Metabolic Profiling of Conyza sumatrensis (Retz.) E. Walker from Lugazi, Uganda

Abstract

Conyza sumatrensis is a plant of the Asteraceae family widespread in the tropical and subtropical regions of all continents. The plant is applied in folk medicine to treat malaria and helminth infections as well as other diseases. In Uganda, for example, the plant is traditionally used against ectoparasites and for wound healing. In this work, we describe a comprehensive analytical approach to investigate the metabolic profile of C. sumatrensis supported by database-assisted annotation and in silico techniques. The study aimed to analyze the metabolic profile of C. sumatrensis using multiple analytical techniques due to the complexity of the plant composition. Therefore, we employed a combination of thin-layer chromatography (TLC), high-performance liquid chromatography (HPLC), gas chromatography-mass spectrometry (GC-MS), and liquid chromatography-tandem mass spectrometry (LC-MS/MS). From the GC-MS experiments, more than 40 compounds could be annotated that had not been found in C. sumatrensis before. A number of these substances are known to possess relevant bioactivities, including antibacterial properties. Indeed, ethanolic extracts demonstrated antimicrobial activity against representative strains of both Gram-positive and Gram-negative bacteria, as shown by agar diffusion and microdilution assays. Using LC-MS/MS data, a feature-based molecular network was constructed to facilitate the comparison of two extraction solvents: water and ethanol. The majority of the features were detected in both of the extracts. However, some features were only detected using one of the extraction solvents. Our work provided valuable insights into the chemical profile of C. sumatrensis and lays the foundation for future research into its medicinal properties.

1. Introduction

Conyza sumatrensis (Retz.) E. Walker is a dicotyledonous herb of the Asteraceae family commonly called Broadleaf Fleabane or Sumatran Fleabane [1] and in many regions of Uganda, “kafumbe” [2]. It spreads invasively worldwide and can be found across Africa [3]. Morphological characteristics include an erect hairy stem, up to 120 cm tall, with spreading and serrated leaves [4]. The plant is distinguished from other species, e.g., C. bonariensis and C. canadensis, by its annual growth, small disciform capitula that lack fully developed ray florets, and narrow lanceolate leaves, which are alternately arranged along the stem [5]. C. sumatrensis is able to develop resistance against popular herbicides, such as glyphosate and chlorimuron-ethyl, revealing a high degree of ecological adaptability [6].

Different traditional medicinal applications of C. sumatrensis have been reported from West and Central Africa, including its use as an antiplasmodial [7] and antihelmintic remedy [8], as well as for its antimicrobial and antioxidant activities in general [1]. In Uganda, the plant is applied by local healers to treat ringworm infections, heal wounds, and control ectoparasites [2]. The herbal medicines are primarily prepared as hot water extracts, decoctions, or steam baths [9]. Essential oils from different parts of the plant have been found to exhibit distinct bioactivities. For example, the leaf oil shows significant antibacterial effects against several pathogens determined with the broth-microdilution method [e.g., Enterococcus faecalis (MIC: >500 µg/mL), Staphylococcus aureus (MIC: 15.62 µg/mL and Proteus mirabilis (MIC: 62.50 µg/mL)]. Antifungal activity was observed with the percentages of mycelia-growth inhibition (GI). The leaf oil showed 80% GI for Candida albicans and 69% for Aspergillus fumigatus and other filamentous fungi between 15.8 and 34.9% at a concentration of 500 µL/mL [10]. Besides antimicrobial effects, various extracts demonstrate antioxidant properties, including radical scavenging activity of the methanolic extract of C. sumatrensis [1]. Previous studies have mostly described the presence or absence of certain natural product classes [1], but only a few specific compounds were identified. A preliminary phytochemical analysis by Jack et al., 2008 [4] revealed tannins, saponins, flavonoids, steroids, and glycosides in steam-distilled fresh leaves of the plant. An essential oil composition analysis identified limonene, β-pinene, and (E)-β-caryophyllene as major compounds [10]. GC-MS analyses of hexane, ethyl acetate, and methanolic extracts of the leaves of C. sumatrensis found spathulenol, (cis)-β-farnesene, cis-pinane, and E-phytol as the most prominent components [1]. High-performance liquid chromatography (HPLC) fingerprinting of aqueous and methanolic extracts from the leaves performed by Boniface et al. revealed 11 and 28 characteristic peaks, respectively, which were not further determined [7]. Even though C. sumatrensis is widely used as a therapeutic plant, significant gaps remain in our knowledge regarding its phytochemistry.

