https://orcid.org/0009-0009-4450-9523
Figures
Abstract
The development of a robust taxonomy for the genus Rubus remains a real challenge, due to apomictic reproduction and hybridization that led to a very large number of “taxa” with undefined frontiers. Using an untargeted metabolomic approach and a two-year leaf sampling along an elevational gradient in the French Alps we provide new tools to assess boundaries between nine Rubus taxa. Multivariate analyses on ionic intensity matrices (of 566 and 668 features for 2023 and 2024 sampling respectively) have shown that R. idaeus species, the two sections Corylifolii and Rubus, and Caesii section related taxa, formed three distinct super groups. Thirty-eight compounds have been annotated, including 8 detected in genus Rubus for the first time. Discriminant compounds have been annotated, allowing to propose ursane triterpenoids, flavonoids glycosides, lignans and gallotannins as taxonomic classifiers of the genus Rubus. Among those, quercetin – and kaempferol – 3-O-methylglutaryl hexoside were detected at higher levels in R. idaeus, whereas neolignans glycosides have been mostly detected in sections Corylifolii and Rubus. The intermediate position of Caesii between the two others did not allow to propose specific biomarkers of the section. The need for a precise and detailed description of the phytochemical profiles of the various Rubus subgroups is reinforced by the growing use of some of these compounds in pharmacology and cosmetics.
Citation: Serviole E, Henriot A, Gallet C, Audoin C (2026) Untargeted metabolomics as a new perspective for assessing taxonomic boundaries in alpine Rubus genus. PLoS One 21(8): e0354840. https://doi.org/10.1371/journal.pone.0354840
Editor: Chun-Hua Wang, Foshan University, CHINA
Received: April 16, 2026; Accepted: July 13, 2026; Published: August 5, 2026
Copyright: © 2026 Serviole et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The minimal data set and accompanying code, where generated, is available at https://github.com/e1000ser/Metabolomics-analysis-of-wild-Rubus-leaves.
Funding: all the funding or sources of support (whether external or internal to your organization) received during this study, were provided by the Research and Technology National Association (ANRT) under the CIFRE project n°2022/1240. There was no additional external funding received for this study.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Rubus L. (brambles), part of the Rosaceae family has one of the widest distribution range among the flowering plants and is particularly well represented in Europe [1–3]. Some of its representatives have been largely exploited for their edible fruits such as raspberries and blackberries (R. idaeus and R. fruticosus aggr. y.). The genus Rubus is a complex group in terms of taxonomy. The number of species to be considered within this genus varies according to the authors and the areas studied but remains in all cases substantial and highly controversial.
The lack of consensus regarding the taxonomy of brambles is mainly due to apomixis, which affects speciation and is the cause of the multiplication of local taxa [4–7]. Apomixis is an asexual reproduction mode producing viable seeds from maternal ovule tissue without fertilization [8], with multiple types (strict or facultative) involving diverse cellular mechanisms [9]. Retention of sexuality in some of these species [10] can promote hybridization phenomena that complicate the classification of the genus to which they belong [6] by producing a large number (often hundreds) of taxa, morphologically similar and that behave biologically like species without consensus on their taxonomic classification [6,11]. The instability of the taxonomic ranks used in classification [11], and the presence of regional or even local taxa [6], result in either an underestimation of the real number of taxa by using aggregates rank, or on the contrary an overestimation by using the rank of subspecies.
The genus Rubus is considered to be one of the most difficult in the European flora in terms of taxonomic classification, which has led to the production of limited and confusing literature [1,3,12–15]. The multitude of taxa has encouraged the batologists (botanists specialized in brambles) to reduce the number of taxa to be considered [1,5,6]. Although the complex state of bramble taxonomy, the following hierarchical classification is the most widely used nowadays: Genus; subgenus; section; sub-section; series; species (Fig 1A). This study concerns the five subgenera generally described in European Rubus, and the diverse and numerous taxa they include (Fig 1A). The organization within this reduced range of Rubus taxa presents some taxonomical divergences and complexities, such as the frequent hybridization events between R. caesius (Sect. Caesii) and other sections: mainly Sect. Rubus [3,5,12,21] but also with the species R. idaeus [4,14], creating overlapping morphological criteria of identification and unstable classification concerning those taxa. Moreover, some authors consider other subgenus, such as Anoplobatus, Chamerubus, Cyclactis that sometimes encompass species belonging to the sections mentioned before [22,23] and tend to blur the taxonomical boundaries.
A: Adapted taxonomic classification of European Rubus taxa from works of [4,5,16–20]. B: Repartition of 8 morphological groups in 3 super groups (underlined) defined by metabolomics analysis. X represents sample group with no morphological assignment.
Phenolic compounds, including tannins, various flavonoids and phenolic acids have been largely described in different parts of the plants in all Rubus taxa investigated [24]. Terpenoids have also been described, including labdane-type diterpenes ursane- and oleanane-type triterpene glucosides, suaviosides and triterpenoids [25–27]. Those compounds probably explained the traditional and worldwide traces of therapeutic uses of Rubus leaves, as infusion to treat numerous ailments [28,29]. Still, no attempt to use such compounds for chimiotaxonomic purposes have been found.
Recently, some studies have shown that metabolomics could provide relevant data to better understand and clarify taxonomic issues within genus such as Myrothamnus, Viola and Ocotea [30–33]. We therefore question the links between the taxonomic complexity and the diversity of chemical profiles in Rubus taxa, using leaf samples from alpine environment that maximize the metabolic variability [34–37] analyzed by untargeted metabolomics (LC-MS-ESI-QTOF). Using a two-year sampling of bramble leaves along an elevational gradient in the French Alps, we ask whether morphologically identical taxa share similar chemical profiles and whether specific metabolites can serve as taxonomic markers within Rubus.
Results and discussion
Multivariate analysis
Ethanol extracts were prepared from dried leaves and UPLC-MS based metabolomic analysis was performed for all samples for both years separately, in positive electrospray ionization (ESI). A similar analysis in negative electrospray ionization was conducted separately on QC samples, and one sample from each taxonomic group (according to morphological identification) from 2024’s sampling, to obtain supplementary spectral informations for compounds which may have a better negative ionization, such as hydrolysable tannins and (glycosylated) flavonoids. Only the data obtained in positive electrospray ionization were processed in Mzmine to obtain ionic intensities matrices for each year (2023: 53 x 566; 2024: 192 x 668).
The principal component analysis (PCA) (Fig 2), the Partial Least Squares – Discriminant Analysis (PLS-DA) (Fig 3), and the hierarchical cluster analysis (HCA) (Fig 4) on both 2023 and 2024 matrices showed segregation into three groups which have been respectively assigned to 3 super groups: ID, C, and RC (Fig 1B). Those groups respectively correspond to the Idaeobatus subgenus (ID), the Caesii section (C), and the two sections Rubus and Corylifolii (RC).
The ID super group consisted of R. idaeus species which were classed in id (2023 and 2024) and id_D (2024) group according to morphological identification. Samples from super group ID were clearly segregated from other samples in all analyses for both years. Projection in PCA and PLS-DA showed a clear separation of the super groups along the first axes for each year (Figs 2, 3) and HCA’s first branch division splits ID supergroup from other samples (Fig 4) displaying a strong clustering of R. idaeus species according to Euclidean distances measures based on the ionic intensity of each sample. Both years show the same tendencies with the R. idaeus (ID supergroup) striking separation against the other taxa. Even if this species can be easily morphologically distinguished, our results support the taxonomic and evolutive distance of R. idaeus from other Rubus taxa already studied [19,38] and confirms their metabolomic divergences.
C super group consists of the morphological group ca (R. caesius species) (2023 and 2024) and the Ca_Ru and Co_Ru_ca groups (2024) that couldn’t be properly identified with morphological analysis, but both show morphological characteristics affiliated to the Caesii section and more precisely to R. caesius species. The third super group RC included the morphological group from the subsection Hiemales (Hi) part of Rubus section (Fig 1), and Corylifolii section (Co, Co_Ru) as well as the x group that couldn’t be identified. PCA and PLS-DA 2024’s projections showed a clear segregation of C super group along first axis, samples being located in the center between ID and RC (Hi, Co, Co_Ru and x). A notable difference between the two years is that in 2023 Co_Ru groups seemed to be more separated from the Hi group than from the ca group (Figs 2, 3). Multivariate analysis showed that C super group was segregated from RC and ID with a relevant clustering displayed by HCA analysis (Fig 4). Trees from both years (Fig 4A and 4B) showed a clear clustering of ID and C super groups. The remaining part of the trees contained all the samples from the Hi, Co and Co_Ru groups. In the 2023 tree, Hi group split among the Co group (Fig 4A) whereas in Fig 4B the Co and Co_Ru groups were well clustered together aside from the Hi samples. Finally, the x group from 2024 data was well clustered among the samples belonging to the Hi group (Fig 4B). The separation between the RC and C groups supports the taxonomical choice to separate the Caesii (C) and Corylifolii (part of RC group) sections. Our results also showed that, despite some slight differences, the Rubus and Corylifolii sections were hardly distinguishable in terms of morphological analysis as well as metabolomics analysis which is consistent with the frequent hybridization events reported in literature [5,12,21,19,39]. The variability observed within the super group RC between the two sampling years could be attributed to the complexity of morphological identification.
