Mass Spectrometry Map Reveals Hidden Chemical Diversity Across the Cabbage Family

Plants cannot run from their enemies, so they fight with chemistry. The mustard family, Brassicaceae, is famous for the pungent sulfur compounds that give mustard, cabbage, and wasabi their bite, but those well-known defenses are only the visible tip of an enormous biochemical iceberg. A new study published in the journal Metabolomics has now mapped the specialized metabolites of fourteen crucifer species in a single, standardized framework, revealing hotspots of chemical novelty that had never been systematically compared before. Led by Felicia C. Wolters of Wageningen University & Research, together with colleagues including Marnix Medema, Klaas Bouwmeester, and Justin van der Hooft, the work combines liquid chromatography tandem mass spectrometry with a battery of computational annotation tools to ask a deceptively simple question: which chemicals do all these related plants share, and which are unique?

The scale of the undertaking is what sets it apart. The team grew fourteen species under tightly controlled greenhouse conditions, harvested both leaf and root tissue at the same developmental stage and time of day, and extracted metabolites using a single, reproducible protocol. Each batch was then run on the same high-resolution Orbitrap mass spectrometer in both positive and negative ionization modes, capturing semi-polar to moderately polar compounds across a mass range of roughly 90 to 1350 daltons. From the raw spectra, the researchers extracted 7,683 mass features in positive mode and 6,433 in negative mode using the software mzmine. Mass features are the detectable ionic fingerprints of individual compounds in a complex mixture, and aligning them across species is the foundation of any comparative metabolomics study.

Extracting features, however, is only half the battle. The central challenge of untargeted metabolomics is that most detected features cannot be matched to known compounds in spectral libraries, and different computational tools often disagree. To overcome this, the team built a consensus annotation pipeline that integrates several state-of-the-art in silico tools: SIRIUS with CANOPUS for molecular formula and chemical class prediction, MS2Query for spectral analogue searching, and the deep-learning tool DreaMS queried against the MassSpecGym library. Only features whose compound class annotations agreed across multiple tools were retained, yielding 5,118 consistently annotated features in positive mode and 2,542 in negative mode. Chemical classes were assigned using the NP.Classifier ontology, a deep-learning-based classification system for natural products. This consensus approach does not deliver unambiguous structural proof, the authors caution, but it makes chemical ontology assignments considerably more reliable than any single tool alone.

With the annotated dataset in hand, the researchers could define what they call the core metabolome of the family. The Brassicaceae phylogeny was recently revised into two subfamilies and five supertribes, and the study focused on the two largest, Camelinodae and Brassicodae, with the early-diverging species Aethionema arabicum as an outgroup. The results showed that only a small fraction of the chemical space is truly conserved: just 2.9 percent of annotated features in positive mode and 1.5 percent in negative mode were shared across all fourteen species. In contrast, the two supertribes shared 13 percent of features in positive mode and 19.2 percent in negative mode, while species-specific features ranged from 7 to 45 percent of the total fingerprint. Conserved profiles broadly tracked phylogenetic distance, suggesting that chemistry, like genes, carries an evolutionary memory.

The most striking findings came from the chemical mavericks. Camelina sativa, the oilseed crop known as false flax, Capsella rubella, and Barbarea vulgaris, a bitter leafy herb, stood out with exceptionally unique metabolite profiles. In Camelina sativa, unique terpenoid features accounted for 33.5 percent of annotations in positive mode and a remarkable 55.6 percent in negative mode, the largest unique terpenoid profile recorded in the study. Terpenoids are one of the largest and most structurally diverse classes of plant natural products, and their abundance in Camelina hints at biosynthetic pathways that have gone largely unexplored in this crop. The finding could have practical consequences for breeding oilseed crops with enhanced pest resistance or novel nutritional chemistry.

Perhaps the most intriguing discovery concerns triterpenoid saponins, soap-like compounds with a characteristic oleanane-type carbon skeleton. Until now, these molecules were believed to be essentially exclusive to Barbarea vulgaris within the Brassicaceae, where they act as feeding deterrents against the diamondback moth. Using MS2LDA, a topic-modeling approach that extracts recurring fragmentation patterns called Mass2Motifs from tandem mass spectra, the team identified substructure motifs corresponding to oleanane and ursane triterpenoid cores and their conjugated glycosides. These motifs were detected not only in Barbarea but across leaf and root tissue of five Camelinodae species, in Brassica carinata, and even in the outgroup Aethionema arabicum. Characteristic fragment peaks at mass-to-charge ratios of 439.36, 119.09, and 105.07, matching library spectra of oleanolic acid, appeared in species where such saponins had never been reported before.

