Olivier Auguste Kirchhofferab,
Luis Quirós-Guerrero
ab,
Jahn Nitschke
c,
Louis-Félix Nothias
abd,
Frédéric Burdet
e,
Laurence Marcourt
ab,
Nabil Hanna
c,
Florence Mehl
e,
Bruno David
f,
Antonio Grondin
f,
Emerson Ferreira Queiroz
ab,
Marco Pagni
e,
Thierry Soldati
c and
Jean-Luc Wolfender
*ab
aInstitute of Pharmaceutical Sciences of Western Switzerland, University of Geneva, CMU, 1211 Geneva, Switzerland. E-mail: Jean-Luc.Wolfender@unige.ch
bSchool of Pharmaceutical Sciences, University of Geneva, CMU, 1211 Geneva, Switzerland
cDepartment of Biochemistry, Faculty of Sciences, University of Geneva, Quai Ernest-Ansermet 30, 1205 Geneva, Switzerland
dUniversité Côte d’Azur, CNRS, ICN, France
eVital-IT, SIB Swiss Institute of Bioinformatics, 1015 Lausanne, Switzerland
fGreen Mission Department, Herbal Products Laboratory, Pierre Fabre Research Institute, Toulouse, France
First published on 23rd April 2025
The rising threat of multidrug-resistant tuberculosis, caused by Mycobacterium tuberculosis, underscores the urgent need for new therapeutic solutions to tackle the challenge of antibiotic resistance. The current study utilized an innovative 3R infection model featuring the amoeba Dictyostelium discoideum infected with Mycobacterium marinum, serving as stand-ins for macrophages and M. tuberculosis, respectively. This high-throughput phenotypic assay allowed for the evaluation of more specific anti-infective activities that may be less prone to resistance mechanisms. To discover novel anti-infective compounds, a diverse collection of 1600 plant NEs from the Pierre Fabre Library was screened using the latter assay. Concurrently, these NEs underwent untargeted UHPLC-HRMS/MS analysis. The biological screening flagged the NE from Stauntonia brunoniana as one of the anti-infective hit NEs. High-resolution HPLC micro-fractionation coupled with bioactivity profiling was employed to highlight the natural products driving this bioactivity. Stilbenes were eventually identified as the primary active compounds in the bioactive fractions. A knowledge graph was then used to leverage the heterogeneous data integrated into it to make a rational selection of stilbene-rich NEs. Using both CANOPUS chemical classes and Jaccard similarity indices to compare features within the metabolome of the 1600 plant NEs collection, 14 NEs rich in stilbenes were retrieved. Among those, the roots of Gnetum edule were flagged as possessing broader chemo-diversity in their stilbene content, along with the corresponding NE also being a strict anti-infective. Eventually, a total of 11 stilbene oligomers were isolated from G. edule and fully characterized by NMR with their absolute stereochemistry established through electronic circular dichroism. Six of these compounds are new since they possess a stereochemistry which was never described in the literature to the best of our knowledge. All of them were assessed for their anti-infective activity and (−)-gnetuhainin M was reported as having the highest anti-infective activity with an IC50 of 22.22 μM.
In this context, the chemical diversity of a registered plant NEs collection (containing 1600 samples) was leveraged. This set covers about 30% of all known botanical families and was systematically analysed by mass spectrometry-based metabolite profiling,3 thereby generating consequential amounts of spectral data from HRMS/MS analyses.
Spectral organization through molecular networking (MN) along with the GNPS platform have established themselves as tools to organize, visualize and compare HRMS/MS data since their inception in 2014.4,5 Yet it has been shown that their use can become challenging with the increasing size of datasets. Furthermore, large datasets tend to be more prone to chromatographic misalignments due to experimental factors. The iterative nature of the data acquisition processes and the fact that samples are likely analysed over large periods of time increases the risks for potential batch-effects.6 The 1600 NEs collection used in this study was no exception in that matter as was described by Allard et al. (2023).3 Although aligning UHPLC-HRMS/MS data does not generally pose any major problem when working with MNs in a dataset-centric manner, problems arise when it comes to incrementing data in an existing dataset as it would require re-processing the entire dataset. This problem changed the paradigm from dataset-centric approaches to sample-centric approaches relying on unaligned datasets organized in knowledge graph (KG) frameworks.7
The present study will establish a method to navigate large unaligned spectral spaces based on plant extracts content. As a proof-of-concept, chemically informed queries will be established to identify extracts likely to contain bioactive NPs. In the frame of the TB disease area, increasing specificity of therapeutic solutions can be addressed with more stringent phenotypic assays. As opposed to screening drug candidates on mycobacteria grown in broth, screening approaches that utilize infection models have been elaborated to better capture the specificities of Mtb's intracellular behavior.8,9 In this study, we employed an innovative 3R infection model using Mycobacterium marinum (Mm, ATCC BAA-535 (https://www.atcc.org/products/baa-535)) alongside the amoeba and professional phagocyte Dictyostelium discoideum (Dd, ATCC MYA-4120 (https://www.atcc.org/products/mya-4120)) as stand-ins for Mtb and macrophages, respectively. We used this model in a high-throughput phenotypic assay which enabled simultaneous assessment of a sample's impact on both the host and the pathogen, making it a particularly stringent assay.10
Combined with the above-mentioned assay, the biological screening of the 1600 NEs library provided an unbiased method for targeting NEs of interest. The current study focuses on one of the hit NEs identified during the bioactivity screening of the above-mentioned 1600 NEs collection. Based on the chemistry of this extract, the available metabolomic data of the KG was then explored to prioritize other potentially bioactive extracts.
In the first step, a selected hit NE was subjected to HPLC bioactivity profiling to identify the LC-peaks containing the bioactivity. MS-based annotation of the corresponding metabolites resulted in the identification of a common scaffold of biological interest. Then, the fragmentation spectra of those metabolites were used as proxies to search for analogues within all species contained in the KG. This strategy aimed at finding NEs containing a maximum number of diverse metabolites derived from a scaffold of interest and guide their isolation.
The screening of the 1600 NEs collection eventually yielded 12 anti-infective hits (hit rate: 0.75%), 8 of which were strict anti-infectives and 4 others were anti-infective Dd inhibitors (not covered in this study, described in the work of J. Nitschke (2024)13). Among those, the roots ethyl acetate extract of a Lardizabalaceae Stauntonia brunoniana (Decne.) Wall. Ex Hemsl. (Q10865526 (https://www.wikidata.org/wiki/Q10865526)) stood out as a strict anti-infective hit with good activity. To our knowledge, no phytochemical investigations were reported for this species at the time of writing (see WikiData query (https://w.wiki/9rLt)), it thus constituted the focus of the current study.
This NE was classified as a “strict anti-infective” as it displayed inhibition of intracellular bacterial growth (during infection) by 76% (Fig. 1B) and was only weakly active on the bacteria grown in vitro (29% inhibition of growth in Fig. 1A). The biological activity on the host was also assessed and did not show significant toxicity (33% inhibition of growth in Fig. 1C). Hence, the ethyl acetate extract of the roots of S. brunoniana was chosen for subsequent HPLC bioactivity profiling in order to highlight the LC-peaks responsible for this activity.
This process highlighted the most intense LC-MS-peaks in the MS trace as being stilbene derivatives. This was also corroborated by online-UV-PDA data for those peaks showing characteristic bands reported for stilbenes at 310 nm and 327 nm.20 The analyses revealed compound A with positive ion (PI) [M + H]+ m/z of 471.1437 for a MF C28H22O7 (Δ = −0.21 ppm), compound B with a PI [M + H]+ m/z of 455.1495 (MF C28H22O6, Δ = 1.32 ppm), compound C with a PI [M + H]+ m/z 697.2077 (MF C42H32O10, Δ = 1.29 ppm) and compound D with PI [M + H]+ m/z of 923.2699 (MF C56H42O13, Δ = 0.11 ppm). Dereplication of these ions based on their MS/MS spectra indicated that compounds A and B corresponded to stilbene dimers, while compound C corresponded to a trimer and compound D to a tetramer. The putative annotated structures obtained are displayed in Fig. 2 and corresponded to the best candidate structures proposed by the CSI:FingerID module of SIRIUS.15,16 Compound E on the other hand seemed to be a structural outlier which, unlike stilbenes, was not detectable in the UV trace.
To assess the bioactivity of these compounds putatively annotated as stilbenes in the historical collection of 1600 NEs, a new NE was prepared by Accelerated Solvent Extraction (ASE). The procedure used consisted in three successive extraction methods with solvents of increasing polarity (hexane, ethyl acetate and methanol respectively). It should be noted that historical ethyl acetate extracts (as reported in Allard et al.3) all underwent a pre-treatment aimed at reducing their content down to medium-polarity compounds, by removing very non-polar components through a DCM rinsing step on silica pads.3 For better scalability, this process was replaced by a hexane-extraction step before the ethyl acetate extraction in the current study to provide a similar removal of non-polar compounds.