Multivariate analyses suggested robust trends, considering the cumulative explication percentage of both axes were 49, 43, 49 and 32% for PCA and PLS-DA projections from Figs 2A, 2B, 3A and 3B respectively. Moreover, the similarities of observed tendencies between two years data reflect a meaningful repeatability of the experiment. All results showed a consistency between the taxonomy of the genus Rubus based on morphological analysis and the chemical profile of the specimens. Indeed the R. idaeus species (id group), the R. caesius species (ca group) and related (Ca_Ru and Co_Ru_ca groups) were clearly separated from the RC group (Rubus and Corylifolii sections). Among the RC group, we observed a trend on which individuals belonging to the Coryfolii section and related (Co and Co_Ru groups) differed from the individuals of the Subsection Hiemales (Hi group). The observed trends in multivariate analysis results were consistent with the taxonomic ranking of Rubus genus classification (Fig 1).
Globally, multivariate analysis suggests strong phytochemical profile identities of taxonomic ranks within the genus Rubus, considering that samples collected along a large environmental gradient were clustered according to their morphological identification rather than by their sampling station.
Compounds annotation
Considering the similarities of trends between 2023 and 2024 and the greater number of stations and samples in the 2024 dataset, this latter was chosen for annotation.
Molecular network
To have a global visualization of the features structural similarities based on their MS/MS fragmentation pattern, we built on Cytoscape a molecular network (Fig 5) of 668 nodes whom 349 were included in 33 clusters and 319 were duets and single nodes. The latter were not included in Fig 5 but the whole molecular network is available in supplementary (S1 Fig.).
Color tag is based on super groups: Purple = ID, Red = C, Orange = RC; chart’s proportions were calculated by ionic intensity average of each feature in the three groups.
Distribution of the three super groups in the molecular network
Average ionic intensity (AII) of the features in the ID, RC and C super groups has been added in each node. Except for C7, C14, C27, C32, AIIs’ tendencies in favor of one or two groups have been observed in every cluster. Most nodes of cluster C1 exhibit higher AIIs for C and RC groups, as in the upper zone of the cluster C2 (C2a). In contrast, most of the nodes of the bottom part (C2b) exhibit higher AII for the ID group. Cluster C3 can also be divided into two parts: on the left nodes have an AII higher in the ID group, whereas in the right part nodes have an AII higher in the RC group. The following clusters are all characterized by a higher representation of the ID group: C4, C5, C6, C9, C15, C18, C21, C25, C28 and C33. Clusters C12, C16 and C24 were specific of RC, whereas Clusters: C22, C26 and C30 were specific of C. Finally, C10, C11, C13, C17, C20 and C31 have shown higher AIIs in RC and C groups, and ID and C specific clusters were C8, C19, C23 and C29.
Annotation & propagation
Thirty-eight compounds were putatively annotated using spectral databases FragHub, homemade experimental library, GNPS, Fiora-fragmented Lotus and data literature (Table 1), and included mainly triterpenoids, lignans, flavonoids, and gallotannins. Compounds listed in Table 1 are level 2 annotations, while level 3 annotations were proposed using propagation based on molecular network clustering and MS/MS fragmentation [80]. Detailed mass spectrometry information for identification was provided as Supporting Information (S2-S3 Figs.). Annotated compounds were placed on the molecular network (Fig 6), putative structures were based on MS and MS/MS spectra and correspond to the first hit in the FragHub database.
Compound numbers refer to Table 1. Color tag is based on super groups: Purple = ID; Red = C; Orange = RC. Putative structures were based on LS-MS and MS/MS spectra, literature and databases matches.
Cluster C1 contained 76 ions detected between 9.47 and 12.69 min, all matching the triterpenoid superclass. Seventeen compounds with a C30 skeleton were annotated (Table 1), based on fragmentation patterns adapted from Gradillas et al. [25] (S2-S3 Figs.). Three main ursane-type classes were identified:
- Tetrahydroxyurs-12-en-28-oic acid isomers (C30H48O6, mw: 504.32 Da) and derivatives: hydroxytormentic acid or isomers derivatives (1); hydroxytormentic acid or isomers (7, 8, 9); dotorioside II or isomers (15, 16, 17). Hexoside moieties produced a parent ion at m/z 684 [M + NH4]+ (15, 16, 17). Compound (17) showed additional MS/MS fragments compared to (15) and (16), suggesting a slightly different structure (S3 Fig.). Putative names were assigned based on reported occurrence in Rubus [25,27,42].
- Trihydroxyurs-12-en-28-oic acid isomers (C30H48O5, mw: 488.35 Da): tormentic acid or isomers (5, 6) [25,40]; rosamultin or isomers (13) [41,42].
- Dihydroxyurs-12-en-28-oic acid isomers (C30H48O4, mw: 472.36 Da): pomolic acid or isomers (2, 3, 4) [25,40].
Additional triterpenoids were annotated outside these classes. Compounds (10, 11, 12) were annotated as ilexgenin A1 (C30H46O6, mw: 502.33 Da) based on database matching and shared m/z 485 and 467 fragments [25], confirmed by negative-mode MS/MS (Table 1). Two other ions with the same database match were not detected in negative mode and were considered 19α-hydroxyursolic acid-type aglycones [25,41]. Compound (14) matched ilexsaponin A1 (C36H56O11, mw: 664.82 Da [M + NH4]+), reported in Ilex pubescens roots [43,44]. The diversity of ursane-type triterpenoids and glycoside isomers detected here is consistent with previous reports in blackberry leaves [25].
The upper part of cluster C2 (C2a, Fig 6) contained glycosylated compounds from multiple superclasses —flavonoids, phenylpropanoids (including coumarins) and triterpenoids — linked by sugar moiety losses (Table 1). Among them, 18 (skimmin or isomer) and 22 (cinnamic acid derivative) are reported here for the first time in Rubus leaves.
The lower part of cluster C2 (C2b, Fig 6) contained flavonoid glycosides. Quercetin and kaempferol aglycones were identified via m/z 303.05 and 287.05 peaks, consistent with Rubus literature [60,69,81]. Seven compounds were annotated: quercetin 3-O-glucuronide (25, C21H18O13, m/z 478.0804 [M + H]+); quercetin 3-O-β-D-(6’‘-O-malonyl)-glucoside or isomer (26, C24H22O15, m/z 551.1015 [M + H]+); quercetin 3-O-[6“-O-(3-hydroxy-3-methylglutaryl)]hexoside (29, C27H28O16, m/z 609.1440 [M+H]+); kaempferol 3-O-glucuronide (24, C21H18O12, m/z 463.0858 [M+H]+); kaempferol 3-O-[6”-O-(3-hydroxy-3-methylglutaryl)]hexoside (27, C27H28O15, m/z 593.1490 [M + H]+), reported here for the first time in Rubus; tiliroside (28, C30H26O13, m/z 595.1437 [M + H]+). Remaining cluster C2 ions showed m/z 303.05 or 287.05 fragments and sugar losses of 146 and 162 Da, suggesting additional kaempferol or quercetin hexoside derivatives.
Cluster C6 contained kaempferol and quercetin aglycones arising from in-source fragmentation of C2b glycosides (S3 Fig.), as well as ellagic acid (32, m/z 303.0117 [M + H]+, C14H6O8) [65,68].
Lignans were annotated in clusters C8 and C17. Compound 34 (cluster C8) was annotated as a furfuranoid lignan (C33H38O13, m/z 643.2381 [M + H]+), matching HHDP-syringaresinol, first reported as a natural product in Ficus hirta [82], and not previously described in Rubus leaves. Remaining C8 nodes are derivatives or in-source fragments of 34, showing characteristic neutral losses of pentose (132 Da), methyl (30 Da) and hydroxyl (34 Da) groups. Compounds 37 and 38 (cluster C17, m/z 547.2136 and 547.2137 [M + Na]+) were annotated as neolignan glycoside isomers (C26H36O11), reported here for the first time in Rubus. Ion at m/z 577.2229 from cluster C17 corresponded to the same structure with an additional methoxy group (C27H38O12). Related neolignan glycosides have been reported in Epimedium sagittatum and Geum japonicum [83,84].
Compound 35 (cluster C13, m/z 802.1078 [M + Na]+) was annotated as a gallotannin (C34H24O22), confirmed by [M − H]− at m/z 783.0655. MS/MS fragments at m/z 303 and 277 (ESI+) and m/z 481, 301 and 275 (ESI−) are characteristic of this compound family [72–75]. A fragmentation pathway is proposed in supplementary materials (S3 Fig.). The seven other cluster C13 nodes were annotated as hydrolysable tannin derivatives: castalagin or vescalagin (C41H26O26) derivatives [74,75,85]; casuarictin (C41H28O26) isomers [73,86–89]; geraniin (C41H28O27) isomers [86,90]; sanguiin H1-10 derivatives or hippophaenin B (C48H32O31) isomer [91,92] (S3 Fig.). Such gallotannins have been reported in Rubus leaves, fruits and seeds [93–96].