Here, we report on the detailed chemical profile of C. sumatrensis, using a broad range of analytical approaches, including TLC, HPLC, GC-MS, and LC-MS/MS. Global Natural Products Social Molecular Networking (GNPS) [11] and in silico annotation tools were employed to visualize and annotate non-targeted MS data. Furthermore, antimicrobial activity of the extracts was investigated against a gram-positive and a gram-negative test organism.

2. Materials and Methods

2.1. Plant Collection, Identification, and Drying

Fresh leaves of C. sumatrensis were collected from Namagunga village, Lugazi district, Uganda (GPS Coordinates: 0°22′31.05″ N, 32°53′3.45″ E). To confirm its identity, the prepared voucher specimens were authenticated at the Department of Pharmacy, Mbarara University of Science and Technology. The leaves were prepared and assigned the collection number IK-2023-001. The plant was further compared with online databases such as the International Plant Name Index (www.ipni.org, accessed on 4 April 2023) and the Kew Royal Botanic Garden (www.theplantlist.org, accessed on 4 April 2023). The leaf samples were air-dried in the shade for two weeks at room temperature and then submitted for grinding and further analysis.

2.2. Extraction of the Powdered Leaf Samples

The air-dried leaves were ground into coarse powder, and two different solvents were used for the extraction. For the preparation of aqueous extracts, 35 g of powdered plant material was used, and 350 mL of water was added and heated at 100 °C for 15 min. After cooling to room temperature, the filtered supernatant was subjected to lyophilization. For the preparation of ethanolic extracts, 350 mL of 70% (v/v) ethanol was used for 35 g of powdered plant material. The extraction was conducted at room temperature and with agitation every 6 h. After 48 h, the mixture was filtered and concentrated in a vacuum evaporator at 35 °C and 50 mbar. Then, the filtrate was dried entirely in a vacuum centrifuge.

2.3. Antimicrobial Testing with Agar Diffusion and Microdilution Assays

We conducted antimicrobial assays to evaluate the efficacy of plant extracts against three distinct microorganisms (Escherichia coli DSM498, Bacillus subtilis DSM109511, and Pseudomonas fluorescens DSM289), which were chosen for their diverse characteristics and relevance. B. subtilis was used as an indicator for activity against Gram-positive bacteria. E. coli and P. fluorescens are indicators for activity against Gram-negative bacteria. The genus Pseudomonas is known for its nonsusceptibility against a variety of antibiotics and is thus of particular clinical interest. The use of these strains allows a comprehensive evaluation of the antimicrobial activity of the extracts against both Gram-positive and Gram-negative bacteria, as they exhibit different properties. For the antimicrobial activity testing in the agar diffusion test, 50 µL of an overnight culture was spread out on Luria-Bertani (LB) medium (10 g/L tryptone, 5 g/L yeast extract, 5 g/L NaCl) or nutrient medium (5 g/L peptone, 3 g/L meat extract, 15 g/L agar). The dry extracts were dissolved in the respective solvent (water or 70% (v/v) ethanol) in a concentration of 50 mg/mL, and 10 µL were applied to filter paper discs (5 mm) and placed on top of the agar medium. Gentamycin (for E. coli and P. fluorescens) or erythromycin (for B. subtilis) at 25 µg/mL was used as a positive control and purified water was used as a negative control. The assay was incubated at 37 °C (E. coli, B. subtilis) or 30 °C (P. fluorescens) for 24 h. Then, the inhibition zones were measured and recorded. To evaluate the minimal inhibition concentration (MIC), the bacteria were grown in a 5 mL preculture overnight at 30 °C for P. fluorescens and 37 °C for the other two strains. Then, the bacteria were diluted with media to approx. 5 × 10⁵ colony-forming units (CFU) per milliliter. The MICs were measured in microtiter plates in a final volume of 200 µL and final extract concentrations of 5, 2.5, 1.25, 0.625, 0.313, and 0.156 mg/mL dissolved in the respective solvent (water or 70% (v/v) ethanol) incubated for 14 h with shaking. To monitor the growth of the bacteria, the optical density at 600 nm was measured every 15 min. P. fluorescens was cultivated in glass tubes with a final volume of 2 mL and final extract concentrations from 2.5 to 0.078 mg/mL at 30 °C for 2 days.