The metabolite profile of the new NE matched well with the historical one (ESI Fig. 1†) and confirmed that compound E (absent in the new NE) was a contaminant. Structure E was reported as a component of meltable ink used in printing21 that was a contaminant common to the 20 adjacent samples in the series of 1600 NEs analysed.
The dried micro-fractions were subjected to biological testing with the same strategy as for screening the 1600 NEs collection. As previously, growth inhibition data from Mm grown in broth or infecting Dd, and on the growth of Dd itself were obtained for all micro-fractions generated. This enabled the generation of an extensive bioactivity profile across the whole chromatogram as displayed in Fig. 3. Results clearly indicated that micro-fractions M24-26 and M44 held most of the bioactivity measured at the NE level (red trace, Mm in infection, Fig. 3). They corresponded to the main UV-active peaks, which were also linked to the major metabolites of the NE according to the ELSD trace (ESI Fig. 1C,† ELSD chromatogram).
Additionally, micro-fractions M37 and M58 also showed moderate anti-infective activities relative to neighbouring fractions. The overlay of all three single-dose assay readouts also demonstrated that none of the fractions seemed to have any major effect on the growth of Dd (in green) nor on Mm in broth (in purple), thus confirming our observations of bioactivity on the plant NE (Fig. 1).
These bioactive micro-fractions were then subjected to UHPLC-PDA-CAD-HRMS/MS measurements to assess their content and purity. The analyses revealed that both M24-26 and M44 contained pure compounds (according to the Charged Aerosol Detector (CAD) semi-quantitative trace) and corresponded to compounds A with PI [M + H]+ m/z of 471.1434 and C with PI [M + H]+ m/z 697.2067 respectively. Meanwhile M37 corresponded to B, with a PI [M + H]+ m/z of 455.1487, while M58 corresponded to D with PI [M + H]+ m/z of 923.2673. This process clearly demonstrated that the activity of the NE was directly linked to its major stilbene constituents (Fig. 2).
The analyses of the micro-fractions revealed that the HPLC bioactivity profiling chromatographic resolution was sufficient to isolate the main constituents of the NE in amounts sufficient for preliminary biological testing. One drawback though was that it did not allow for precise weight measurements. The real potency of such stilbenes in terms of anti-infective activities was consequently hard to assess at this stage.
These results on S. brunoniana pushed for the evaluation of the 1600 NEs collection to find plants containing a large diversity of stilbenes, while including those highlighted in S. brunoniana. To do so, the chemical- and spectral space of the 1600 NEs collection was explored with dedicated computational approaches. The aim was to efficiently identify NEs rich in stilbene for targeted isolation at a scale compatible with detailed biological assessment and unambiguous structural identification of the largest possible number of stilbene derivatives.
For this, ENPKG organizes and stores all unaligned HRMS and HRMS/MS data along with associated information such as retention time, peak area, annotations (GNPS/ISDB/SIRIUS structural annotations and CANOPUS chemical class annotations), and more, on a sample-by-sample basis. The data is stored as semantic Resource Description Framework (RDF)-type triples (subject-predicate-object), allowing for incremental data storage from independent metabolite profiling series. Additionally, ENPKG can incorporate several data types like biological readouts (such as anti-infective activities) and previous phytochemical knowledge. The recovery of meaningful information from the KG is performed through federated SPARQL queries (https://www.w3.org/TR/2008/REC-rdf-sparql-query-20080115/)24 and in the frame of this study, is entirely reliant data from the publicly available and previously published instance of ENPKG.7 They allow for the retrieval and combination of data from the ENPKG triple store. In the case of the 1600 NE-collection, this corresponded to over 200 million triples that were stored (features, different types of annotations, biological readouts, …). For example, the structure and chemical class annotations from SIRIUS could be retrieved and linked to specific NEs. Of special interest in this study was the conversion of the MS/MS data of each feature into documents (i.e. lists) of peaks and losses through Spec2Vec25 as part of the ENPKG workflow. Each peak in an MS/MS spectrum was represented by the word ‘peak@xxx.xx’, and the neutral losses were calculated as precursorm/z minus peakm/z and represented as ‘loss@xxx.xx’ (see Fig. 4A). The peaks and losses of each MS/MS of each feature were stored as triples (e.g. document of feature 62 → has_spec2vec_peak → peak of value 923.27, …) in the KG (ESI Fig. 2†). This enabled the search and comparison of spectral fingerprints (documents) through SPARQL queries. In the context of this study and to identify NEs rich in stilbenes, a two-step strategy was envisioned, based on annotations at the chemical class level and spectral similarity. First, a SPARQL query (ESI query 1†) that counted the features annotated as ‘oligomeric stilbenes’ by CANOPUS17 was applied for each sample over the whole collection of 1600 NEs. This chemical class was chosen as the structures of compounds A–D all fell into this category according to NPClassifier.18 This query produced a list of 163 NEs with at least one feature annotated as ‘oligomeric stilbene’ (ESI Table 1†).
![]() | ||
Fig. 4 Overview of the workflow using Jaccard Similarity Indices (JSI). (A) Spectrum to document: each feature's spectral information was transformed into a document of peaks and losses using Spec2Vec within the ENPKG workflow. The peaks in the MS/MS spectra were represented as words (e.g., ‘peak@xxx.xx’) and the neutral losses as ‘loss@xxx.xx’, corresponding to the mass differences between two peaks.25 (B) Search for structural analogues: the search for structural analogues using Jaccard Similarity Indices (JSI) began with the definition of a proxy MS/MS scan of the molecule of interest contained in the Knowledge Graph (KG). In this example, the proxy scan was the feature annotated as ‘dimer A’, focusing on stilbene-like structures. This information was used in a SPARQL query to calculate the JSI and retrieve all features within the KG that met the defined JSI threshold (JSI %3e 0.25). The results table listed each feature, with an in-house script grouping the features sample-wise to obtain the total count of features tagged as stilbene-like structures. (C) JSI calculation: illustration of the JSI calculation between two documents (proxy and a given scan) in the KG through the SPARQL query. (D) MS/MS spectra comparison: MS/MS spectra mirror comparison between the MS/MS of the proxy feature and three different scans in a given sample. The cosine similarity (CS) value was shown as well [link scan 5 vs. scan 1, mirror plot] [link scan 5 vs. scan 8, mirror plot]. |
Then, to complement the NE selection process, a method was implemented to directly navigate the spectral space and search for stilbene fragmentation patterns based on peaks and losses in all the spectral documents stored in the ENPKG. For this, a SPARQL query applying Jaccard similarity index (JSI)26 was developed (see Fig. 4B). JSIs measured the similarity between two MS/MS spectra (stored as ‘text’ documents in the KG) by comparing their intersection with their union. The JSI is determined by dividing the number of unique common observations in two sets by the total number of unique observations in either set (see Fig. 4C). The results ranged from 0 to 1, with 0 meaning that two features are completely dissimilar and 1 meaning that they are identical features. To perform this calculation in the ENPKG, a “proxy” or “feature query” (a reference MS/MS spectrum) included in the KG had to be selected.
The spectral documents of the features dereplicated as stilbenes (dimer A, trimer C, and tetramer D) in S. brunoniana mentioned above were used individually as proxies in the SPARQL query. Fig. 4C provided an overview of the calculations for ‘Compound A’ as a proxy against one of the features (feature 1) of one of the samples (extract A) in the ENPKG. The Spec2Vec document of the proxy was automatically compared to that of ‘feature 1’ in sample ‘Extract A’. The intersection and union were computed, and the ratio corresponded to the Jaccard similarity index (JSI = 0.68). This calculation was repeated for all features in the ENPKG (ca. 1 million features). The results returned by the query included the sample codes, the scan number, retention time, and parent mass of each ‘similar’ feature in addition to the JSI. This provided a list of features that could be sorted according to JSI values. The top feature candidates were those most similar to the proxy in the whole dataset. This procedure was repeated for the other proxies of S. brunoniana (ESI query 2, 3, and 4†).
An in-house script (https://github.com/luigiquiros/Anti-infective-stilbenes-publication) (available on github (https://github.com/luigiquiros/Anti-infective-stilbenes-publication)) was used to combine the results of the three different queries for compounds A, C, and D, to express the results as the number of stilbene-like features per sample. Features were only considered if they had a JSI higher than the established threshold of 0.25. This value was established based on the iterative application of the same type of query over several types of compounds, as well as an observation of the statistical distribution of JSI indices for a set query (see ESI Fig. 3†). An acceptable match between the proxy and the features present in other NEs containing it, as well as structural analogues, was obtained with JSI > 0.25. To prove the acceptability of this threshold, an additional query that added the information of chemical classes for each feature when calculating JSI scores was implemented. Among the 1000 features with the highest JSI scores compared to compound A, 137 features had a JSI ≥ 0.25 (ESI Fig. 3†). Of those 137, 43 features were annotated as “oligomeric stilbenes”, 2 features were annotated as “unknown”, and the remainder (92 features) were not annotated by CANOPUS. No other chemical class than the correct “oligomeric stilbenes” appeared within the features with JSI ≥ 0.25, which suggested that the rate of “false positives” arising using this threshold is likely low.