Compound 36 (cluster C16, m/z 331.15) was annotated as an eremophilane sesquiterpenoid (C19H24O6, exact mass: 348.15 Da), detected as [M + H − H2O]+ due to in-source water loss [97]. Two isomeric compounds with identical MS/MS peaks (m/z 331, 313, 287, 255, 189, 151) have been reported in Camellia japonica and Casearia arborea [76,77], but to our knowledge not previously in Rubus.
Cluster C5 ions could not be properly annotated. They shared an MS/MS fragment at m/z 274 and displayed even measured masses, suggesting the presence of an even number of nitrogen atoms or NH4+ adducts.
Biomarkers
Very Important for Projection (VIP) ions scores explaining the differences between groups have been extracted and listed for the first two components considering a significancy threshold of 1 according to the literature [98].
VIP scores and relative repartition of the three groups (ID, C, RC) were represented in the molecular network (Fig 7). Table 2 shows a summary of the main annotated compound’s VIP scores and statistics tests output.
Color represents VIP’s scores extracted from first component of PLS-DA. Only the 9 first single nodes out of 279 were represented.
Most of the highest-ranking VIPs were found in clusters C1, C2, C5, C6, C13, C16 and C17 (Table 2).
In cluster C1, 27 of 76 features were significant VIPs (Fig 7). Among triterpenoids, 15 (dotorioside II or isomer) was particularly discriminatory. This compound and related ursane-type triterpenoids have been reported in Rubus and other species for their bioactivities [99–101], and 15 appears to be a relevant biomarker of the Corylifolii and Rubus sections showing higher AIIs in the RC group (Table 2). Similarly, 14 (ilexsaponin A1 or isomers) and 1 (hydroxytormentic acid or isomers derivatives) were more represented in the RC group, whereas 11 (ilexgenin A1) was more characteristic of R. idaeus (Table 2). Compounds 13, and 16, and their in-source fragments also showed higher AIIs in the RC group (Table 2). The shared annotation of 15, 16 and 17 likely reflects closely related structural isomers that could not be distinguished here. Still, AIIs’ trends of these three compounds revealed by tests showed divergences highlighting the importance of the numerous triterpenoids ursane type isomers. Overall, ursane-type triterpenoids appear as important contributors to Rubus taxa differentiation.
In cluster C16, all five significant VIPs (ranks 11, 15, 20, 33, 65), annotated as eremophilane-type sesquiterpenoids, followed a RC > C > ID AIIs’ gradient. These compounds are known for antibacterial and anti-inflammatory activities [102–104] and are, to our knowledge, reported here for the first time in Rubus.
In cluster C2, 20 of 40 features had significant VIP scores (Fig 7). Most flavonoid glycosides — including kaempferol 3-O-[6“-O-(3-hydroxy-3-methylglutaryl)]hexoside (27), quercetin 3-O-[6”-O-(3-hydroxy-3-methylglutaryl)]hexoside (29), compound 30 and their derivatives — were more abundant in the ID group, whereas kaempferol 3-O-glucuronide was more abundant in RC (Fig 7). Cluster C6 (quercetin and kaempferol aglycones, containing 5 significant ions) exhibited the same trend, as these ions largely arise from in-source fragmentation of C2 compounds. Compound 29 was especially associated with R. idaeus, as were 27, 30, and the kaempferol and quercetin moieties contained in C6 cluster. Flavonol glycosides acylated with 3-hydroxy-3-methylglutaric acid are known in Rubus for antioxidant, anti-inflammatory and antimicrobial activities [24,105,106] and have been described in R. idaeus and R. fructicosus cultivars [24,60,66]. Porter et al. [107] did not detect related compounds in R. idaeus or 12 other wild Rubus species, which may be explained by the promoting effect of alpine environments on many specialized metabolite production [108].
In cluster C13, 7 of 8 features had significant VIP scores (Fig 7). Compound 35 and most cluster ions showed lower AIIs in the ID group, with three ions (VIP ranks 40, 57 and 93) also differing between RC and C, suggesting that gallotannins are relevant markers for taxonomic classification (Table 2).
All six neolignan glycosides from cluster C17 were biomarkers of the RC group (Fig 7, Table 2). Compounds 37 and 38 had the highest VIP scores in the network (ranks 1 and 4) and followed a RC > C > ID AIIs’ gradient, as did most cluster ions, except ion at m/z 577.2229 (rank 27), which was higher in C than in RC (Table 2). Neolignan glycosides are known for protective roles against herbivores and microorganisms [109]. Xu et al. [110] identified five lignans (rubuslins) in R. idaeus leaves and rhizomes. Here, most annotated neolignan glycosides were more abundant in Corylifolii, Rubus and Caesii sections, and further work should confirm their value as section-level taxonomic markers.
In cluster C5, 10 of 19 features were significant VIPs, all more abundant in the ID group. These compounds could not be annotated and deserve further structural characterization.
Overall, results revealed a strong phytochemical structure across Rubus taxa. Three main chemical groups were identified: R. idaeus, Rubus/Corylifolii, and Caesii. R. idaeus was characterized by higher abundance of flavonol glycosides, while Rubus/Corylifolii were enriched in sesqui- and triterpenoids, neolignans and gallotannins. Caesii showed an intermediate profile, explaining the scarcity of clear biomarkers for this section. Further analyses are needed to identify markers at more specific taxonomic levels within these sections.
Using untargeted LC-QTOF/MS/MS metabolomics on Rubus leaf extracts from 9 alpine taxa, several polyphenols and terpenoids were putatively annotated as relevant taxonomic classifiers. This is one of the first attempts to validate Rubus taxonomy through phytochemical diversity. Triterpenoids, flavonoids, gallotannins and lignans have been previously reported in Rubus [24,25,60,66,93–96] but rarely compared across wild taxa for taxonomic purposes. Oszmiański et al. [66] analyzed fruit phenolics in 23 wild Rubus taxa and noted the lack of comparative chemical studies, detecting 34 phenolics without discussing their distribution. Our results support a more structured chemical differentiation. Eight compounds — coumarin glycoside (18), cinnamic acid derivatives (22), flavonoid glycosides (27, 30), lignans (34, 37, 38) and sesquiterpenoid (36) — are reported here for the first time in Rubus leaves. This study provides a scientific basis for selecting Rubus taxa for targeted extraction of leaf compounds in a context of cosmetic or pharmaceutical valorization.
Materials and methods
Part of the experiments, in particular plant sampling, were conducted in a natural zone located in Haute-Savoie (France). Field permit was delivered by Asters (Conservatoire Espaces Naturel Haute-Savoie) and an “Internationally recognized certificate of compliance constituted from information on the permit or its equivalent made available to the Access and Benefit-sharing Clearing-House” has been delivered by the United Nations Environment Program (ABSCH-IRCC-FR-259097–1).
Sampling sites
The study area (around 4000 ha) is located within the limestone Tournette range (Massif des Bornes, Northern Alps, France) characterized by mostly calcareous soils and wet mountain climate with an average annual cumulated pluviometry of 1.7 m (Weather station Thônes, Les Besseaux (74), Alt. 630m, 1991–2020 data). After a comprehensive campaign led by the “Conservatoire d’Espaces Naturels de Haute-Savoie” (Asters) [111] ten stations (diameter 19 m) were sampled in 2023 based on their accessibility and the presence of Rubus L. specimens. In 2024, a new survey added 18 additional stations, to increase both the diversity of environmental conditions and the number of potential taxa. The 28 sites spread over an altitudinal gradient of 856 m to 1913 m represent several environmental habitats. Additionally, a station located on Clarins private production site named “Domain” (1153 m a.s.l.) was sampled because of its easier access to spontaneous Rubus L. specimens. Those specimens were used as references to monitor individual metabolome variation during the harvest season. Phytosociological and pedological descriptions of the stations are compiled in the supplementary (Table S2).
Leaf sampling procedure
Two sampling campaigns were carried out in 2023 from the 5th to 12th September and in 2024 from the 20th August to 17th September. The taxa of interest were annotated as precisely as possible in the field using the criteria of different identification keys [5,12,112]. The term ‘morph’ used hereinafter refers to individuals with the same annotation result. Individuals of interest were counted and marked with a colored twine. To achieve robust sampling, 3–10 leaves depending on their size were collected from three to five individuals of the same morph at similar stages of development, i.e., after the growth peak period and before senescence. Before collecting the leaves, photographs of each individual were taken. Rubus have lignified stems adorned with prickles that can be divided into several ‘canes’ of different ages. The leaves were collected from the floricans (stems bearing inflorescences) and primocanes (leafy stems from the current year, also known as turions) and stored in chemically neutral paper bags. The samples were then dried in open air on racks for an average of ten days according to the protocol used at Clarins Laboratories. The 2023 sampling of 54 samples was carried out at 10 stations (10 stations x [3–8 replicates]). In 2024, the sampling was carried out at 28 stations, yielding to 192 samples (28 stations x [3–8 replicates]).
