The genus Rubus (brambles) is taxonomically complex in Europe due to apomixis, hybridization and the resulting proliferation of locally differentiated taxa. Apomictic reproduction produces viable seeds without fertilization and, combined with retained sexual reproduction in some lineages, promotes hybrid formation and morphologically similar taxa with unclear taxonomic status. This instability of ranks and frequent hybrid events—especially involving R. caesius and other sections including Sect. Rubus and R. idaeus—has made consensus classification difficult.
Phenolic compounds, tannins, flavonoids, phenolic acids and various terpenoids (including labdane-type diterpenes and ursane- and oleanane-type triterpene glucosides) are well documented across Rubus taxa and have traditional therapeutic uses. Despite this phytochemical knowledge, chemotaxonomic approaches for Rubus have been lacking.
This study tested whether leaf metabolomic profiles, obtained by untargeted metabolomics, can inform boundaries among Rubus taxa. Leaves were sampled over two years along an elevational gradient in the French Alps to maximize metabolic variability and assessed by LC-MS (UPLC-MS-ESI-QTOF) to determine whether morphologically assigned taxa share chemical signatures and whether specific metabolites can act as taxonomic markers.
Dried leaf material from several Rubus taxa in alpine environments was extracted in ethanol. Untargeted UPLC-MS analyses were performed in positive electrospray ionization (ESI) for all samples for both years. Negative ESI analyses were also performed on quality control samples and representative samples from each morphological group in 2024 to supplement spectral information for compounds that ionize better in negative mode, such as hydrolysable tannins and some glycosylated flavonoids. Data from the positive ionization mode were the primary basis for multivariate analyses.
Two yearly ionic intensity matrices were generated: one containing 566 features for 2023 and one containing 668 features for 2024. Multivariate statistics were applied to these matrices to explore sample clustering and discriminant features.
Multivariate analyses of the ionic intensity matrices revealed consistent clustering of samples into three overarching metabolomic supergroups. These corresponded to (1) R. idaeus samples, (2) taxa belonging to sections Corylifolii and Rubus, and (3) taxa related to section Caesii, which occupied an intermediate position between the other two groups. The grouping indicates that leaf chemical profiles can reflect larger taxonomic partitions within the sampled alpine Rubus taxa, despite morphological ambiguities caused by apomixis and hybridization.
The intermediate metabolomic position of section Caesii samples limited the ability to extract section-specific biomarkers for Caesii; its profiles overlapped with the other sections to a degree that precluded unique marker assignment.
Across both sampling years, 38 compounds were annotated from the metabolomic datasets. Among these, eight compounds were detected in the genus Rubus for the first time in this study. Annotation benefitted from combined positive-mode data and targeted negative-mode spectral information for certain compound classes. The annotated compounds span several chemical families known from Rubus studies, including phenolics and terpenoids.
Multivariate feature selection highlighted several compound families as discriminant between groups and therefore of potential taxonomic relevance. These include ursane triterpenoids, flavonoid glycosides, lignans (including neolignan glycosides), and gallotannins.
Specific patterns reported in the datasets include higher levels of quercetin- and kaempferol-3-O-methylglutaryl hexoside in R. idaeus samples, whereas neolignan glycosides were predominantly detected in taxa from sections Corylifolii and Rubus. Because section Caesii samples occupied an intermediate chemical space, the study did not propose exclusive biomarkers for that section.
These discriminant compounds provide candidate biochemical markers that could complement morphological and genetic data when assessing taxonomic boundaries within Rubus.
The findings demonstrate that untargeted metabolomics of leaf extracts can separate sampled alpine Rubus taxa into consistent chemical groups and identify discriminant metabolites. This approach offers a complementary perspective to morphology and genetics for resolving taxonomic ambiguities produced by apomixis and hybridization.
The authors emphasize the need for precise and detailed phytochemical descriptions across Rubus subgroups, both to clarify taxonomic relationships and because several of the compounds annotated have growing relevance in pharmacology and cosmetics. Robust chemotaxonomic characterization could therefore support both scientific classification and applied development of plant-derived compounds.
The minimal dataset and analysis code for the study are publicly available through a GitHub repository provided by the authors. Funding for the work was provided by the Research and Technology National Association (ANRT) under a CIFRE project (number reported in the source). The authors declared no competing interests. Specific methodological details, full lists of annotated compounds and figure data are available in the published article and supporting information.