This information was added to the list of the previously obtained CANOPUS chemical class count (ESI Table 1†). From this combined list, only the top 10% of NEs (according to their CANOPUS chemical class count), with enough dry plant material available, were considered for further analyses. This reduced the list to 14 NEs presented in Table 1. The best candidates according to the JSI query results were used to select NEs for in depth phytochemical analysis as shown in the following section.
Species | Organ | CANOPUS class count | JSI feature count* | Bacterial growth inhibition in infection (%) | Amoebal growth inhibition in infection (%) | Bacterial growth inhibition in broth (%) |
---|---|---|---|---|---|---|
Gnetum edule | Roots | 60 | 36 | 52 | 0 | 18 |
Stauntonia brunoniana | Roots | 48 | 33 | 76 | 33 | 29 |
Ampelocissus arachnoidea | Multiple | 95 | 9 | 3 | 0 | 0 |
Gnetum edule | Stems | 36 | 8 | 20 | 13 | 1 |
Shorea roxburghii | Bark | 68 | 5 | 46 | 0 | 8 |
Holoptelea integrifolia | Roots | 48 | 5 | 67 | 53 | 22 |
Rheum rhabarbarum | Roots | 37 | 5 | 41 | 10 | 25 |
Shorea roxburghii | Leaves | 33 | 4 | 0 | 0 | 0 |
Parashorea dussaudii | Stems | 32 | 4 | 0 | 0 | 13 |
Rheum officinale | Roots | 35 | 4 | 34 | 11 | 27 |
Hopea chinensis | Roots | 56 | 3 | 37 | 17 | 14 |
Hopea chinensis | Stems | 36 | 3 | 0 | 0 | 9 |
Hopea helferi | Stems | 29 | 3 | 12 | 8 | 0 |
Hopea helferi | Roots | 35 | 2 | 16 | 3 | 0 |
Gnetum edule (Willd.) Blume (roots) stood out as the most promising candidate with the highest number of stilbene features and an improved bioactivity profile with lower toxicity on the host in infection (although it also has lower anti-infective activity than S. brunoniana). Interestingly, the richest plant in terms of CANOPUS stilbene class count was Ampelocissus arachnoidea with 95 ions, but this did not translate in a significant number of stilbenes flagged by the JSI metric (highlighting only 9 stilbene-like features). It can be explained by the fact that stilbenes within A. arachnoidea were structurally different from the stilbene proxies used to calculate JSI (according to annotations made, differing notably in the way stilbene monomers were branched together). This NE was also barely bioactive, which suggested that not all ‘oligomeric stilbenes’ flagged in this manner were equal in terms of biological activities. Those results thus highlighted the importance of using the second filter, which targeted the stilbene sub-structure. To explore potential differences among stilbenes themselves, a feature-based molecular network (FBMN27) was created using only the LC-HRMS/MS data of the 14 selected NEs.
![]() | ||
Fig. 5 Stilbene clusters selected from the Molecular Network (MN) of the 14 stilbene-containing extracts. All four compounds annotated in Fig. 2 are found in the left-hand-side cluster (compounds A–D). Pie charts inside each node represent the relative intensity of the corresponding ion within each of the extracts they are contained in. Nodes of the left-hand-side cluster show ions most intense in two extracts: Stauntonia brunoniana and Gnetum edule. The right-hand-side cluster shows other nodes annotated as stilbenes which were mostly found in the extract of G. edule but were not found in S. brunoniana. |
Interestingly, another feature (with positive m/z of 713.2004) appearing abundantly in G. edule was not previously identified and additional stilbene isomers, not present in S. brunoniana, were spotted in the same cluster (cluster II, Fig. 5). These observations were in line with the data in Table 1, suggesting that G. edule may have higher chemical diversity when it comes to stilbene derivatives, with 60 features identified by CANOPUS as ‘oligomeric stilbenes’ against only 48 in S. brunoniana. For all those reasons, it was decided that G. edule would be the best candidate to undergo an in-depth phytochemical investigation and obtain a representative set of stilbenes for bioactivity assessments.
A total of six structures among those in Fig. 6 were known (1, 3, 5, 7–9) and matched previously reported 1H and 13C NMR data. Gnetin D (1, Q104666844 (https://www.wikidata.org/wiki/Q104666844)) and gnetin C (3, Q11300131 (https://www.wikidata.org/wiki/Q11300131)) were previously isolated from the woods of Gnetum leyboldii lianas (Q15049867 (https://www.wikidata.org/wiki/Q15049867)),29 latifolol (5, Q105143065 (https://www.wikidata.org/wiki/Q105143065)) was isolated from the stem of Gnetum latifolium (Q15050020 (https://www.wikidata.org/wiki/Q15050020)),30 macrostachyol A (7) was isolated from the roots of Gnetum macrostachyum (Q17013764 (https://www.wikidata.org/wiki/Q17013764)),31 gnemonol B (8, Q105384666 (https://www.wikidata.org/wiki/Q105384666)) was isolated from the roots of Gnetum gnemon (Q72368 (https://www.wikidata.org/wiki/Q72368))32 and gnemontanin G (9) was isolated from the caulis of Gnetum montanum Markgr. (Q10879811 (https://www.wikidata.org/wiki/Q10879811)).33
Among these stilbenes isolated from G. edule, four of them matched the four peaks annotated as stilbenes in S. brunoniana and serving as proxies (A–D) for this type of compounds in the whole 1600 NEs dataset. These stilbenes were gnetin D (1, compound A), gnetin C (3, compound C), latifolol (5, compound B) and macrostachyol A (7, compound D). The final structures obtained were in good agreement with the core skeleton dereplicated from the micro-fractions. The full de novo identification process enabled the unambiguous establishment of the stereochemistry and ascertainment of the hydroxylation pattern on this type of molecules. Comparison of the metabolite profiling on high-resolution profiles between G. edule and S. brunoniana also indicated that the retention times (RTs) and MS/MS spectra for all four reference compounds were matching between both NEs (as was shown on the FBMN, Fig. 5). All this evidence therefore confirmed that the biological activity in S. brunoniana was indeed linked to these four stilbenes.
Additionally, five of the isolated stilbenes were found to be isomers of known compounds, yet their stereochemistry did not match any reported structures. One of them was identified as a new stilbene and named gnetoline A (4) and four of them were enantiomers of known compounds (2, 6, 10 and 11). The absolute stereochemistry of latifolol (5) was also established here, whereas previous reports had only established its relative stereochemistry.
Compound 2 was isolated as a brown amorphous powder with an [M + H]+ of m/z 471.1433 corresponding to a MF of C28H22O7 (Δ = −1.06 ppm). The 1H and 13C NMR data (Table 2, Fig. S2.4. and S2.5., ESI†) showed that it corresponded to gnetupendin C,20 the 7′-8′ cis isomer of gnetin D. Both the 3JH-7′-H-8′ value of 7.7 Hz and the dipolar correlation from H-7′ to H-8′ and from H-6′ and H-10′/14′ confirmed the cis configuration of H-7′ and H-8′. However, the specific rotation value ([α]20D +69.6 for 2 against [α]20D −220.0 reported for gnetupendin C20) indicated that 2 was the enantiomer of (−)-gnetupendin C. The absolute configuration of the latter had not been defined previously and the Electronic Circular Dichroism (ECD) measurements carried out on compound 2 did not provide concluding information. Therefore compound 2 was identified as (+)-gnetupendin C with a cis-relative stereochemistry (7′S,8′R or 7′R,8′S).