To provide a morphological reference and to confirm our identifications, one or more individual’s representative from the sample’s replicates intended for metabolomic analysis were collected to create an herbarium according to the protocol described in the literature [113]. These individuals were pressed and dried using newspaper and homemade press. The 2023 and 2024 herbarium contain respectively 21 and 53 specimens. Voucher specimens are available for consultation at the Grenoble Alpes University (Grenoble 38, France).
The presence of spontaneous R. idaeus at the ‘Domain’ site enabled weekly sampling in 2024 to monitor individual metabolome variation during the harvest season.
Morphological annotations
Morphological annotations of the herbarium specimens were conducted using a botanical magnifying glass, following the criteria of various identification keys [4,5,12,112] and finally reviewed by David Mercier, Pascal Duboc and Jean-Marie Royer using both herbarium and field photographs. Some of the identifications must still be considered carefully due to the complexity of the genus and the poor documentation relative to Rubus of alpine mountains. Taxonomic ranks larger than species: subgenus, sections, subsection, series have then been used for some specimens. Morphological characters, including cane shape and pruinosity, armature type, leaf architecture, lower leaflet indumentum, presence or absence of stipitate glands, fruit colour and pruinosity were analyzed to annotate specimens. Rubus idaeus was readily identifiable by its pinnate leaves with a silvery-tomentose lower surface and its red fruits detaching freely from the receptacle. Distinction among the sections Rubus, Hiemales, Corylifolii, and Caesii was more challenging, because of the untypical aspects of alpine Rubus specimens collected for the present study. Section Rubus was identified according to the following main characters: digitately compound leaves, absence of stipitate glands, robust prickles, and black fruits remaining firmly attached to the receptacle. Presence of stipitate glands on stems and/or peduncles allowed to precise the identification to the Hiemales subsection. Section Corylifolii was identified according to the following main characters: often arching to sprawling habit, and ternate to digitately compound leaves. Specimens from Caesii section were identified according to the following main characters: stems and fruits bearing a conspicuous glaucous (pruinose) bloom; leaflets often ternate and lobed; fruits black and commonly pruinose. Detailed results of morphological identification are available in the supplementary data (Table S1). 2023 et 2024 samples were grouped in 5 and 9 taxonomic groups respectively (Table 3) based on the morphological analysis of each specimen and the taxonomic classification of Rubus L. to attribute homogenous taxonomic rank to the studied samples. Five groups based on morphological identification were considered in 2023: ca (R. caesius species); Co (Coryfolii section); Co_Ru (specimens that could not be distinguished from the Rubus and Coryfolii sections); Hi (Hiemales Subsection); id (R. idaeus species). The same five groups were found in 2024 sampling, along with four new groups: Ca_Ru (specimens that could not be distinguished from the Caesi and Rubus sections); Co_Ru_ca (specimens that could not be distinguished from the Coryfolii section, the Pallidi serie and the caesius species); id Domain (R. idaeus species sampled at the Domain station); x (specimens that couldn’t be identified) (Table 3, Fig 1). Fig 1B presents a simplified taxonomical Rubus scheme and the nine groups distribution.
Details about morphological identifications are available in supplementary data and S1 Table.
LC-MS/MS analysis
Rubus leaves were crushed to obtain pieces inferior to 0.5 mm. Extraction was carried out with a solvent (EtOH 80%) ratio to dry mass of 10 and an agitation-precession (Precellys ©) program with a rotation speed of 8000 rpm in 3 cycles of 15 seconds separated by 5-second breaks. The supernatant was filtered at 0.2 μm and then stored in vials in the dark at 8°C before being analyzed by UPLC-UV-ESI-QTOF. A biological pooled QC sample consisting of 10 μL of each sample extract was used to validate the repeatability of the metabolomics dataset and analyzed every ten runs during the whole sequence to constitute a technical replicate. Samples run order was randomly set using the function “RAND” from Microsoft Excel, except for the first four runs which were constituted by three blanks and then one QC to start the analysis. Liquid Chromatography was performed on Waters (Waters, Milford, Massachusetts, USA) Acquity I-Class UPLC with PDA coupled to a Waters Xevo G2-XS QTOF on a Thermoscientific C18 Hypersil Gold reverse phase column (2.1 x100 mm, particle size 1.9 µm). The chromatographic method used for the 2023 samples is as follows: Solvent A: H2O implemented with formic acid (0.1%) and ammonium formate (5 mM). Solvent B: methanol (100%). 0–0.5 min 5% B; 0.5–17 min 5- > 55% B; 17–20 min 55 - > 98% B; 20–28 min 98% B; 28–29 min 98%- > 5% B; 29–35 min 5% B. Flow rate: 0.4 μL/min. Mass spectrometer was set to perform Data Dependent Acquisition (DDA), with a detection window of 50–2000 Da, a MS detection threshold of 1000, and a scan time of 0.2 s. The MSMS parameters were programmed to fragment 3 ions, with a scan time of 0.2 s and a cumulative intensity threshold of 100,000. All spectra were acquired in continuum mode. A ramp was used to set the collision energy: 15–35 eV for molecules with a low molecular weight (50 Da) and 25–55 eV for molecules with a high molecular weight (1200 Da). All samples were analyzed in positive mode. Run time for each sample was 35 minutes. 2024 samples were analyzed with an optimized method based on the 2023 results. Time analysis was reduced to 22 minutes and a two steps gradient using acetonitrile instead of methanol appeared to be more suitable for the separation of compounds: 0–0.5 min 5% B; 0.5–1.5 min 5–10% B; 1.5–6 min - > 10% B; 6–7 min 10- > 20% B; 7–9 min 20% B; 9–13.5 min 20- > 98% B; 13.5–17.5 min 98% B; 17.5–20 min 98- > 5% B; 20–22 min 5% B. Flow rate: 0.4 μL/min. The DDA acquisition method was also optimized: detection window: 50–1200 Da, a MS detection threshold of 5 000, and a scan time of 0.2 s. The MSMS parameters were programmed to fragment 3 ions, with a scan time of 0.05 s and a cumulative intensity threshold of 10 000. All spectra were acquired in continuum mode. A ramp was used to set the collision energy: 5–40 eV for molecules with low molecular weight (50 Da) and 20–70 eV for molecules with high molecular weight (1200 Da). All samples were analyzed in positive mode. To verify the precision of our analysis during time, and the MS/MS detection, a quercetin standard provided by Extrasynthese (Genay, France) was analyzed every 10 runs. In addition, to assist with annotation, QC sample plus one sample of each taxonomic group were analyzed in negative mode using the same chromatographic and DDA parameters.
Metabolomics data processing
Waters’ raw data were converted to mzML files using MSconvert (Version: 3.0.24124-ba8a4fd) and corrected by Precursor Corrector script (Elnur Garayev’s script, Windows_x64_v0.5) with the exact Lockspray’s Leucine m/z and the maximal error between precursors set by default. Data were processed on MZmine v.2.53 [114,115]. We used the MZmine 2 data-preprocessing workflow proposed by Olivon et. al [116] with a few adaptations to our data concerning the intensity threshold on the different steps. The 2023 et 2024 datasets were treated separately with their respective parameter adaptations. Ionic intensity (peak intensity) matrices of respectively 1322 and 1730 ions were generated for 2023 and 2024.
Data analysis
MetaboAnalyst v.6.0 was used for statistical analysis preprocessing. Filters were used to remove features with low-quality, low repeatability based on QC samples (RSDs), and low variance based on interquantile range (IQR). Those filters were respectively set at thresholds of 80%, 55% and 40%. After this filtering step the 2023 and 2024 matrices contained respectively 566 and 668 features. Missing values have been replaced by 1/5 of the minimum positive value of the feature. The matrices were then centered and reduced using median normalization, cube root transformation and Pareto scaling algorithms to perform multivariate statistical analysis and analysis of variance (ANOVA). Filtered and normalized matrices were exported from MetaboAnalyst and imported on Rstudio (2025.9.1) running with R v.4.5.1. to conduct statistical analysis. The package Mixomics was used for multivariate analysis. First, Principal Component Analyses (PCA) was performed on both 2023 and 2024 matrices, using all biological samples, blanks, and Quality Controls to confirm the repeatability of the runs during the analysis (S4 Fig.). One sample from 2023 analysis belonging to the Hi group sampled at station 3 has been removed from the dataset because its PCA projection showed a trend that differed sharply from the rest of the group. Photographs of this sample were examined and compared with those of other replicas, and visual differences were noted in the appearance of the leaves (Fig. S5). The following multivariate analyses were performed on a dataset from which the QCs and blanks were removed. The most explanatory principal components were chosen to perform the PCA and Partial Least Squares Discriminant Analysis (PLS-DA). Dendrograms were calculated with the Dendextend package, using Ward’s clustering algorithms and Euclidean distance measures. VIP extraction has been realized on MetaboAnalyst v.06 by downloading the “Imp. Features” table.