2.4. HPLC-UV Fingerprints of Ethanolic and Aqueous Extracts

For HPLC-UV measurements, an Agilent InfinityLab LC System 1260 Infinity II Prime LC (Agilent, Santa Clara, CA, USA) was used with a module lineup comprising a G7104C 1260 Flexible pump, G7129C 1260 vial sampler, G7116A 1260, column oven, G1364F 1260 fraction collector, G1330B 1260 thermostatic unit, and a G7115A 1260 DAD WR UV detector. Solvent system A was 0.1% formic acid and solvent system B was acetonitrile with 0.1% formic acid, with a flow rate of 1 mL/min. The program gradient started with 90% solvent A for 5 min, which decreased to 20% A in 20 min, was held for 3 min, and then increased to 90% A and held at 90% A for 2 min. The HPLC system was equipped with an EC 250/4 Nucleosil 100-5 C18 analytical column (Macherey-Nagel, Düren, Germany). The dried extracts were dissolved in water:acetonitrile (90:10) at a concentration of 20 mg/mL. The injection volume was 20 µL of the extract, the column oven was at 25 °C, and the detection wavelengths were 205, 220, 250, and 310 nm to detect the widest possible range of different UV-active substances.

2.5. GC-MS Headspace Analysis

For headspace analysis, a QP2010 GC-MS with an HS-20 headspace sampler (Shimadzu, Kyoto, Japan) was used, and 50 mg of dried pulverized raw plant material from leaves, placed into 20 mL GC-MS vials (N = 3), was measured. After an equilibration for 8 min at 70 °C, 1 mL of headspace volume was drawn at a pressurizing gas pressure of 50 kPa. All experiments were run in trap mode; compounds from the headspace sample were adsorbed to a Tenax^® trap cooled to 5 °C, which was then dried for 3 min at 31 kPa and subsequently desorbed at 300 °C for injection at a split ratio of 5. Helium (Alphagaz, Air Liquide, Düsseldorf, Germany) was used as carrier gas at a flow of 0.95 mL/min. The GC program was as follows: starting at 35 °C for two minutes, then increasing the temperature by 5 °C/min until 150 °C and 15 °C/min until 300 °C. After electron ionization at 230 °C and 70 eV, the ions were detected using current tune parameters. The data were analyzed using the LabSolutions program GCMSsolution version 4.20 (Shimadzu, Kyoto, Japan). The chromatograms were processed with the automated peak-finding algorithm of the “Post run” module at a slope of 600/min. An automated similarity search against the commercial mass spectral library NIST14 (National Institute of Standards and Technology, Gaithersburg, MD, USA) was run for tentative identification of the peaks. The peak area was obtained from a representative file. The indicated relative peak areas refer to the sum of all peak areas over the entire chromatogram.

2.6. TLC Analysis

The TLC was conducted with the same extracts of C. sumatrensis as mentioned in Section 2.3 in a concentration of 1 mg/mL. Five microliters of the extract were applied on the start line of the TLC plate. Chloroform: methanol, 8:2 (v/v), was used as the mobile phase and p-anisaldehyde/H₂SO₄ was utilized as the detection agent. For the preparation of the dyeing reagent, 0.5 mL p-anisaldehyde was mixed with 50 mL glacial acetic acid and 1 mL concentrated sulfuric acid. After the run, the TLC plate was completely immersed in the staining reagent and heated in the drying oven until the spots appeared, and the plates were cooled to room temperature. The spots were then visually detected under visible and UV light at 254 and 366 nm.