Position | 2 | 10 | ||
---|---|---|---|---|
1H | 13C | 1H | 13C | |
1 | 128.2 | 115.3 | ||
2 | 7.41 d (8.6) | 127.9 | 156.1 | |
3 | 6.77 d (8.6) | 115.5 | 6.33 d (2.5) | 102.6 |
4 | 157.2 | 158.1 | ||
5 | 6.77 d (8.6) | 115.5 | 6.25 dd (8.5, 2.5) | 107.2 |
6 | 7.41 d (8.6) | 127.9 | 7.35 d (8.5) | 127.2 |
7 | 7.03 d (16.3) | 127.9 | 7.19 d (16.4) | 123.3 |
8 | 6.91 d (16.3) | 125.7 | 6.85 d (16.4) | 124.7 |
9 | 139.0 | 140.1 | ||
10 | 6.70 d (1.3) | 98.1 | 6.54 d (1.3) | 97.8 |
11 | 161.3 | 161.7 | ||
12 | 116.8 | 114.3 | ||
13 | 154.2 | 154.6 | ||
14 | 6.49 d (1.3) | 107.3 | 6.44 d (1.3) | 106.1 |
1′ | 114.5 | 118.7 | ||
2′ | 154.4 | 155.3 | ||
3′ | 6.16 d (2.3) | 101.6 | 6.32 d (2.4) | 102.5 |
4′ | 156.9 | 157.9 | ||
5′ | 5.94 dd (8.3, 2.3) | 105.4 | 6.14 dd (8.4, 2.4) | 105.9 |
6′ | 6.75 d (8.3) | 127.4 | 6.84 d (8.4) | 126.5 |
7′ | 5.84 d (7.7) | 84.8 | 5.53 d (3.4) | 87.6 |
8′ | 4.45 d (7.7) | 48.6 | 4.18 d (3.4) | 52.8 |
9′ | 141.6 | 145.7 | ||
10′ | 5.63 d (2.2) | 106.8 | 6.04 d (2.2) | 105.5 |
11′ | 157.1 | 158.1 | ||
12′ | 5.78 t (2.2) | 100.6 | 6.02 t (2.2) | 100.6 |
13′ | 157.1 | 158.1 | ||
14′ | 5.63 d (2.2) | 106.8 | 6.04 d (2.2) | 105.5 |
2-OH | — | 9.59 s | ||
4-OH | 9.55 s | 9.40 s | ||
13-OH | 9.23 s | 9.20 s | ||
2′-OH | 9.37 s | 9.54 s | ||
4′-OH | 8.97 s | 9.24 s | ||
11′-OH | 8.71 s | 9.04 s | ||
13′-OH | 8.71 s | 9.04 s |
Compound 10 was also a dimer of stilbene isolated as a brown amorphous powder with an [M + H]+ of m/z 487.1382 corresponding to a MF of C28H22O8 (Δ = −1.03 ppm). The 1H and 13C NMR data (Table 2, Fig. S10.4. and S10.5.†) followed patterns reported for (+)-gnetumontanin A.34 Additional measurements distinguished 10 from (+)-gnetumontanin A: opposing specific rotation values ([α]20D −13.4 for 10 against [α]22D +17 reported for (+)-gnetumontanin A34). Compound 10 was thus identified as (−)-gnetumontanin A.
Compound 4 was a trimer of stilbene isolated as brown amorphous powder with [M + H]+ of m/z 697.2068 corresponding to MF of C42H32O10 (Δ = 0 ppm). Its NMR data (Table 3 and Fig. S.4.4†) showed similarities with that of latifolol (5),30 leading to the conclusion that they share the same planar structure. For latifolol, the relative configuration of each dihydrofurane ring was identified as trans based on the 3JH-7′–H-8′ and 3JH-7′′–H-8′′ values (4.0 and 4.2 Hz, respectively) and from the ROESY correlations from H-7′ (or H-7′′) to H-10′/H-14′ (or H-10′′/14′′) and from H-8′ (or H-8′′) to H-2′/6′ (or H-2′′). For compound 4, the relative configuration was defined as trans for H-7′ and H-8′ (3JH-7′–H-8′ = 3.9 Hz; and same ROESY as those described for latifolol), and cis for H-7′′ and H-8′′ (3JH-7′–H-8′ = 7.9 Hz; and ROESY correlations from H-6′′ to H-10′′/14′′). As the absolute configuration of latifolol (5) did not seem to have been defined previously, ECD calculations were carried out for the 4 possible isomers and compared with the experimental data (S5.1†). The best fit was observed for the 7′R,8′R,7′′S,8′′S isomer, for which the specific rotation value −12.4 (c 0.12, MeOH) was measured (literature:30 [α]20D −42 (c 0.15, MeOH)). The absolute configuration of latifolol (5) was thus defined here as (E)-7′R,8′R,7′′S,8′′S for the first time. For compound 4, the presence of an unidentified stilbene dimer mixed with it meant that potential distortions in the ECD spectrum could be expected. The experimental ECD (S4.1†) trace did match that of the 7′R,8′R,7′′R,8′′S-configuration and compound 4 was thus identified as a new (E)-7′R,8′R,7′′R,8′′S-isomer of latifolol (5) and was named gnetoline A (4).
Position | 4 | 6 | 11 | |||
---|---|---|---|---|---|---|
1H | 13C | 1H | 13C | 1H | 13C | |
1 | 128.1 | 128.1 | 128.0 | |||
2 | 7.42 d (8.6) | 127.9 | 7.41 d (8.7) | 127.9 | 7.09 d (8.7) | 127.7 |
3 | 6.76 d (8.6) | 115.5 | 6.76 m | 115.5 | 6.68 d (8.7) | 115.5 |
4 | 157.3 | 157.3 | 157.3 | |||
5 | 6.76 d (8.6) | 115.5 | 6.6 m | 115.5 | 6.68 d (8.7) | 115.5 |
6 | 7.42 d (8.6) | 127.9 | 7.41 d (8.7) | 127.9 | 7.09 d (8.7) | 127.7 |
7 | 7.03 d (16.3) | 128.2 | 7.03 d (16.3) | 128.2 | 6.83 d (16.3) | 128.8 |
8 | 6.92 d (16.3) | 125.5 | 6.91 d (16.3) | 125.5 | 6.58 d (16.3) | 122.0 |
9 | 139.7 | 139.7 | 134.7 | |||
10 | 6.68 d (1.1) | 97.7 | 6.68 d (0.9) | 97.8 | 118.9 | |
11 | 161.5 | 161.5 | 160.8 | |||
12 | 113.9 | 113.9 | 6.23 d (2.0) | 96.0 | ||
13 | 154.6 | 154.6 | 158.4 | |||
14 | 6.50 d (1.1) | 107.3 | 6.50 d (0.9) | 107.3 | 6.59 d (2.0) | 102.9 |
1′ | 132.3 | 132.2 | 112.8 | |||
2′ | 7.16 d (8.6) | 126.7 | 7.15 d (8.6) | 126.8 | 158.9 | |
3′ | 6.76 d (8.6) | 115.3 | 6.76 m | 115.3 | 115.1 | |
4′ | 157.2 | 157.3 | 154.9 | |||
5′ | 6.76 d (8.6) | 115.3 | 6.76 m | 115.3 | 6.31 d (8.5) | 108.6 |
6′ | 7.16 d (8.6) | 126.7 | 7.15 d (8.6) | 126.8 | 7.05 d (8.5) | 127.6 |
7′ | 5.40 d (3.9) | 91.9 | 5.40 d (4.0) | 91.9 | 5.61 d (6.6) | 88.2 |
8′ | 4.35 d (3.9) | 54.2 | 4.34 d (4.0) | 54.2 | 4.65 d (6.6) | 53.9 |
9′ | 144.5 | 145.1 | 145.5 | |||
10′ | 6.19 s | 99.7 | 6.16 d (1.3) | 99.4 | 6.10 d (2.2) | 105.7 |
11′ | 160.9 | 161.0 | 158.6 | |||
12′ | 115.8 | 113.1 | 6.07 t (2.2) | 101.1 | ||
13′ | 154.2 | 154.5 | 158.6 | |||
14′ | 6.15 s | 107.3 | 6.13 d (1.3) | 107.1 | 6.10 d (2.2) | 105.7 |
1′′ | 114.6 | 131.7 | 118.6 | |||
2′′ | 154.3 | 7.10 d (8.6) | 127.2 | 154.8 | ||
3′′ | 6.13 d (2.3) | 101.6 | 6.76 m | 115.3 | 6.32 d (2.5) | 102.4 |
4′′ | 156.8 | 157.2 | 157.6 | |||
5′′ | 5.93 dd (8.4, 2.3) | 105.4 | 6.76 m | 115.3 | 6.07 dd (8.4, 2.5) | 105.9 |
6′′ | 6.75 d (8.4) | 127.4 | 7.10 d (8.6) | 127.2 | 6.55 d (8.4) | 125.8 |
7′′ | 5.82 d (7.9) | 85.0 | 5.27 d (5.7) | 92.4 | 5.54 d (2.7) | 88.2 |
8′′ | 4.42 d (7.9) | 48.5 | 4.22 d (5.7) | 54.3 | 4.10 d (2.7) | 53.3 |
9′′ | 141.6 | 144.7 | 145.5 | |||
10′′ | 5.63 d (2.2) | 106.8 | 5.97 d (2.2) | 105.4 | 6.08 d (2.2) | 105.6 |
11′′ | 157.1 | 158.3 | 158.1 | |||
12′′ | 5.78 t (2.2) | 100.5 | 6.03 t (2.2) | 100.9 | 6.04 t (2.2) | 100.7 |
13′′ | 157.1 | 158.3 | 158.1 | |||
14′′ | 5.63 d (2.2) | 106.8 | 5.97 d (2.2) | 105.4 | 6.08 d (2.2) | 105.6 |
4-OH | 9.55 s | 9.49 s or 9.56 s | 9.57 s | |||
13-OH | 9.40 s | 9.40 s | 9.37 s | |||
4′-OH | 9.48 s | 9.49 s or 9.56 s | 9.37 s | |||
11′-OH | — | — | 9.15 s | |||
13′-OH | 9.18 s | 9.23 s | 9.15 s | |||
2′′-OH | 9.33 s | — | 9.53 s | |||
4′′-OH | 8.95 s | 9.49 s or 9.56 s | 9.17 s | |||
11′′-OH | 8.72 s | 9.10 s | 9.07 s | |||
13′′-OH | 8.72 s | 9.10 s | 9.07 s |
Compound 6 was also a trimer of stilbene isolated as brown amorphous powder with [M + H]+ of m/z 681.2120 corresponding respectively to a MF of C42H33O9 (Δ = 0.15 ppm). The 1H and 13C NMR data (Table 3, Fig. S6.4. and S6.5.†) followed patterns of trans,trans-configuration reported for gnetin E,29 with no absolute configuration established in previous reports. ECD measurements (S6.1†) were therefore carried out to establish the absolute configuration 7′R,8′R,7′′S,8′′S for (−)-gnetin E (6) for the first time.