Molecular network and annotation
Molecular network representations were generated on MetGem V.1.5.2 from the MZmine ion intensity matrix containing only ions with associated MS2 fragmentation (106 samples x 668 ions). Despite duplicate and isotopes filtering during data processing, we observed the occurrence of duplicate ions and fragment nodes in the network, supposedly the result of in-source fragmentation. Choice has been made to keep those nodes in the network to conserve a maximum of spectral information to facilitate annotation. Still, potential in-source fragmentation phenomena have been carefully considered during nodes’ annotation process. Cosine score compilation parameters: m/z tolerance 0.02; minimum matched peaks: 2; Top K: 10; minimal cosine score value: 0.70; max. connected component size: 1000. An initial annotation attempt was made using spectral data libraries available on MetGem, an internal experimental library generated with standards, the Lotus library fragmented by FIORA, and the FragHub database downloaded in September 2025. Metgem’s libraries’ matches parameters were set as follow: m/z tolerance 0.02; Minimum matched peaks: 2; Minimum Cosine Score Values: 0.2. The matching parameters were voluntarily set loose to maximize the number of annotations. All outputs were then compared between the different databases and literature occurrences. Peak informations (molecular mass, MSMS fragmentation, retention time and UV absorbance) were also compared with literature. The molecular network displays have been produced on Cytoscape version 3.10.4 by loading the same.mgf file as the one used in Metgem.
Supporting information
S1 Table. Morphological identifications of Rubus specimens collected in 2023 and 2024. Content of columns from right to left: sampling station.
Taxonomic identification and taxonomic rank; other possible identification; herbarium number.
Detailed morphological annotation below.
https://doi.org/10.1371/journal.pone.0354840.s001
(DOCX)
S2 Table. Site-specific and pedological characteristics of the stations.
Erosion, soil moisture, soil compaction, and root abundance are discrete values represented on a five degrees scale: 0–5. Soil Texture: A/a: Clay, L/l: Loamy, S/s: sandy. PH was measured in water. Latitude and longitude are expressed in WGS84 (GPS).
https://doi.org/10.1371/journal.pone.0354840.s002
(DOCX)
S1 Fig. Complete molecular network (668 nodes) based on MS/MS fragmentation by LC-MS ESI Q-TOF performed on Rubus leaf extracts.
Color tag is based on taxa groupment shown by multivariate analysis: purple = ID group; orange = RC group; red = C group, chart’s proportions were calculated by ionic intensity average of each feature in the three groups.
https://doi.org/10.1371/journal.pone.0354840.s003
(DOCX)
S2 Fig. Main triterpenoids complementary information.
A: Triterpenoids focus on TIC chromatograms (RT: 9.20–12.65 min). Peaks annotated as triterpenoids are colored, and corresponding ions detected in MS1 are represented in the same color rectangle. Parent ions considered for annotation are bold later; in-source fragments are identified by *. Compounds number appear in parentheses. B: Annotated compounds’ putative structures. C: Fragmentation pathway proposed for each represented structure adapted from Gradillas et al., 2019 [27]. Dashed arrows indicate that proposed molecular formula and adduct weren’t in the original publication.
https://doi.org/10.1371/journal.pone.0354840.s004
(DOCX)
S3 Fig. Complementary spectral information of annotated compounds and molecular network propagation.
3.1: Cluster C1. 3.2: Triterpenes from the tetrahydroxyurs-12-en-28-oic acid isomer class; compound 7, compound 8, compound 9, compound 15, compound 16, compound 17. 3.3: Triterpenes from the trihydroxyurs-12-en-28-oic acid class; compound 5, compound 13. 3.4: Triterpenes from the trihydroxyurs-12-ene-23(or24),28-dioic acid isomer class: compound 2, compound 3, compound 4. 3.5: Spectral data of triterpenes from ilexgenin A1 and ilexsaponin A1 isomers and derivatives; compound 11, compound 10, compound 12, compound 14. 3.6: Compound 18. S3.7: Clusters C6, C2b. S3.8: Compound 31, compound 32, compound 33. S3.9: Compound 24, compound 25, compound 26, compound 27. S3.10: Cluster C17. S3.11: Compound 38. S3.12: Cluster C13. S3.13: Compound 35. S3.14: Cluster C16. S3.15: Compound 36.
https://doi.org/10.1371/journal.pone.0354840.s005
(DOCX)
S4 Fig. Principal Component Analysis of the ionic intensity matrices figuring QC samples from the LC-MS ESI Q-TOF analysis of Rubus leaves: sampled in A: 2023 (53 samples; 21 QC x 566 features) and B: 2024 (192 samples; 26 QC x 668 features).
https://doi.org/10.1371/journal.pone.0354840.s006
(DOCX)
S5 Fig. Supporting information concerning the removing of the outlier sample from 2023’s dataset.
A: Principal Component Analysis of the ionic intensity matrices from the LC-MS ESI Q-TOF analysis of Rubus leaves sampled in 2023 (54 samples; 21 QC x 566 features), the red arrow identifies the outlier sample that has been removed. B: Photographs of the outlier sample. C: Photographs of a replica from the same taxonomic group sampled at the same station as the outlier sample. Comparison of photographs B and C reveals differences in the color of the upper and lower surfaces, the overall shape, and the dissection of the leaf blades.
https://doi.org/10.1371/journal.pone.0354840.s007
(DOCX)
Acknowledgments
The authors thank David Mercier, Pascal Duboc, and Jean-Marie Royer for their valuable contribution to specimen’s morphological identification. We would also like to acknowledge Olivier Billant and Christophe Manzon from Asters for their help in field work and in the characterization of sampling zones. Work team from Domaine Clarins is also acknowledged for their assistance in field work. We would also like to thank the Herbarium of Grenoble Alpes University and Christophe Perrier for taking care of the voucher specimens.
References
- 1. Haveman R, Ronde I de. The role of the Weberian Reform in European Rubus research and the taxonomy of locally distributed species – which species should we describe?. Nordic Journal of Botany. 2012;31(2):145–50.
- 2.
Royal Botanic Gardens K. Plants of the World Online. Rubus L. 2026 https://powo.science.kew.org/taxon/urn:lsid:ipni.org:names:30000199-2
- 3. Sochor M, Vašut RJ, Sharbel TF, Trávníček B. How just a few makes a lot: Speciation via reticulation and apomixis on example of European brambles (Rubus subgen. Rubus, Rosaceae). Mol Phylogenet Evol. 2015;89:13–27. pmid:25882835
- 4.
Ferrez Y, Royer J-M. Le genre Rubus dans le Nord-Est de la France. Description, détermination, écologie et phytosociologie. SBFC, CBNFC-ORI &GREFFE; 2021.
- 5. Mercier D. Le genre Rubus L. (Rosaceae) dans le massif armoricain et ses abords: une nouvelle approche, et une première espèce à réviser, R. caesius L. E.R.I.C.A. 2012; 25:97–116.
- 6. Stace CA. Species recognition in agamosperms — The need for a pragmatic approach. Folia Geobot. 1998;33(3):319–26.
- 7. Trávníček B, Oklejewicz K, Zieliński J. Rubus ambrosius (Rubus subsect. Rubus, Rosaceae), a new bramble species from the eastern part of Central Europe. Folia Geobotanica. 2005; 40:421–34.
- 8. Bicknell RA, Koltunow AM. Understanding apomixis: recent advances and remaining conundrums. Plant Cell. 2004;16 Suppl(Suppl):S228-45. pmid:15131250
- 9. Whitton J, Sears CJ, Baack EJ, Otto SP. The Dynamic Nature of Apomixis in the Angiosperms. International Journal of Plant Sciences. 2008;169(1):169–82.
- 10. Šarhanová P, Vašut RJ, Dančák M, Bureš P, Trávníček B. New insights into the variability of reproduction modes in European populations of Rubus subgen. Rubus: how sexual are polyploid brambles?. Sex Plant Reprod. 2012;25(4):319–35. pmid:23114637
- 11. Ferrez Y, Tison J-M. Contribution à la connaissance des Alchemilla du massif jurassien. Première partie Alchemilla section Alpinae Buser. Les Nouvelles Archives de la Flore jurassienne. 2009; 7:43–56. https://www.researchgate.net/publication/323839867
- 12. Duboc P. genre Rubus: un défi pour les botanistes contemporains. BIOM. 2022; 3:1–9.
- 13.
Ferrez Y, Royer J-M. Le genre Rubus en Franche-Comté, résultats des premières investigations. Les Nouvelles Archives de la Flore jurassienne et du nord-est de la France. 2010; 8:57–66.
- 14. Royer J-M. Observations nouvelles sur les Rubus du Nord-Est de la France. Bulletin de la Société Botanique du Centre-Ouest - Nouvelle Série -. 2009; 40:29–48.
- 15. Weber HE. Former and modern taxonomic treatment of the apomicticRubus complex. Folia Geobot Phytotax. 1996;31(3):373–80.