2.7. LC-MS/MS Instrument Setting

For LC-MS/MS measurements, an Ultimate 3000 UHPLC from ThermoFisher Scientific (Waltham, MA, USA) coupled to an Impact II QTOF mass spectrometer from Bruker Daltonics (Billerica, MA, USA) was used for analysis of the same extracts of C. sumatrensis as mentioned in Section 2.3 in positive ion mode with automated collision-induced dissociation (DDA). Five microliters of the cooled extracts (10 °C) were injected and separated at a flow rate of 0.6 mL/min on a YMC-Triart C18 100 × 3.0 mm S-5 µm 12 nm column (YMC Europe GmbH, Dinslaken, Germany) kept at 30 °C. The gradient started with 15% solvent A (0.1% formic acid in acetonitrile) and increased to 80% solvent A in 20 min, increasing to 100% A for the next 2 min and was then held for 12 min; solvent B was 0.1% formic acid. Nitrogen was used as a nebulizer (3.0 bar) and dry gas (8.0 L/min, 250 °C). The capillary was set to 3700 V and the charging voltage to 2000 V (focus not active), with a scan range from m/z 100 to 1000 and DAD detection from 190 to 500 nm. MS/MS analysis was in auto mode with three precursors and active exclusion after five spectra, with a release after 0.3 min. Compass 1.8 (Hystar 3.2 and DataAnalysis 4.2 from Bruker, Billerica, MA, USA) was used for spectra visualization.

2.8. Construction of a Feature-Based Molecular Network

For the generation of the feature-based molecular network, the data were processed as suggested by Heuckeroth et al. (2024) [12], and the network was created as recommended in Nothias et al. (2020) [13]. LC-MS/MS data were centroided and converted to an open format (.mzML) using the peak picking algorithm of MSConvert [14] (version 3.0.19) from ProteoWizard (https://proteowizard.sourceforge.io, accessed on 21 May 2024). Subsequent recalibration was performed using the m/z value corrector from github (https://github.com/, accessed on 21 May 2024). The recalibrated raw data files were further processed using MZmine 4.0 [15], and the feature detection process comprised the following steps:

Import MS-data, Import spectral library, Project metadata import, Mass detection (MS1), Mass detection (MS2), ADAP chromatogram builder, Local minimum feature resolver, 13C isotope filter, Isotopic peaks finder, Join aligner, Peak finder (multithreaded), Feature list row filter, Feature list blank subtraction, Duplicate peak filter, Correlation grouping (metaCorrelate), Spectral library search, Ion identity networking.