Compound 11 is a trimer of stilbene with an [M + H]+ of m/z 713.2018 corresponding to a MF of C42H32O11 (Δ = 0.14 ppm). The NMR data (Table 3, Fig. S11.4. and S11.5.†) was similar to that reported for (−)-gnetuhainin M.35 Measurements of specific rotation values ([α]20D−44.9 for 11 against [α]25D −32.9 reported for (−)-gnetuhainin M) also suggested that the isolated compound matched with the reported (−)-gnetuhainin M.35 Only the trans,trans relative stereochemistry was established in that report however, there were still four different absolute stereoisomers possible (Fig. 7). The experimental ECD trace was compared to the four traces of the possible absolute stereoisomers calculated using time-dependant density functional theory (TD-DFT).36–38 The experimental trace matched that of isomer 7′S,8′S,7′′R,8′′R, allowing for the assignment of the absolute stereochemistry (E)-7′S,8′S,7′′R,8′′R for (−)-gnetuhainin M (Fig. 7) for the first time.
Compound | IC50 Mm in infection (μM) | IC50 Dd in infection (μM) | IC50 Mm in broth (μM) |
---|---|---|---|
1 | >100 | >100 | >100 |
2 | 66.67 | >100 | >100 |
3 | >100 | >100 | >100 |
4 | 66.67 | 66.67 | >100 |
5 | 66.67 | 66.67 | >100 |
6 | 66.67 | >100 | >100 |
7 | 66.67 | 66.67 | >100 |
8 | 66.67 | 66.67 | >100 |
9 | >100 | >100 | >100 |
10 | >100 | >100 | >100 |
11 | 22.22 | >100 | >100 |
The biological readouts of single compounds did align well with the biological readout expected at the NE level. In fact, none of these compounds showed any activity on Mm in broth and their effect on the growth of the host Dd remained largely limited, qualifying most of them as strict anti-infectives, as was the case for the NE. IC50 values of isolated compounds were generally high (i.e. lower activity), compared to the strong activities observed in fractions from the HPLC bioactivity profiling. This can be attributed to the inherent drawback of the latter approach, in that it doesn't allow for precise weight measurements of micro-fractions. In that process, all micro-fractions are tested at a single concentration according to an average weight (calculated as the total amount of extract injected divided by the number of micro-fractions). This induced increased uncertainties, resulting in bioactivities being likely overestimated for major compounds of an extract.
In absolute terms, compound 11 was the best candidate for anti-infective activities with an IC50 of 22.22 μM. In terms of its chemical structure, it was also an outlier compared to other compounds, which was observable in the form of its linkage between monomers 1 and 2 (Fig. 9). It was fair to assume therefore that this structural difference may have had an impact on the activity levels observed. Additionally, in spite of their appearance as complex three-dimensional entities, these compounds are in fact fairly rigid in terms of conformational movements.
Clearly, all 14 NEs contained an abundance of stilbenes. In the case of G. edule, they were clearly abundant in the roots, in the woody stems however they were not detectable on the CAD trace, but the MS detection revealed their presence. The CAD semi-quantitative analysis revealed that there seemed to be no direct correlation between the general abundance of stilbenes and the biological activity of the NE. In fact, this was also reflected in the isolated stilbenes, exhibiting very variable levels of anti-infective activity. The alignment of all CAD traces revealed that most of the compounds isolated from G. edule (roots) and those flagged in S. brunoniana seemed to be specific to these two NEs, with many stilbenes not appearing in any other NE (2, 4, 5, 7, 11). The other stilbene-rich plants highlighted by the JSI search in the KG were likely to contain other forms of stilbene (generally more polar), eluting before 8 minutes.
Considering the significant size of the dataset and unlike in MNs approaches, the exploration of the spectral space was based on MS/MS data stored through the Spec2Vec pipeline. This is computationally friendly format, which comes with an inherent drawback being the absence of intensity values in the resulting data format (documents). The JSI score developed in this study was indeed effective in handling such format. One major resulting pitfall, highlighted by low JSI value, was that the noise in the MS/MS spectra could not be filtered out based on intensity values in the current KG. Future developments of the KG will aim to overcome this problem by integrating the MS/MS peaks intensities information during raw data processing. Such adaptation should allow direct filtering based on spectral intensities, a critical parameter that was lost in spectral document conversion. As shown however, this did not impede the relevance of JSI scores to successfully highlight chemical similarity across samples. This enabled the prioritization of extracts similar in composition to a reference extract with annotated bioactive LC-peaks as proxies.
This approach also yielded chemical novelty, with 6 newly characterized isomers out of 11 chemical entities isolated from G. edule. (−)-Gnetuhainin M (11) in particular stood out from all other compounds, not only due to its specific structural features, but also for its anti-infective activity being significantly higher, with an IC50 of 22.22 μM for the inhibition of bacterial growth in infection. These structures were reminiscent of trans-δ-viniferins, which are another type of stilbene dimers widely reported for various biological activities,39 including anti-infective effects. Furthermore, studies on trans-δ-viniferins obtained by biotransformation showed that O-methylation on some of the phenol groups appeared to have a beneficial effect on the anti-infective activity, confirming interest for this type of scaffold.11 Selective O-methylation of (−)-gnetuhainin M (11) might therefore lead to further improvement of its anti-infective activity.
An additional observation concerning biological activities was the pro-infective effect observed with individual compounds at mid-level concentrations, which remained largely unexplained. This pro-infective effect was reminiscent of the behaviour displayed by rapamycin (ESI Fig. 4†), a known autophagy inducer (inhibitor of the mTOR pathway which negatively regulates autophagy).40 Further mechanism of action studies on this class of compounds might provide additional information to explain such a phenomenon. Altogether, the combination of metabolomic data in a KG with stringent in cellulo anti-infective assays holds promises to highlight other scaffolds of interest in the chemical/spectral space of large metabolomic datasets.
The centroid data-dependent MS2 (dd-MS2) scan acquisition events were performed in discovery mode, triggered by apex detection with a trigger detection (%) of 300 with a maximum injection time of 120 ms, performing 1 microscan. The top 3 abundant precursors (charge states 1 and 2) within an isolation window of 1.2 m/z were considered for MS/MS analysis. For precursor fragmentation in the HCD mode, a normalized collision energy of 15, 30 and 45% was used. Data was recorded in profile mode (use EASY-IC(TM): ON).
The chromatographic separation was done on a Waters BEH C18 column (50 × 2.1 mm i.d., 1.7 μm, Waters, Milford, MA) using the following gradient (time (min), % B): 5% B from 0 to 0.5 min; from 5% B to 100% B between 0.5 and 7 min; 100% B from 7 to 8 min, from 100% B to 5% B from 8 to 8.10 min; 5% B from 8.10 to 10 min. The mobile phases were H2O (A) and MeCN (B) both containing 0.1% FA. The flow rate was set to 600 μL min−1, the injection volume was 2 μL and the column was kept at 40 °C. The PDA detector was used from 210 to 400 nm with a resolution of 1.2 nm. The CAD detector was kept at 40 °C, with 5 bar N2 and power function 1 for a data collection rate of 20 Hz.
As described in Nitschke et al.11 and Mottet et al.,41 D. discoideum Ax2(ka) expressing mCherry at the act5 locus42 was infected with M. marinum M strain expressing the lux operon (luxCDABE)43,44 by spinoculation.