- 16. Ferrez Y, Bornand C. Nouvelles observations de taxons de Rubus (sous-genre Rubus) dans le canton de Vaud. Bulletin du Cercle vaudois de botanique. 2019; 48:125–40. https://www.researchgate.net/publication/338052395
- 17. Lazare J-J. RUBI de France: appel à contribution batologique dans le cadre de la préparation de l’Atlas Florae Europaeae 15. Le Journal de botanique. 2006; 34:17–32. Available from:
- 18. Meng Q, Manghwar H, Hu W. Study on Supergenus Rubus L.: Edible, Medicinal, and Phylogenetic Characterization. Plants (Basel). 2022;11. pmid:35567211
- 19. Sochor M. Diversity, phylogenesis and evolutionary mechanisms in the genus Rubus. PhD. Thesis, Palacký University Olomouc. 2016.
- 20. Tomaszewski D, Zieliński J, Gawlak M. Foliar indumentum in central‐European Rubus species (Rosaceae) and its contribution to the systematics of the group. Nordic Journal of Botany. 2013;32(1):1–10.
- 21. Trávníček B, Sochor M, Kosiński P, Király G. Taxonomy of the Rubus gothicus group in south-eastern central Europe. Preslia. 2021;93(4):321–40.
- 22. Huang T-R, Chen J-H, Hummer KE, Alice LA, Wang W-H, He Y, et al. Phylogeny of Rubus (Rosaceae): Integrating molecular and morphological evidence into an infrageneric revision. Taxon. 2023; 72:278–306.
- 23. Michael K. Clarification of basal relationships in Rubus (Rosaceae) and the origin of Rubus chamaemorus. Masters Theses & Specialist Projects. 2006; 250. http://digitalcommons.wku.edu/theses/250
- 24. Patel AV, Rojas-Vera J, Dacke CG. Therapeutic Constituents and Actions of Rubus Species. Current Medicinal Chemistry. 2004; 11:1501–12.
- 25. Gradillas A, Martínez-Alcázar MP, Gutiérrez E, Ramos-Solano B, García A. A novel strategy for rapid screening of the complex triterpene saponin mixture present in the methanolic extract of blackberry leaves (Rubus cv. Loch Ness) by UHPLC/QTOF-MS. J Pharm Biomed Anal. 2019;164:47–56. pmid:30343243
- 26. Li B-Z, Wang B-G, Jia Z-J. Pentacyclic triterpenoids from Rubus xanthocarpus. Phytochemistry. 1998;49(8):2477–81.
- 27. Xiao-Hong Z, Kasai R, Ohtani K, Tanaka O, Rui-Lin N, Chong-Ren Y, et al. Oleanane and ursane glucosides from Rubus species. Phytochemistry. 1992;31(10):3642–4.
- 28. Hummer KE. Rubus Pharmacology: Antiquity to the Present. Hortscience. 2010; 45:1587–91.
- 29. Vukelic-Nikolic M, Dragićević A, Stojanović N, Vasiljević P, Pavlović D. Beyond the Traditional Applications of Raspberry (Rubus idaeus) Leaf: An in vitro, in vivo and in silico Study. Rec Nat Prod. 2025:518–32.
- 30. Antonio AS, Veiga-Junior VF, Wiedemann LSM. Ocotea complex: A metabolomic analysis of a Lauraceae genus. Phytochemistry. 2020;173:112314. pmid:32120118
- 31. Bentley J, Moore JP, Farrant JM. Metabolomics as a complement to phylogenetics for assessing intraspecific boundaries in the desiccation-tolerant medicinal shrub Myrothamnus flabellifolia (Myrothamnaceae). Phytochemistry. 2019;159:127–36. pmid:30611872
- 32. Chervin J, Talou T, Audonnet M, Dumas B, Camborde L, Esquerré-Tugayé M-T, et al. Deciphering the phylogeny of violets based on multiplexed genetic and metabolomic approaches. Phytochemistry. 2019;163:99–110. pmid:31035059
- 33. Zidorn C. Plant chemophenetics - A new term for plant chemosystematics/plant chemotaxonomy in the macro-molecular era. Phytochemistry. 2019;163:147–8. pmid:30846237
- 34. Defossez E, Pitteloud C, Descombes P, Glauser G, Allard P-M, Walker TWN, et al. Spatial and evolutionary predictability of phytochemical diversity. Proc Natl Acad Sci U S A. 2021;118(3):e2013344118. pmid:33431671
- 35. Ganzera M, Guggenberger M, Stuppner H, Zidorn C. Altitudinal variation of secondary metabolite profiles in flowering heads of Matricaria chamomilla cv. BONA. Planta Med. 2008;74(4):453–7. pmid:18484542
- 36. McDougal KM, Parks CR. Elevational Variation In Foliar Flavonoids Of Quercus Rubra L. (fagaceae). American J of Botany. 1984;71(3):301–8.
- 37. Spitaler R, Schlorhaufer PD, Ellmerer EP, Merfort I, Bortenschlager S, Stuppner H, et al. Altitudinal variation of secondary metabolite profiles in flowering heads of Arnica montana cv. ARBO. Phytochemistry. 2006;67(4):409–17. pmid:16405933
- 38. Wang Y, Wang X, Chen Q, Zhang L, Tang H, Luo Y, et al. Phylogenetic insight into subgenera Idaeobatus and Malachobatus (Rubus, Rosaceae) inferring from ISH analysis. Mol Cytogenet. 2015;8:11. pmid:25674160
- 39. Šarhanová P, Sharbel TF, Sochor M, Vašut RJ, Dancák M, Trávnícek B. Hybridization drives evolution of apomicts in Rubus subgenus Rubus: evidence from microsatellite markers. Ann Bot. 2017;120(2):317–28. pmid:28402390
- 40. Falev DI, Onuchina AA, Faleva AV, Voronov IS, Kosyakov DS, Ulyanovskii NV. Analysis of pentacyclic triterpenoids and phytosterols in cloudberry (Rubus chamaemorus L.) by LC-MS/MS. Nat Prod Res. 2026;40(3):652–9. pmid:39428696
- 41. Yang B, Li H, Ruan Q-F, Xue Y-Y, Cao D, Zhou X-H, et al. A facile and selective approach to the qualitative and quantitative analysis of triterpenoids and phenylpropanoids by UPLC/Q-TOF-MS/MS for the quality control of Ilex rotunda. J Pharm Biomed Anal. 2018;157:44–58. pmid:29758469
- 42. Seto T, Tanaka T, Tanaka O, Naruhashi N. β-glucosyl esters of 19α-hydroxyursolic acid derivatives in leaves of Rubus species. Phytochemistry. 1984;23(12):2829–34.
- 43. Lin L-P, Kong X, Chen L, Chen L. Chemical constituents from the roots of cultivated Ilex pubescens. Biochemical Systematics and Ecology. 2019;82:13–5.
- 44. Liu W-Y, Feng F, Yu C-X, Xie N. Qualitative and quantitative analysis of the main constituents of Radix Ilicis Pubescentis by LC-coupled with DAD and ESI-MS detection. Nat Prod Commun. 2010;5(1):23–6. pmid:20184013
- 45. Li W, Fu H, Bai H, Sasaki T, Kato H, Koike K. Triterpenoid saponins from Rubus ellipticus var. obcordatus. J Nat Prod. 2009;72(10):1755–60. pmid:19795885
- 46. Yi T, Zhu L, Tang Y-N, Zhang J-Y, Liang Z-T, Xu J, et al. An integrated strategy based on UPLC-DAD-QTOF-MS for metabolism and pharmacokinetic studies of herbal medicines: Tibetan “Snow Lotus” herb (Saussurea laniceps), a case study. J Ethnopharmacol. 2014;153(3):701–13. pmid:24661968
- 47. De Nisco M, Manfra M, Bolognese A, Sofo A, Scopa A, Tenore GC, et al. Nutraceutical properties and polyphenolic profile of berry skin and wine of Vitis vinifera L. (cv. Aglianico). Food Chem. 2013;140(4):623–9. pmid:23692745
- 48. Ruiz A, Hermosín-Gutiérrez I, Vergara C, von Baer D, Zapata M, Hitschfeld A, et al. Anthocyanin profiles in south Patagonian wild berries by HPLC-DAD-ESI-MS/MS. Food Research International. 2013;51(2):706–13.
- 49. Qi L-W, Chen C-Y, Li P. Structural characterization and identification of iridoid glycosides, saponins, phenolic acids and flavonoids in Flos Lonicerae Japonicae by a fast liquid chromatography method with diode-array detection and time-of-flight mass spectrometry. Rapid Commun Mass Spectrom. 2009;23(19):3227–42. pmid:19725056
- 50. Abu-Reidah IM, Ali-Shtayeh MS, Jamous RM, Arráez-Román D, Segura-Carretero A. HPLC-DAD-ESI-MS/MS screening of bioactive components from Rhus coriaria L. (Sumac) fruits. Food Chem. 2015;166:179–91. pmid:25053044
- 51. Vrhovsek U, Masuero D, Gasperotti M, Franceschi P, Caputi L, Viola R, et al. A versatile targeted metabolomics method for the rapid quantification of multiple classes of phenolics in fruits and beverages. J Agric Food Chem. 2012;60(36):8831–40. pmid:22468648
- 52. Rebello LPG, Lago-Vanzela ES, Barcia MT, Ramos AM, Stringheta PC, Da-Silva R, et al. Phenolic composition of the berry parts of hybrid grape cultivar BRS Violeta (BRS Rubea × IAC 1398-21) using HPLC–DAD–ESI-MS/MS. Food Research International. 2013;54(1):354–66.