The noise was assessed to correspond to 2.0 × 10³ for MS1 and 3.5 × 10² for MS2, removing all data points below that threshold, followed by executing the ADAP chromatogram builder with a minimum group size of three. The local minimum feature resolver was set to a chromatographic threshold of 30%. The 13C isotope filter and isotopic peaks finder considered a mass tolerance of 0.003 m/z or 3 ppm, and the join aligner features with a sample-to-sample m/z tolerance of 0.008 or 8 ppm and a retention time tolerance of 0.4 min were combined in the aligned feature list. With the multithreaded peak finder, missing peaks were complemented with a m/z tolerance of 0.01 or 20 ppm. With the feature list rows filter function, all the features that were not present in at least three samples were filtered out. With the blank filter, features present in the blank samples were removed, and only signals threefold as high as in the blank sample were considered. With the duplicate peak filter, peaks with a m/z tolerance of 0.0015 or 1 ppm were duplicated and removed. In the correlation grouping, peaks with 60% intensity overlap and a retention time overlap of 0.06 min were grouped. A spectral library search was conducted with an imported library from MassBank of North America (MoNA, https://mona.fiehnlab.ucdavis.edu/, accessed 19 November 2024). The used databases were Fiehn HILIC, GNPS, natural product library, LipidBlast2022, MassBank, ReSpect, TOF alkaloids, and Fiehn Natural Products. The ion identity network was performed with a m/z tolerance of 0.003 or 3 ppm. Then, the feature quantification table (.csv), spectral information (.mgf), and files for statistical evaluation with MetaboAnalyst (https://www.metaboanalyst.ca, accessed on 7 November 2024) were exported with the molecular networking (for GNPS) module, with the SIRIUS export module and the export for statistics module. The .csv, .mgf file, and the metadata table were uploaded to the GNPS platform [11] (accessible under: https://gnps.ucsd.edu, accessed on 4 June 2024) using WinSCP (version 6.3.3 from Martin Přikryl, https://winscp.net, accessed on 3 June 2024) as an FTP client. Networks of the features obtained from the processed data were then generated using the Feature-Based-Molecular Network (FBMN) workflow [13] on the GNPS platform [11]. The precursor ion mass tolerance and the MS/MS fragment ion tolerance were set to 0.07 Da. The molecular networking was generated with a minimum cosine score of 0.6 and more than four matched peaks for the connecting nodes, or if the nodes share the respective top 10 most similar nodes. The maximum size of a molecular family was 100, and a GNPS library search was used for spectra annotation [11]. The library spectra were filtered as the input data. All matches between the network spectra and library spectra were required to have a score above 0.7 and at least six matched peaks. Then, the MS/MS spectra were annotated with DEREPLICATOR [16]. The GNPS workflow output of the network is accessible with the ID c5615bc3bccb4c148479f8fb6250b25d. The generated network was visualized using Cytoscape (version 3.10.2) [17]. Compound identification and annotation were based on the molecular masses and MS/MS spectra via the GNPS databases, as well as comparison with the relevant literature. For compound class prediction, SIRIUS software (https://boecker-lab.github.io/docs.sirius.github.io/, accessed on 29 October 2024) [18,19] (version 5.8.5) was used with the following libraries: Bio Database, COCONUT (https://coconut.naturalproducts.net/, accessed on 29 October 2024)), plantcyc (https://www.plantcyc.org/, accessed on 29 October 2024)), KNApSAcK (https://www.knapsackfamily.com/KNApSAcK/, accessed on 29 October 2024)), and Natural products https://bioinf-applied.charite.de, accessed on 29 October 2024)). CANOPUS [19] class prediction was mapped on the network based on the ClassyFire prediction [20]. ClassyFire classification was compared with the natural product classifier (NPC) [21].

2.9. MS2LDA Analysis

MS2LDA analysis was conducted within the GNPS workflow. It employs the latent Dirichlet allocation (LDA) algorithm to identify motifs (Mass2Motifs) of interconnected fragments and neutral losses within MS2 spectra. It accepts outputs from FBMN, facilitating a direct link between MS2LDA annotations and molecular networks, and can be executed on the GNPS web platform (GNPSDocumentation/ms2lda/) or through the MS2LDA web application [22]. The MS2LDA results can be accessed under the task ID 13aa01d2295a4db281c4eb14ea7d73af.

2.10. Visualization

The Venn diagram was generated with R and Rstudio (version 2023.12.1, build 402) with the package “VennDiagram” (version 1.7.3) from the feature list generated with MZmine. Features were considered present if detected in at least two replicates above the intensity threshold according to feature detection parameters set in MZmine. The hierarchical visualization of the compound class prediction was performed with the online data visualization tool Flourish^® (https://flourish.studio, accessed 17 July 2024, 2024, Canva UK Operations Ltd., London, UK).

3. Results and Discussion

3.1. The Extraction of C. sumatrensis Plants with Water and Ethanol Results in Similar Dry Weights

In our goal to examine the phytochemical profile of C. sumatrensis, we aimed to reflect the traditional medicinal preparations of this plant in Uganda. Typically, the plant material is applied as a paste, a poultice, or a decoction [9]. Recent studies of C. sumatrensis often worked with ethanol extracts [4]. We thus selected aqueous decoction and ethanolic maceration as common extraction methods and compared them in the course of our study. First, we determined their extraction efficiency. The extraction of 35.0 g powdered C. sumatrensis plant material yielded 2.8 g dry extract (8.0%) from the aqueous decoction and 3.9 g (11.1%) from ethanolic maceration. This indicates that ethanolic maceration is a slightly more efficient extraction method.