Briefly, the day before the experiment M. marinum was cultivated in 7H9 broth (Becton Dickinson, Difco Middlebrook 7H9) supplemented with 10% OADC (Becton Dickinson) and 0.05% tyloxapol (Sigma Aldrich) and 50 μg mL−1 kanamycin at 32 °C overnight with continuous shaking. Additionally, the day before the experiment 107 amoebae were plated in HL5-C in a 10 cm Petri dish (Falcon). On the day of the experiment, a volume of the M. marinum culture corresponding to a multiplicity of infection of 25 with respect to the number of amoeba in the Petri dish was taken and added to the amoeba, subsequently the Petri dishes were centrifuged twice at 500×g, as described in Mottet et al.41 To remove extracellular bacteria, dishes were washed with fresh HL5-C and cells were resuspended in HL5-C with 5 U mL−1 penicillin and 5 μg mL−1 of streptomycin (Gibco) to inhibit extracellular growth of bacteria during the course of the experiment.
For testing fractions and purified compounds, 20 μL of infected cell suspension was plated into each well of a 384-well plate (Interchim FP-BA8240) to an effective cell number of 104 cells per well. Fractions or compounds including a vehicle control (0.3% DMSO final concentration) and a positive control (rifabutin, 10 μM final concentration) were added using an electronic multipipette (Sartorius). Subsequently, the well plates were sealed with a gas impermeable membrane (H769.1, Carl Roth), briefly centrifuged and intracellular bacterial growth was monitored using an Agilent BioTek H1 plate reader, an Agilent BioTek BioStack plate stacker, by recording luminescence over 72 hours at 25 °C with readings taken every hour. Fluorescence was also recorded to monitor amoeba growth.
For testing bacteria in broth, the pre-culture was diluted to a bacterial density of 3.75 × 105 bacteria per mL in 7H9 medium. Plating bacteria and compounds or fractions was performed analogously to the infection assay described above. Bacteria growth was monitored with the Agilent BioTek H1 plate reader by recording luminescence at 32 °C.
For both assays, growth curves were obtained by measuring the luminescence and fluorescence as a proxy for bacterial growth and host growth, respectively, for 72 hours with time-points taken every hour. The “normalized residual growth” was computed by calculating the area under the curve (AUC, trapezoid method) and normalization to the vehicle control (0.3% DMSO, set at 1) and a baseline curve (set at 0). The baseline curve was calculated by taking the median of the first measurement of all wells in a plate and extrapolating it over the full time course. The threshold for hit detection was arbitrarily fixed at a cut-off of normalized residual growth ≤0.5. Normalized values were averaged over technical and biological replicates (all experiments on isolated compounds have at least n = 3 and N = 3, whereas the primary extract screening and micro-fractions testing had values of n = 1 and N = 3).
This procedure was also applied to screening NEs, with slight modifications. The day before the experiment we pre-plated 10 μL HL5-C using a dispenser (Thermo Multidrop), subsequently we pre-plated 2.2 μL of dissolved NEs from 96-well plates into quadrants 1, 2 and 3 of 384-well plates, whereas quadrant 4 was used for positive and vehicle controls. Pre-plating of NEs was performed using a liquid handler (Agilent Bravo). In total, 24 NE plates were distributed in triplicates into eight 384-well plates, amounting to 24 assay plates. The prepared plates were sealed and stored at 4 °C overnight. On the day of the experiment, nine 10 cm Petri dishes were infected (as described before) and grouped into three pools. The cell suspension was adjusted to 106 cells per mL and 10 μL were plated into the 24 assay plates, resulting in 104 cells per well, as used for conventional infection experiments. Plates were sealed and placed in the plate stacker that supplies the plate reader. The same procedure was used to screen the same NEs on Mm in broth. For both assays, the first timepoints and a timepoint after 72 hours was recorded. Subsequently, we normalized the end point, first with the median of the full assay plate at the first time point, and second with the endpoint of the vehicle controls in the respective assay plate.
For IC50 estimation, we used a rudimentary approach of interpolating the sample concentration between the two normalized residual growth values which were closest to a value of 0.5.
(+)-Gnetin D (1) [α]20D +15.7 (c 0.10, MeOH); UV (MeOH) λmax (logε) 226 (4.27), 287 (3.89), 310 (4.02), 328 (4.04), 347 (3.76) nm; 1H NMR (DMSO, 600 MHz) δ 9.54 (1H, s), 9.24 (2H, s), 9.05 (2H, s), 8.22 (1H, s), 7.41 (2H, d, J = 8.7 Hz), 7.02 (1H, d, J = 16.3 Hz), 6.90 (1H, d, J = 16.2 Hz), 6.85 (1H, d, J = 8.4 Hz), 6.76 (2H, d, J = 8.6 Hz), 6.66 (1H, s), 6.45 (1H, s), 6.33 (1H, d, J = 2.3 Hz), 6.15 (1H, dd, J = 8.4, 2.3 Hz), 6.05 (2H, d, J = 2.2 Hz), 6.04 (1H, d, J = 2.2 Hz), 5.54 (1H, d, J = 3.4 Hz), 4.20 (1H, d, J = 3.5 Hz); 13C NMR (DMSO, 151 MHz) δ 161.8, 158.2, 157.9, 157.2, 155.3, 154.6, 145.7, 139.4, 128.2, 127.9, 126.6, 125.6, 118.7, 115.6, 114.8, 107.1, 105.9, 105.5, 102.5, 100.7, 97.7, 87.7, 52.8 (NP-MRD ID: NP0061299 (https://np-mrd.org/natural_products/NP0061299)); HRESIMS m/z 471.1434 [M + H]+ (calcd for C28H23O7+ 471.1438 Δ = −0.85 ppm), MS/MS spectrum: CCMSLIB00012474988 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012474988), m/z 469.1291 [M − H]− (calcd for C28H21O7− 469.1293, Δ = −0.43 ppm).
(+)-Gnetupendin C (2) [α]20D +69.6 (c 0.16, MeOH); UV (MeOH) λmax (logε) 226 (4.57), 287 (4.19), 310 (4.32), 328 (4.34), 347 (4.06) nm; 1H NMR (DMSO, 600 MHz) δ 9.54 (1H, s), 9.37 (1H, s), 9.24 (1H, s), 8.97 (1H, s), 8.71 (2H, s), 7.41 (2H, d, J = 8.7 Hz), 7.03 (1H, d, J = 16.3 Hz), 6.91 (1H, d, J = 16.2 Hz), 6.79–6.73 (3H, dd, J = 11.2, 8.5 Hz), 6.70 (1H, s), 6.49 (1H, d, J = 1.3 Hz), 6.16 (1H, d, J = 2.3 Hz), 5.94 (1H, dd, J = 8.3, 2.3 Hz), 5.84 (1H, d, J = 7.7 Hz), 5.78 (1H, t, J = 2.2 Hz), 5.63 (2H, d, J = 2.2 Hz), 4.45 (1H, d, J = 7.7 Hz); 13C NMR (DMSO, 151 MHz) δ 161.3, 157.2, 157.1, 156.9, 154.4, 154.2, 141.6, 138.9, 128.2, 127.9, 127.9, 127.4, 125.7, 116.8, 115.5, 114.5, 107.3, 106.8, 105.4, 101.6, 100.5, 98.1, 84.8, 48.6 (NP-MRD ID: NP0332862 (https://np-mrd.org/natural_products/NP0332862)); HRESIMS m/z 471.1433 [M + H]+ (calcd for C28H23O7+ 471.1438, Δ = −1.06 ppm), MS/MS spectrum: CCMSLIB00012474991 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012474991), m/z 469.1289 [M − H]− (calcd for C28H21O7− 469.1293, Δ = −0.85 ppm).
Gnetin C (3) [α]20D −4.61 (c 0.09, MeOH); UV (MeOH) λmax (logε) 226 (4.23), 287 (3.74), 310 (3.86), 328 (3.88), 347 (3.61) nm; 1H NMR (DMSO, 600 MHz) δ 9.48 (1H, s), 9.31 (1H, s), 9.11 (2H, s), 8.22 (1H, s), 7.41 (2H, d, J = 8.6 Hz), 7.11 (2H, d, J = 8.6 Hz), 7.02 (1H, d, J = 16.3 Hz), 6.91 (1H, d, J = 16.3 Hz), 6.76 (4H, t, J = 8.3 Hz), 6.65 (1H, s), 6.48 (1H, s), 6.05 (1H, s), 5.99 (2H, d, J = 2.2 Hz), 5.31 (1H, d, J = 4.5 Hz), 4.23 (1H, d, J = 4.5 Hz); 13C NMR (DMSO, 151 MHz) δ 161.5, 158.4, 157.3, 157.2, 154.6, 145.0, 139.6, 132.2, 128.1, 128.1, 127.9, 126.8, 125.5, 115.5, 115.3, 114.0, 107.2, 105.3, 100.9, 97.6, 92.0, 54.3 (NP-MRD ID: NP0061298 (https://www.npmrd.org/natural_products/NP0061298)); HRESIMS m/z 455.1487 [M + H]+ (calcd for C28H23O6+ 455.1489, Δ = −0.44 ppm), MS/MS spectrum: CCMSLIB00012475006 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012475006), m/z 453.1338 [M − H]− (calcd for C28H21O6− 453.1344, Δ = −1.32 ppm).