- 53. Koponen JM, Happonen AM, Auriola S, Kontkanen H, Buchert J, Poutanen KS, et al. Characterization and fate of black currant and bilberry flavonols in enzyme-aided processing. J Agric Food Chem. 2008;56(9):3136–44. pmid:18426212
- 54. Mertz C, Cheynier V, Günata Z, Brat P. Analysis of phenolic compounds in two blackberry species (Rubus glaucus and Rubus adenotrichus) by high-performance liquid chromatography with diode array detection and electrospray ion trap mass spectrometry. J Agric Food Chem. 2007;55(21):8616–24. pmid:17896814
- 55. Oszmiański J, Wojdyło A, Gorzelany J, Kapusta I. Identification and characterization of low molecular weight polyphenols in berry leaf extracts by HPLC-DAD and LC-ESI/MS. J Agric Food Chem. 2011;59(24):12830–5. pmid:22098480
- 56. Lorenz P, Bunse M, Klaiber I, Conrad J, Laumann-Lipp T, Stintzing FC, et al. Comprehensive Phytochemical Characterization of Herbal Parts from Kidney Vetch (Anthyllis vulneraria L.) by LC/MSn and GC/MS. Chem Biodivers. 2020;17(10):e2000485. pmid:32860459
- 57. Samet S, Ayachi A, Fourati M, Mallouli L, Allouche N, Treilhou M, et al. Antioxidant and Antimicrobial Activities of Erodium arborescens Aerial Part Extracts and Characterization by LC-HESI-MS2 of Its Acetone Extract. Molecules. 2022;27. pmid:35889269
- 58. Faleva AV, Ul’yanovskii NV, Onuchina AA, Falev DI, Kosyakov DS. Comprehensive Characterization of Secondary Metabolites in Fruits and Leaves of Cloudberry (Rubus chamaemorus L.). Metabolites. 2023;13(5):598. pmid:37233639
- 59. Zan T, Piao L, Wei Y, Gu Y, Liu B, Jiang D. Simultaneous determination and pharmacokinetic study of three flavonoid glycosides in rat plasma by LC-MS/MS after oral administration of Rubus chingii Hu extract. Biomed Chromatogr. 2018;32(3):10.1002/bmc.4106. pmid:28976589
- 60. Pavlović AV, Papetti A, Zagorac DČD, Gašić UM, Mišić DM, Tešić ŽLj, et al. Phenolics composition of leaf extracts of raspberry and blackberry cultivars grown in Serbia. Industrial Crops and Products. 2016;87:304–14.
- 61. Berhow MA, Bennett RD, Poling SM, Vannier S, Hidaka T, Omura M. Acylated flavonoids in callus cultures of Citrus aurantifolia. Phytochemistry. 1994;36(5):1225–7. pmid:7765362
- 62. Hu L, Zhou C, Huang Y-C, Wang Y, Wei G, Liang Z, et al. HPLC coupled with electrospray ionization multistage MS/MS and TLC analysis of flavones-C-glycosides and bibenzyl of Dendrobium hercoglossum. J Sep Sci. 2020;43(20):3885–901. pmid:32803831
- 63. Alkhudaydi HMS, Muriuki EN, Spencer JPE. Determination of the Polyphenol Composition of Raspberry Leaf Using LC-MS/MS. Molecules. 2025;30. pmid:40005280
- 64. March RE, Miao X-S. A fragmentation study of kaempferol using electrospray quadrupole time-of-flight mass spectrometry at high mass resolution. International Journal of Mass Spectrometry. 2004;231(2–3):157–67.
- 65. Navarro M, Moreira I, Arnaez E, Quesada S, Azofeifa G, Vargas F, et al. Flavonoids and Ellagitannins Characterization, Antioxidant and Cytotoxic Activities of Phyllanthus acuminatus Vahl. Plants (Basel). 2017;6(4):62. pmid:29244711
- 66. Oszmiański J, Nowicka P, Teleszko M, Wojdyło A, Cebulak T, Oklejewicz K. Analysis of Phenolic Compounds and Antioxidant Activity in Wild Blackberry Fruits. Int J Mol Sci. 2015;16(7):14540–53. pmid:26132562
- 67. Zhang X, Sandhu A, Edirisinghe I, Burton-Freeman B. An exploratory study of red raspberry (Rubus idaeus L.) (poly)phenols/metabolites in human biological samples. Food Funct. 2018;9(2):806–18. pmid:29344587
- 68. Zhang H, Cha S, Yeung ES. Colloidal graphite-assisted laser desorption/ionization MS and MS(n) of small molecules. 2. Direct profiling and MS imaging of small metabolites from fruits. Anal Chem. 2007;79(17):6575–84. pmid:17665874
- 69. Grochowski DM, Strawa JW, Granica S, Tomczyk M. Secondary metabolites of Rubus caesius (Rosaceae). Biochemical Systematics and Ecology. 2020;92:104111.
- 70. Liu Q, Li J, Gu M, Kong W, Lin Z, Mao J, et al. High-Throughput Phytochemical Unscrambling of Flowers Originating from Astragalus membranaceus (Fisch.) Bge. var. mongholicus (Bge.) P. K. Hsiao and Astragalus membranaceus (Fisch.) Bug. by Applying the Intagretive Plant Metabolomics Method Using UHPLC-Q-TOF-MS/MS. Molecules. 2023;28. pmid:37630367
- 71. Scigelova M, Hornshaw M, Giannakopulos A, Makarov A. Fourier Transform Mass Spectrometry. Mol. Cell. Proteomics. 2011;10.
- 72. Boulekbache-Makhlouf L, Meudec E, Mazauric J-P, Madani K, Cheynier V. Qualitative and semi-quantitative analysis of phenolics in Eucalyptus globulus leaves by high-performance liquid chromatography coupled with diode array detection and electrospray ionisation mass spectrometry. Phytochem Anal. 2013;24(2):162–70. pmid:22930658
- 73. D’Urso G, Sarais G, Lai C, Pizza C, Montoro P. LC-MS based metabolomics study of different parts of myrtle berry from Sardinia (Italy). JBR. 2017; 7:217–29.
- 74. Fernandes TA, Antunes AMM, Caldeira I, Anjos O, de Freitas V, Fargeton L, et al. Identification of gallotannins and ellagitannins in aged wine spirits: A new perspective using alternative ageing technology and high-resolution mass spectrometry. Food Chem. 2022;382:132322. pmid:35158268
- 75. Singh A, Bajpai V, Kumar S, Sharma KR, Kumara B. Profiling of Gallic and Ellagic Acid Derivatives in Different Plant Parts of Terminalia arjuna by HPLC-ESI-QTOF-MS/MS. Nat Prod Commun. 2016;11(2):239–44. pmid:27032211
- 76. Bi L, Liu H, Liu R, Chen Q, Yan H, Ni W, et al. Rapid identification of radical scavenging compounds from Camellia japonica leaves through the integration of feature-based molecular networking and statistical approach. LWT. 2025;223:117730.
- 77. Santos AL, Rodrigues MT, Michelli AP, Tamura RE, de Rubió IGS, Soares MG, et al. Exploring the Antiproliferative Activity of Flavolignans From the Leaves of Casearia arborea (Salicaceae). Chem Biodivers. 2025;22(12):e01489. pmid:40911853
- 78. Chernonosov AA, Karpova EA, Karakulov AV. Metabolomic profiling of three Rhododendron species from Eastern Siberia by liquid chromatography with high-resolution mass spectrometry. South African Journal of Botany. 2023;157:622–34.
- 79. Su XD, Li W, Ma JY, Kim YH. Chemical constituents from Epimedium koreanum Nakai and their chemotaxonomic significance. Nat Prod Res. 2018;32(19):2347–51. pmid:29157003
- 80. Schymanski EL, Jeon J, Gulde R, Fenner K, Ruff M, Singer HP, et al. Identifying small molecules via high resolution mass spectrometry: communicating confidence. Environ Sci Technol. 2014;48(4):2097–8. pmid:24476540
- 81. Gudej J. Kaempferol and quercetin glycosides from Rubus idaeus L. leaves. Acta Pol Pharm. 2003;60(4):313–5. pmid:14714861
- 82. Ye X-S, Tian W-J, Liu X-Z, Zhou M, Zeng D-Q, Lin T, et al. Lignans and phenylpropanoids from the roots of Ficus hirta and their cytotoxic activities. Nat Prod Res. 2022;36(15):3840–9. pmid:33648391
- 83. Cheng X, Qin J, Zeng Q, Zhang S, Zhang F, Yan S, et al. Taraxasterane-type triterpene and neolignans from Geum japonicum Thunb. var. chinense F. Bolle. Planta Med. 2011;77(18):2061–5. pmid:21800281
- 84. Matsushita H, Miyase T, Ueno A. Lignan and terpene glycosides from Epimedium sagittatum. Phytochemistry. 1991;30(6):2025–7.