3.2. TLC-UV/Vis Analysis Suggests the Presence of Steroids and Other Terpenes in the Plant Extracts

In order to obtain an initial overview of the chemical composition and a first comparison of the aqueous and ethanolic extracts of C. sumatrensis, we carried out a TLC UV/Vis analysis. TLC is known as a very convenient, cost-effective, and rapid method to separate non-volatile metabolites. The use of suitable detection reagents allows a preliminary identification of specific compound classes. With respect to the available literature [23,24], we focused on the detection of terpenes using anisaldehyde/H₂SO₄. Terpenes represent a highly diverse and extensive group of natural products [25]. Potential effects of terpenoid compounds in plants include functions in defense mechanisms, responses to environmental stresses [26], and the attraction of pollinators [27].

Anisaldehyde/sulphuric acid produces distinct visible color changes with the separated analytes, including those which might not be detected in standard HPLC-UV analysis [24]. Indeed, after derivatization, the TLC chromatograms displayed various colored spots indicating the presence of different classes of terpenes. Examination of the plates under visible light and UV light revealed the presence of triterpenes and steroids in both the aqueous and ethanolic extracts of C. sumatrensis but monoterpenes only in ethanolic extracts (

Table 1. Preliminary investigation of phytochemical constituents in aqueous and ethanolic extracts of C. sumatrensis with TLC analysis and detection with anisaldehyde/H₂SO₄.

	Ethanolic Extracts	Aqueous Extracts
Monoterpenes	+	-
Triterpenes	+	+
Steroids	+	+

+ detected; - not detected.

), as detailed below.

In detail, the TLC chromatograms displayed a single blue spot with an R_f of 0.75 in the chromatogram, suggesting a distinct monoterpene in the ethanolic extract. Smaller terpenes, such as monoterpenes and sesquiterpenes, are typically volatile; therefore, their presence in the extracts was unexpected. However, the gentle shade-drying process might have helped to preserve some of these volatile compounds adsorbed to cell components or trapped inside intact cell wall compartments. In contrast, larger terpenoids, including diterpenes, sesterterpenes, and triterpenes, are predominantly non-volatile [28]. Three blue to violet-colored spots with R_f values of 0.50, 0.83, and 0.90 for ethanolic extracts (data shown in Figure S1 and Table S1) and two similarly colored spots for the aqueous extracts with R_f values of 0.72 and 0.77 indicate the presence of such terpenes.

In the TLC chromatograms, distinct gray spots were detected as tentative steroids [29] with R_f-values of 0.36 and 0.20 in the ethanolic extracts and 0.87, 0.38, and 0.13 in the aqueous extracts. These findings suggest the presence of different steroids and other terpenes in aqueous and ethanolic extracts of C. sumatrensis. Steroids, a subclass of terpenes, are mainly present in plants as phytosterols. The antimicrobial, anti-inflammatory, antioxidant, and anti-cancer properties of terpenes in general and of phytosterols in particular have been extensively studied [30]. Moreover, phytosterols stabilize plant cell membranes and regulate membrane fluidity [31]. In TLC analysis, carbohydrates can be observed in different colors (gray, red, turquoise) depending on their structure [24]. Neither ethanolic nor aqueous extracts showed any definitive carbohydrate-related signals in our analyses. However, four specific pink–red colored signals with R_f-values of 0.96, 0.93, 0.66, and 0.56 only occurred in the aqueous extract and could be interpreted either as triterpenes, glycosides, or carbohydrates. Overall, the TLC analysis of extracts of C. sumatrensis indicated a rich terpenoid profile, which prompted us to a more detailed investigation.

3.3. HPLC-UV Fingerprinting Suggests Similarity of Aqueous and Ethanol Extracts

The next step of our phytochemical profiling of C. sumatrensis involved the HPLC-UV fingerprinting of the aqueous and the ethanolic extracts. This method is widely used for authentication and quality assessment of herbal medicines, e.g., by regulatory drug authorities, and provides a broad perspective on the composition of UV-active metabolites due to its high resolution and sensitivity. In our case, HPLC fingerprints from the ethanolic and aqueous extracts (Figure 1) showed similar profiles.