(−)-Gnetoline A (4) [α]20D +19.7 (c 0.17, MeOH); UV (MeOH) λmax (logε) 226 (4.47), 287 (3.94), 310 (4.02), 328 (4.05), 347 (3.80) nm; 1H NMR (DMSO, 600 MHz) δ 9.55 (2H, s), 9.40 (1H, s), 9.33 (1H, s), 9.18 (1H, s), 8.95 (1H, s), 8.72 (2H, s), 7.41 (2H, d, J = 8.9 Hz), 7.16 (2H, d, J = 8.7 Hz), 7.03 (1H, d, J = 15.5 Hz), 6.92 (1H, d, J = 16.3 Hz), 6.78–6.73 (5H, m), 6.68 (1H, s), 6.50 (2H, s), 6.19 (1H, s), 6.15 (1H, d, J = 1.4 Hz), 5.93 (1H, dd, J = 8.4, 2.3 Hz), 5.82 (1H, d, J = 7.9 Hz), 5.78 (1H, t, J = 2.2 Hz), 5.63 (2H, d, J = 2.2 Hz), 5.40 (1H, d, J = 3.9 Hz), 4.42 (1H, d, J = 7.8 Hz), 4.35 (1H, d, J = 3.8 Hz); 13C NMR (DMSO, 151 MHz) δ 161.53, 160.91, 157.26, 157.21, 157.06, 156.85, 154.61, 154.31, 154.20, 144.46, 141.61, 139.74, 132.26, 128.17, 128.09, 127.90, 127.35, 126.72, 125.50, 115.80, 115.53, 115.31, 114.55, 113.90, 107.27, 106.84, 105.56, 101.59, 100.53, 99.72, 97.71, 91.94, 84.98, 54.23, 48.52 (NP-MRD ID: NP0332863 (https://np-mrd.org/natural_products/NP0332863)); HRESIMS m/z 697.2068 [M + H]+ (calcd for C42H33O10+ 697.2068, Δ = 0 ppm), MS/MS spectrum: CCMSLIB00012475011 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012475011), m/z 695.1917 [M − H]− (calcd for C42H31O10− 695.1923, Δ = −0.86 ppm).
(−)-Latifolol (5) [α]20D −12.4 (c 0.12, MeOH); UV (MeOH) λmax (logε) 226 (4.43), 287 (3.93), 310 (4.04), 328 (4.07), 347 (3.80) nm; 1H NMR (DMSO, 600 MHz) δ 9.57 (1H, s), 9.49 (2H, s), 9.40 (1H, s), 9.27 (1H, s), 9.16 (1H, s), 9.06 (2H, s), 7.42 (2H, d, J = 8.6 Hz), 7.16 (2H, d, J = 8.1 Hz), 7.04 (1H, d, J = 16.2 Hz), 6.92 (1H, d, J = 16.2 Hz), 6.87 (1H, d, J = 8.3 Hz), 6.77 (4H, d, J = 8.2 Hz), 6.68 (1H, s), 6.52 (1H, s), 6.31 (1H, m), 6.17 (2H, m), 6.11 (1H, s), 6.03 (3H, d, J = 1.1 Hz), 5.54 (1H, d, J = 4.4 Hz), 5.39 (1H, d, J = 4.0 Hz), 4.34 (1H, d, J = 4.0 Hz), 4.23 (1H, d, J = 4.5 Hz); 13C NMR (DMSO, 151 MHz) δ 161.5, 161.3, 158.1, 158.0, 157.3, 157.3, 155.6, 154.6, 154.5, 145.6, 144.8, 139.8, 132.2, 128.2, 128.2, 127.9, 127.1, 126.8, 125.5, 118.3, 115.6, 115.3, 113.9, 113.8, 107.3, 106.9, 106.1, 105.6, 102.5, 100.7, 99.5, 97.8, 92.0, 87.9, 54.3, 52.7, 48.6 (NP-MRD ID: NP0028328 (https://np-mrd.org/natural_products/NP0028328)); HRESIMS m/z 697.2069 [M + H]+ (calcd for C42H33O10+ 697.2068, Δ = 0.14 ppm), MS/MS spectrum: CCMSLIB00012475005 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012475005), m/z 695.1917 [M − H]− (calcd for C42H31O10− 695.1923, Δ = −0.86 ppm).
(−)-Gnetin E (6) [α]20D −2.06 (c 0.13, MeOH); UV (MeOH) λmax (logε) 226 (4.45), 287 (3.91), 310 (4.03), 328 (4.04), 347 (3.78) nm; 1H NMR (600 MHz, DMSO) δ 9.56 (1H, s), 9.49 (2H, s), 9.40 (1H, s), 9.23 (1H, s), 9.10 (2H, s), 7.41 (2H, d, J = 8.7 Hz), 7.15 (2H, d, J = 8.6 Hz), 7.10 (2H, d, J = 8.5 Hz), 7.03 (1H, d, J = 16.3 Hz), 6.91 (1H, d, J = 16.3 Hz), 6.75 (6H, t, J = 8.7 Hz), 6.68 (1H, s), 6.50 (1H, s), 6.16 (1H, s), 6.13 (1H, s), 6.03 (1H, t, J = 2.2 Hz), 5.97 (2H, d, J = 2.2 Hz), 5.40 (1H, d, J = 4.0 Hz), 5.27 (1H, d, J = 5.7 Hz), 4.34 (1H, d, J = 3.9 Hz), 4.22 (1H, d, J = 5.7 Hz); 13C NMR (151 MHz, DMSO) δ 161.5, 161.0, 158.3, 157.3, 157.2, 154.6, 154.5, 145.1, 144.7, 139.7, 132.2, 131.7, 128.2, 128.1, 127.9, 127.2, 126.8, 125.5, 115.5, 115.3, 115.3, 113.9, 113.1, 107.3, 107.1, 105.4, 100.9, 99.4, 97.8, 92.4, 91.9, 54.3, 54.2 (NP-MRD ID: NP0332864 (https://www.npmrd.org/natural_products/NP0332864)); HRESIMS m/z 681.2120 [M + H]+ (calcd for C42H33O9+ 681.2119, Δ = 0.15 ppm), MS/MS spectrum: CCMSLIB00012474987 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012474987), m/z 679.1970 [M − H]− (calcd for C42H31O9− 679.1974, Δ = −0.59 ppm).
Macrostachyol A (7) [α]20D −3.87 (c 0.12, MeOH); UV (MeOH) λmax (logε) 226 (4.59), 287 (4.01), 310 (4.07), 328 (4.10), 347 (3.84) nm; 1H NMR (DMSO, 600 MHz) δ 9.56 (2H, s), 9.49 (2H, s), 9.40 (1H, s), 9.31 (1H, s), 9.26 (1H, s), 9.16 (1H, s), 9.04 (2H, s), 7.41 (2H, d, J = 8.7 Hz), 7.16 (2H, d, J = 8.7 Hz), 7.13 (2H, d, J = 8.6 Hz), 7.03 (1H, d, J = 16.3 Hz), 6.91 (1H, d, J = 16.2 Hz), 6.85 (1H, d, J = 8.4 Hz), 6.80–6.72 (6H, m), 6.68 (1H, m), 6.50 (1H, m), 6.31 (1H, d, J = 2.3 Hz), 6.16 (3H, ddd, J = 8.0, 6.5, 1.8 Hz), 6.10 (2H, dd, J = 11.9, 1.2 Hz), 6.01 (3H, s), 5.51 (1H, d, J = 4.6 Hz), 5.41 (1H, d, J = 4.0 Hz), 5.35 (1H, d, J = 5.1 Hz), 4.35 (1H, d, J = 4.0 Hz), 4.32 (1H, d, J = 5.1 Hz), 4.23 (1H, d, J = 4.7 Hz); 13C NMR (DMSO, 151 MHz) δ 161.5, 161.1, 161.0, 158.1, 158.0, 157.3, 157.2, 155.6, 154.6, 154.5, 154.5, 145.5, 145.2, 144.5, 139.8, 132.2, 131.8, 128.2, 128.1, 127.9, 127.2, 127.1, 126.8, 125.5, 118.2, 115.5, 115.3, 115.3, 113.9, 113.7, 113.0, 107.3, 107.2, 106.9, 106.0, 105.6, 102.5, 100.7, 99.7, 99.6, 97.8, 92.3, 91.9, 87.8, 54.2, 52.6 (NP-MRD ID: NP0332865 (https://np-mrd.org/natural_products/NP0332865)); HRESIMS m/z 923.2700 [M + H]+ (calcd for C56H43O13+ 923.2698, Δ = 0.22 ppm), MS/MS spectrum: CCMSLIB00012475007 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012475007), m/z 921.2527 [M − H]− (calcd for C56H41O13− 921.2553, Δ = −2.82 ppm).