- 85. Chang Z, Zhang Q, Liang W, Zhou K, Jian P, She G, et al. A Comprehensive Review of the Structure Elucidation of Tannins from Terminalia Linn. Evid Based Complement Alternat Med. 2019;2019:8623909. pmid:31885669
- 86. Díaz-de-Cerio E, Gómez-Caravaca AM, Verardo V, Fernández-Gutiérrez A, Segura-Carretero A. Determination of guava (Psidium guajava L.) leaf phenolic compounds using HPLC-DAD-QTOF-MS. Journal of Functional Foods. 2016;22:376–88.
- 87. García-Villalba R, Espín JC, Tomás-Barberán FA, Rocha-Guzmán NE. Comprehensive characterization by LC-DAD-MS/MS of the phenolic composition of seven Quercus leaf teas. Journal of Food Composition and Analysis. 2017;63:38–46.
- 88. Grace MH, Warlick CW, Neff SA, Lila MA. Efficient preparative isolation and identification of walnut bioactive components using high-speed counter-current chromatography and LC-ESI-IT-TOF-MS. Food Chem. 2014;158:229–38. pmid:24731336
- 89. Medic A, Jakopic J, Hudina M, Solar A, Veberic R. Identification and quantification of the major phenolic constituents in Juglans regia L. peeled kernels and pellicles, using HPLC-MS/MS. Food Chem. 2021;352:129404. pmid:33676122
- 90. Świątek Ł, Sieniawska E, Sinan KI, Maciejewska-Turska M, Boguszewska A, Polz-Dacewicz M, et al. LC-ESI-QTOF-MS/MS Analysis, Cytotoxic, Antiviral, Antioxidant, and Enzyme Inhibitory Properties of Four Extracts of Geranium pyrenaicum Burm. f.: A Good Gift from the Natural Treasure. Int J Mol Sci. 2021; 22. Epub 2021/07/16. pmid:34299238
- 91. Duckstein SM, Lotter EM, Meyer U, Lindequist U, Stintzing FC. Phenolic Constituents from Alchemilla vulgaris L. and Alchemilla mollis (Buser) Rothm. at Different Dates of Harvest. Z. Naturforsch. C J Biosci. 2012; 67:529–40.
- 92. Zheng W-H, Bai H-Y, Han S, Bao F, Zhang K-X, Sun L-L, et al. Analysis on the Constituents of Branches, Berries, and Leaves of Hippophae rhamnoides L. by UHPLC-ESI-QTOF-MS and Their Anti-Inflammatory Activities. Natural Product Communications. 2019;14(8).
- 93. Hager TJ, Howard LR, Liyanage R, Lay JO, Prior RL. Ellagitannin composition of blackberry as determined by HPLC-ESI-MS and MALDI-TOF-MS. J Agric Food Chem. 2008;56(3):661–9. pmid:18211030
- 94. Hussein SAM, Ayoub NA, Nawwar MAM. Caffeoyl sugar esters and an ellagitannin from Rubus sanctus. Phytochemistry. 2003;63(8):905–11. pmid:12895538
- 95. Kähkönen M, Kylli P, Ollilainen V, Salminen J-P, Heinonen M. Antioxidant activity of isolated ellagitannins from red raspberries and cloudberries. J Agric Food Chem. 2012;60(5):1167–74. pmid:22229937
- 96. Sójka M, Janowski M, Grzelak-Błaszczyk K. Stability and transformations of raspberry (Rubus idaeus L.) ellagitannins in aqueous solutions. Eur Food Res Technol. 2018;245(5):1113–22.
- 97. Bitchagno GTM, Koffi JG, Simo IK, Kagho DUK, Ngouela AS, Lenta BN, et al. LC-ToF-ESI-MS Patterns of Hirsutinolide-like Sesquiterpenoids Present in the Elephantopus mollis Kunth Extract and Chemophenetic Significance of Its Chemical Constituents. Molecules. 2021;26(16):4810.
- 98. Chen J, Wang W, Lv S, Yin P, Zhao X, Lu X, et al. Metabonomics study of liver cancer based on ultra performance liquid chromatography coupled to mass spectrometry with HILIC and RPLC separations. Anal Chim Acta. 2009;650(1):3–9. pmid:19720165
- 99. Ardenghi JV, Kanegusuku M, Niero R, Filho VC, Monache FD, Yunes RA, et al. Analysis of the mechanism of antinociceptive action of niga-ichigoside F1 obtained from Rubus imperialis (Rosaceae). J Pharm Pharmacol. 2006;58(12):1669–75. pmid:17331332
- 100. Choi J, Lee K-T, Ha J, Yun S-Y, Ko C-D, Jung H-J, et al. Antinociceptive and antiinflammatory effects of Niga-ichigoside F1 and 23-hydroxytormentic acid obtained from Rubus coreanus. Biol Pharm Bull. 2003;26(10):1436–41. pmid:14519951
- 101. Tonin TD, Thiesen LC, Oliveira Nunes MLd, Broering MF, Donato MP, Goss MJ, et al. Rubus imperialis (Rosaceae) extract and pure compound niga-ichigoside F1: wound healing and anti-inflammatory effects. Naunyn Schmiedebergs Arch Pharmacol. 2016; 389:1235–44. pmid:27527496
- 102. Chen W-H, Yang H-Y, Wang Y-X, Wen M-L, Yang Z-D, Chen J-J. Two New Eremophilane-Type Sesquiterpenoids from Ligularia sagitta. Chem Biodivers. 2022;19(11):e202200762. pmid:36177989
- 103. Li-Bin L, Xiao J, Zhang Q, Han R, Xu B, Yang S-X, et al. Eremophilane Sesquiterpenoids with Antibacterial and Anti-inflammatory Activities from the Endophytic Fungus Septoria rudbeckiae. J Agric Food Chem. 2021;69(40):11878–89. pmid:34605647
- 104. Zhou M, Duan F, Gao Y, Peng X, Meng X, Ruan H. Eremophilane sesquiterpenoids from the whole plant of Parasenecio albus with immunosuppressive activity. Bioorg Chem. 2021;115:105247. pmid:34411979
- 105. Gong ES, Li B, Li B, Podio NS, Chen H, Li T, et al. Identification of key phenolic compounds responsible for antioxidant activities of free and bound fractions of blackberry varieties’ extracts by boosted regression trees. J Sci Food Agric. 2022;102(3):984–94. pmid:34302364
- 106. Puupponen-Pimiä R, Nohynek L, Juvonen R, Kössö T, Truchado P, Westerlund-Wikström B, et al. Fermentation and dry fractionation increase bioactivity of cloudberry (Rubus chamaemorus). Food Chem. 2016;197(Pt A):950–8. pmid:26617039
- 107. Porter EA, van den Bos AA, Kite GC, Veitch NC, Simmonds MSJ. Flavonol glycosides acylated with 3-hydroxy-3-methylglutaric acid as systematic characters in Rosa. Phytochemistry. 2012;81:90–6. pmid:22721781
- 108. Treutter D. Significance of flavonoids in plant resistance: a review. Environ Chem Lett. 2006; 4:147–57.
- 109. Teponno RB, Kusari S, Spiteller M. Recent advances in research on lignans and neolignans. Nat Prod Rep. 2016;33(9):1044–92. pmid:27157413
- 110. Xu Y, Li L-Z, Cong Q, Wang W, Qi X-L, Peng Y, et al. Bioactive lignans and flavones with in vitro antioxidant and neuroprotective properties from Rubus idaeus rhizome. Journal of Functional Foods. 2017;32:160–9.
- 111.
Manzon C. Contribution aux connaissances des genres Alchemilla et Rubus sur le massif de la Tournette.; 2022.
- 112.
Ferrez Y, Royer J-M. Identification de dix espèces communes de Rubus du nord-est de la France. Les Nouvelles Archives de la Flore jurassienne et du nord-est de la France. 2015; 13:121–89.
- 113.
Belhacène L, Gritti C, Sirvent L, Argagnon O. Compte rendu de la première sortie batologique du groupe Rubus de la SBOcc: le Haut-Languedoc. Société botanique d’Occitanie; 2021.
- 114. Katajamaa M, Miettinen J, Oresic M. MZmine: toolbox for processing and visualization of mass spectrometry based molecular profile data. Bioinformatics. 2006;22(5):634–6. pmid:16403790
- 115. Pluskal T, Castillo S, Villar-Briones A, Oresic M. MZmine 2: modular framework for processing, visualizing, and analyzing mass spectrometry-based molecular profile data. BMC Bioinformatics. 2010;11:395. pmid:20650010
- 116. Olivon F, Grelier G, Roussi F, Litaudon M, Touboul D. MZmine 2 Data-Preprocessing To Enhance Molecular Networking Reliability. Anal Chem. 2017;89(15):7836–40. pmid:28644610