Notably, UV detection revealed consistently higher peak intensities in ethanolic extracts compared to those obtained using aqueous extraction. The chromatograms showed seven major peaks that are common to both the ethanolic and aqueous extracts. These results indicate a highly similar chemical composition of UV-active compounds in both extracts. Further peaks were visible at 205, 220, and 310 nm (see Figure S2). Only the early-eluting peaks (at approximately 2–3 min) lacked absorbance at 310 nm. Peaks eluting between 10 and 16 min (especially peak numbers 5–7) may correspond to different flavonoid structures. The UV spectra of the seven most intense peaks appeared similar in both ethanolic and aqueous extracts and were not further characterized (see Figure S3).

3.4. GC-MS Allows for Comprehensive Annotation of Volatile Compounds in Dried Leaf Samples

Headspace GC-MS analysis provided detailed information on volatile thermally stable compounds of the raw leaf material to identify individual plant metabolites. The headspace was thermally extracted directly from the dried leaves without prior liquid extraction. This convenient approach, avoiding lengthy procedures such as prior distillation or liquid extraction, is particularly suitable for analyzing volatile compounds, which are often lost or degraded during liquid extraction; the latter is also often challenged by the selection of a suitable solvent matching subsequent analysis. The high sensitivity and selectivity of GC-MS ensures the generation of accurate and reliable data to enable compound annotation [32]. A first assessment of our data and search against the mass spectral database (Figure 2B) showed that most of the annotated peaks are terpenes (44%), followed by other hydrocarbons (36%). Of the terpenes, 65% are predicted as sesquiterpenes, 30% as monoterpenes, and only 5% as diterpenes. The lowest abundant volatile compounds are fatty acid esters, lactones, and carboxylic acids.

A detailed analysis assigned hexanal (6.78 min), α-pinene (10.09 min), D-limonene (13.16 min), α-copaene (23.40 min), norpinane (23.74 min), aristolene (24.56 min), bergamotene (25.31 min), α-humulene (25.45 min), β-caryophyllene (25.56 min), and neophytadiene (30.23 min) to the ten most prominent signals (Figure 2A). This corresponds well with previous GC-MS-based studies, which also identified limonene [33], α-bergamotene, and caryophyllene in n-hexane and ethyl acetate extracts of C. sumatrensis [1]. In accordance, α-pinene, α-copaene, α-humulene, and D-limonene were reported from the essential oil of this plant, contributing to its bioactive properties [10]. The three largest peaks by area in our analysis were assigned to terpenes (aristolene, neophytadiene, and D-limonene). In total, 47 volatile compounds were detected in the powdered leaf samples of C. sumatrensis and annotated with a spectral library (

Table 2. Volatile compounds detected with GC-MS headspace analysis. The compounds were assigned and ordered according to their corresponding chemical class based on their basic structure. Within each class, compounds are ordered according to their peak area; retention times and RI are listed in the last columns. The peak area was obtained from a representative file. The indicated relative peak areas refer to the sum of all peak areas over the entire chromatogram.

