, B; p-hydroxybenzoic acid, C; 5-O-caffeoylquinic acid, D; caffeic acid, E; p-coumaric acid, F; aesculetin, G; isoorientin, H; quercetin-3-O-glucoside, I; naringin, J; luteolin, K; hispidulin, L; isorhamnetin, M; digoxin, N; deslanoside, O; lanatoside C, P. For the interpretation of population abbreviations in the figure legend, please refer to Table 1 .

Figure 4

Principal component analysis (PCA) biplot with the first two PCs explaining 84.38% of the total variance (A) among D. grandiflora accessions. For fifteen D. grandiflora populations (each labeled by a different symbol) 95% confidence ellipses are presented. Participation of the variables (compounds) in the first two PCs is indicated by the corresponding loading plots. Linear discriminant analysis (LDA) (B) represents the same fifteen populations of D. grandiflora with 95% confidence ellipses. Heatmap of the scaled quantitative data (C) with the samples arranged according to the hierarchical cluster analysis (Ward’s method of cluster agglomeration). Compounds’ labels are the same as in Figure 3 . For the interpretation of population abbreviations in the figure legend, please refer to Table 1 .

Figure 5

Principal component analysis (PCA) biplot with the first two PCs explaining 83.90% of the total variance (A) among D. lanata accessions. For six D. lanata populations (each labeled by a different symbol) 95% confidence ellipses are presented. Participation of the variables (compounds) in the first two PCs is indicated by the corresponding loading plots. Linear discriminant analysis (LDA) (B) represents the same six populations of D. lanata with 95% confidence ellipses. Heatmap of the scaled quantitative data (C) with the samples arranged according to the hierarchical cluster analysis (Ward’s method of cluster agglomeration). Compounds’ labels are the same as in Figure 3 . For the interpretation of population abbreviations in the figure legend, please refer to Table 1 .

Figure 6

Principal component analysis (PCA) biplot with the first two PCs explaining 83.84% of the total variance (A) among D. ferruginea accessions. For seven D. ferruginea populations (each labeled by a different symbol) 95% confidence ellipses are presented. Participation of the variables (compounds) in the first two PCs is indicated by the corresponding loading plots. Linear discriminant analysis (LDA) (B) represents the same seven populations of D. ferruginea with 95% confidence ellipses. Heatmap of the scaled quantitative data (C) with the samples arranged according to the hierarchical cluster analysis (Ward’s method of cluster agglomeration). Compounds’ labels are the same as in Figure 3 . For the interpretation of population abbreviations in the figure legend, please refer to Table 1 .