The authors are careful about interpretation. Some of the glycosylation-related fragmentation patterns required careful scrutiny because in-source fragmentation, in which molecules break apart inside the instrument before mass analysis, can masquerade as genuine sugar-loss signatures. Molecular families with highly similar retention times flagged this phenomenon in the Camelina data. Nevertheless, the convergence of multiple independent lines of evidence, including compound class predictions, spectral library matches, and substructure motifs, makes a compelling case that triterpenoid saponin-like chemistry is far more widespread in the crucifer family than decades of targeted phytochemistry had suggested. The result is a textbook example of how untargeted, computationally guided approaches can overturn long-held assumptions about the distribution of natural products.

The study also probed alkaloids, finding that tryptophan-derived alkaloid annotations were enriched across all species, while anthracillic acid alkaloids were particularly abundant in Isatis tinctoria, the historic dye plant woad, and ornithine alkaloids dominated in Aethionema arabicum. Molecular networking further suggested the presence of the phytoalexin camalexin, previously known mainly from Arabidopsis thaliana and Camelina, in the root tissue of Capsella rubella. Indole ring substructure motifs were widely distributed through the network, and the researchers propose systematically fingerprinting closely related species in the Arabidopsidae and Camelinae tribes to determine how specific camalexin production really is within the supertribe.

To test whether chemistry can reconstruct evolution, the team extended the analysis to seventeen species by integrating a second batch of LC-MS/MS data using the MS-Cluster algorithm, which computes consensus spectra while ignoring retention time shifts between batches. After filtering for high-confidence structural annotations from SIRIUS, 895 unique structures remained. Hierarchical clustering of these annotations, combining leaf and root profiles, largely reproduced the species phylogeny, with Brassicodae and Camelinodae species falling into distinct groups. Notable exceptions were informative: Lepidium sativum consistently grouped with Isatis tinctoria, hinting at phytochemical relatedness within the Lepidae tribe, while Camelina sativa persistently appeared as an outgroup in every clustering scenario, an enigma the authors suggest could be resolved by sampling more Camelinodae species. Tissue-specific clustering also revealed that leaf and root profiles can be conserved to different degrees, consistent with the idea that different organs face different selective pressures from herbivores, pathogens, and microbes.

Beyond its specific findings, the study offers a blueprint. The authors argue that current estimates of the total number of unique plant compounds are extrapolated from a biased sample of species and ignore phylogenetic distance, and they advocate clade-wide, standardized metabolic fingerprinting to correct this bias. Because specialized metabolites are often stress-induced, they note that a single snapshot under benign growth conditions inevitably underestimates the biochemical repertoire, and future designs should systematically elicit hormone-mediated stress responses. The modular pipeline, from growth chamber to consensus annotation, is extendable to other plant families, and all data have been deposited in public repositories including MetaboLights and MassIVE. For a family that supplies much of the world’s vegetables, oils, and model organisms, the message is clear: the chemical frontier of the Brassicaceae is far wider, and far stranger, than anyone had measured.

Subject of Research: Comparative metabolomic fingerprinting of specialized metabolite diversity across Brassicaceae species

Article Title: Systematic mass-spectrometry-guided metabolic fingerprinting elucidates diversity of specialized metabolites across the Brassicaceae

Article References: Wolters, F. C., Woldu, T., Schranz, M. E., Medema, M. H., Bouwmeester, K., & van der Hooft, J. J. J. (2026). Systematic mass-spectrometry-guided metabolic fingerprinting elucidates diversity of specialized metabolites across the Brassicaceae. Metabolomics, 22(5), Article 161. https://doi.org/10.1007/s11306-026-02537-y

Image Credits: AI Generated

DOI: 10.1007/s11306-026-02537-y

Keywords: Brassicaceae, metabolomics, mass spectrometry, specialized metabolites, triterpenoid saponins, chemotaxonomy, LC-MS/MS, molecular networking, Camelina sativa, plant biochemistry, computational annotation, phylogenetics