Gnemonol B (8) [α]20D −6.18 (c 0.10, MeOH); UV (MeOH) λmax (logε) 226 (4.58), 287 (4.04), 310 (4.09), 328 (4.10), 347 (3.90) nm; 1H NMR (DMSO, 600 MHz) δ 9.56 (2H, s), 9.48 (2H, s), 9.39 (1H, s), 9.31 (1H, s), 9.23 (1H, s), 9.09 (2H, s), 7.41 (2H, d, J = 8.7 Hz), 7.16–7.12 (4H, dd, J = 11.8, 8.7 Hz), 7.09 (2H, d, J = 8.6 Hz), 7.03 (1H, d, J = 16.3 Hz), 6.91 (1H, d, J = 16.3 Hz), 6.77–6.74 (8H, dd, J = 8.7, 2.8 Hz), 6.67 (1H, d, J = 1.2 Hz), 6.50 (1H, d, J = 1.3 Hz), 6.19–6.10 (4H, m), 6.02 (2H, t, J = 2.1 Hz), 5.96 (1H, d, J = 2.2 Hz), 5.41 (1H, d, J = 4.1 Hz), 5.37 (1H, d, J = 5.0 Hz), 5.25 (1H, d, J = 5.9 Hz), 4.33 (2H, dd, J = 9.5, 4.5 Hz), 4.22 (1H, d, J = 5.9 Hz); 13C NMR (DMSO, 151 MHz) δ 161.5, 161.0, 160.9, 158.3, 158.1, 157.3, 157.2, 154.6, 154.5, 154.5, 145.2, 144.9, 139.8, 132.2, 131.8, 131.6, 128.2, 128.1, 127.9, 127.3, 127.1, 126.8, 125.5, 115.5, 115.3, 115.3, 113.9, 113.1, 113.0, 107.3, 107.2, 107.2, 105.4, 100.8, 99.6, 97.7, 92.4, 92.2, 91.9, 54.4, 54.2, 54.2 (NP-MRD ID: NP0140175 (https://www.npmrd.org/natural_products/NP0140175)); HRESIMS m/z 907.2742 [M + H]+ (calcd for C56H43O12+ 907.2749, Δ = −0.77 ppm), MS/MS spectrum: CCMSLIB00012475004 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012475004), m/z 905.2579 [M − H]− (calcd for C56H41O12− 905.2604, Δ = −2.76 ppm).
(−)-Gnemontanin G (9) [α]20D −16.8 (c 0.06, MeOH); UV (MeOH) λmax (logε) 226 (4.09), 287 (3.29), 310 (2.84), 328 (2.94), 347 (2.51) nm; 1H NMR (DMSO, 600 MHz) δ 9.44 (1H, s), 9.19 (1H, s), 8.97 (2H, s), 8.85 (1H, s), 8.62 (1H, s), 7.10 (2H, d, J = 8.6 Hz), 6.72 (2H, d, J = 8.6 Hz), 6.57 (1H, d, J = 8.5 Hz), 6.22 (1H, dd, J = 8.4, 2.4 Hz), 6.17 (1H, d, J = 2.4 Hz), 6.04–6.01 (3H, m), 5.99 (1H, d, J = 2.1 Hz), 5.53 (1H, d, J = 2.1 Hz), 4.72 (1H, d, J = 8.4 Hz), 3.95 (1H, d, J = 7.1 Hz), 3.54–3.49 (1H, m), 3.27 (1H, t, J = 7.1 Hz); 13C NMR (DMSO, 151 MHz) δ 157.9, 157.1, 157.0, 156.5, 154.7, 154.2, 147.0, 144.8, 130.1, 129.7, 129.0, 121.3, 115.4, 114.8, 108.4, 106.2, 104.3, 103.4, 102.6, 101.8, 100.3, 77.3, 56.1, 55.6, 48.4, 47.0 (NP-MRD ID: NP0332866 (https://np-mrd.org/natural_products/NP0332866)); HRESIMS m/z 471.1432 [M + H]+ (calcd for C28H23O7+ 471.1438, Δ = −1.27 ppm), MS/MS spectrum: CCMSLIB00012474989 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012474989), m/z 469.1289 [M − H]− (calcd for C28H21O7− 469.1293, Δ = −0.85 ppm).
(−)-Gnetumontanin A (10) [α]20D −13.4 (c 0.05, MeOH); UV (MeOH) λmax (logε) 226 (3.88), 287 (3.38), 305 (3.41), 333 (4.57), 342 (3.54) nm; 1H NMR (DMSO, 600 MHz) δ 9.59 (1H, d, J = 8.0 Hz), 9.54 (1H, d, J = 7.2 Hz), 9.40 (1H, s), 9.24 (1H, s), 9.21 (1H, d, J = 5.2 Hz), 9.04 (2H, s), 7.35 (1H, d, J = 8.5 Hz), 7.19 (1H, d, J = 16.4 Hz), 6.85 (1H, d, J = 16.4 Hz), 6.84 (1H, d, J = 8.3 Hz), 6.54 (1H, s), 6.44 (1H, d, J = 1.3 Hz), 6.32 (2H, t, J = 2.7 Hz), 6.25 (1H, dd, J = 8.5, 2.4 Hz), 6.14 (1H, dd, J = 8.3, 2.4 Hz), 6.04 (2H, d, J = 2.2 Hz), 6.02 (1H, t, J = 2.3 Hz), 5.53 (1H, d, J = 3.4 Hz), 4.18 (1H, d, J = 3.4 Hz); 13C NMR (DMSO, 151 MHz) δ 161.7, 158.1, 157.9, 156.1, 155.3, 154.6, 145.7, 140.1, 127.2, 126.5, 124.7, 123.3, 118.7, 115.3, 114.3, 107.2, 106.1, 105.9, 105.5, 102.6, 102.5, 100.7, 97.8, 87.6, 52.8 (NP-MRD ID: NP0332867 (https://np-mrd.org/natural_products/NP0332867)); HRESIMS m/z 487.1382 [M + H]+ (calcd for C28H23O8+ 487.1387, Δ = −1.03 ppm), MS/MS spectrum: CCMSLIB00012474986 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012474986), m/z 485.1240 [M − H]− (calcd for C28H21O8− 485.1242, Δ = −0.41 ppm).
(−)-Gnetuhainin M (11) [α]20D −44.9 (c 0.12, MeOH); UV (MeOH) λmax (logε) 226 (4.63), 287 (4.11), 310 (4.16), 328 (4.19), 347 (3.90) nm; 1H NMR (DMSO, 600 MHz) δ 9.57 (1H, s), 9.53 (1H, s), 9.37 (2H, s), 9.17 (1H, s), 9.15 (2H, s), 9.07 (2H, s), 7.09 (2H, d, J = 8.7 Hz), 6.83 (1H, d, J = 16.3 Hz), 6.68 (2H, d, J = 8.6 Hz), 6.59 (1H, d, J = 2.2 Hz), 6.57 (1H, d, J = 16.3 Hz), 6.55 (1H, d, J = 8.4 Hz), 6.33–6.29 (3H, m), 6.23 (1H, d, J = 2.0 Hz), 6.10 (2H, d, J = 2.2 Hz), 6.09–6.05 (4H, m), 6.04 (1H, t, J = 2.1 Hz), 5.61 (1H, d, J = 6.6 Hz), 5.54 (1H, d, J = 2.6 Hz), 4.65 (1H, d, J = 6.6 Hz), 4.10 (1H, d, J = 2.6 Hz); 13C NMR (DMSO, 151 MHz) δ 160.8, 158.9, 158.6, 158.4, 158.1, 157.6, 157.3, 154.9, 154.8, 145.6, 145.5, 134.7, 128.8, 128.0, 127.7, 127.6, 125.8, 122.0, 118.9, 118.6, 115.5, 115.1, 112.8, 108.6, 105.9, 105.7, 105.6, 102.9, 102.4, 101.1, 100.7, 96.0, 88.2, 88.2, 53.9, 53.3 (NP-MRD ID: NP0332868 (https://np-mrd.org/natural_products/NP0332868)); HRESIMS m/z 713.2018 [M + H]+ (calcd for C42H33O11+ 713.2017, Δ = 0.14 ppm), MS/MS spectrum: CCMSLIB00012474996 (https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00012474996), m/z 711.1870 [M − H]− (calcd for C42H31O11− 711.1872, Δ = −0.28 ppm).
Footnote |
† Electronic supplementary information (ESI) available: Figures, NMR data, ECD spectra and HRMS data for all isolated compounds. See DOI: https://doi.org/10.1039/d4ra08421g |
This journal is © The Royal Society of Chemistry 2025 |