Compound Class	Compound Name	Peak Area [%] *	Retention Time [min]	RI
Monoterpenes	D-Limonene	7.59	13.16	1034
	α-Pinene	2.35	10.09	931
	p-Cymene	1.33	13.01	1029
	β-Pinene	0.81	11.51	979
	γ-Terpinene	0.58	14.09	1065
	Camphene	0.14	10.61	948
Sesquiterpenes	Aristolene	13.03	24.56	1414
	β-Caryophyllene	5.54	25.56	1448
	α-Humulene	5.44	25.45	1444
	α-Copaene	3.40	23.40	1376
	β-Bergamotene	2.25	25.31	1439
	β-Selinene	1.72	26.15	1467
	α-Selinene	1.32	26.29	1472
	γ-Muurolene	1.10	25.85	1457
	Salvial-4(14)-en-1-one	0.87	27.80	1523
	β-Cubebene	0.85	24.82	1423
	β-Bourbonene	0.78	23.62	1383
	δ-Cadinene	0.77	26.65	1484
	cis-Calamenene	0.60	26.70	1486
	β-Copaene	0.55	25.99	1462
Diterpene	Neophytadiene	12.94	30.23	1604
Hydrocarbons	Norpinane	1.67	23.74	1387
	2,4-Dimethyl-1-decene	0.86	14.66	1084
	3,3-Dimethyl-octane	0.63	12.48	1011
	2,5,5-Trimethyl-heptane	0.52	12.61	1015
	2,4-Dimethyl-1-heptene	0.44	7.239	836
	5-Ethyl-2-methyl-octane	0.43	13.99	1061
	4,6-Dimethyl-dodecane	0.43	19.60	1249
	3,8-Dimethyl-undecane	0.43	19.80	1255
	Dodecane	0.42	18.46	1211
	Heptadecane	0.41	20.58	1281
	3,7-Dimethyl-undecane	0.35	15.41	1109
	5-Methyl-undecane	0.30	14.17	1067
	2,6,7-Trimethyl-decane	0.25	18.82	1223
	Dotriacontane	0.25	26.06	1464
	Hexadecane	0.24	21.88	1325
	3,7-Dimethyl-decane	0.17	19.36	1241
Lactone	(2,6,6-Trimethyl-2-hydroxycyclohexylidene)acetic acid lactone	1.72	26.80	1489
Carboxylic acid	Acetic acid	1.59	2.56	680
Fatty acid ester	Phytyl palmitate	0.30	30.59	1616
Aldehydes	Hexanal	6.78	6.11	798
	Pentanal	1.43	3.87	723
Alcohols	3,7,11,15-Tetramethyl-2-hexadecen-1-ol	1.25	30.43	1610
	1-Penten-3-ol	0.79	3.66	716
	1-Octen-3-ol	0.41	11.61	982

* The TIC peak area was determined using GCMSsolution version 4.20 (Shimadzu, Kyoto, Japan) as a semiquantitative indicator for relative abundance and refers to the sum of all peak areas over the entire chromatogram.

). Of those, 41 compounds could be newly assigned to this plant, including major constituents, e.g., aristolene, norpinane, and neophytadiene. The relative abundance of the metabolites was determined by calculating the share of total peak area [%] as a semiquantitative indicator. For a better and more detailed overview and semiquantitative evaluation based on the peak area of the compounds, please refer to Table 2.

Aristolene (peak area: 13.03%, retention time: 24.56 min) and caryophyllene (peak area: 5.54%, retention time: 25.56 min) were identified as the most abundant sesquiterpenes. The presence of aristolene has been previously noted in the analysis of other essential oils [34], but to the best of our knowledge, the therapeutic relevance of aristolene has not yet been investigated. Caryophyllene, however, is a widely studied phytochemical and has diverse therapeutic properties, e.g., analgesic, anti-inflammatory, antimicrobial, antifungal, and antioxidant [35]. There is also evidence suggesting potential antidepressant and anxiolytic effects associated with caryophyllene [36]. The most prominent monoterpene was suggested as limonene (peak area: 7.59%, 13.16 min). It is well-investigated, e.g., for its anti-cancer, anti-inflammatory, and antimicrobial activities [37]. Neophytadiene (peak area: 12.94%, retention time: 30.23 min) is the only suggested diterpene and the second-highest signal overall. It has shown anti-inflammatory activity and neuropharmacological effects and has been suggested as a potential anxiolytic and antidepressant agent [38]. In summary, the headspace analysis supported major constituents of C. sumatrensis from previous reports but also adds metabolic depth and several new compounds with interesting bioactivity and possible indications. Whether these bioactivities are found in aqueous formulations, which are often used in traditional medicine, remains to be shown. And clearly, the preliminary status of our compound annotation requires further confirmation. There is also still room for optimization to enhance the sensitivity, specificity, and overall analytical efficiency of these methods. Future studies should focus on refining these techniques to improve the detection and quantification of minor compounds, which could further elucidate the plant’s phytochemical profile.

3.5. LC-MS/MS Phytochemical Profile of Ethanolic and Aqueous Extracts

To obtain a more comprehensive phytochemical profile of C. sumatrensis, we applied LC-MS/MS analysis and generated feature-based mass spectral similarity networks. Generally, LC-MS is a proficient method to detect complex secondary metabolites, particularly non- and semi-volatile or thermally labile compounds [32]. In addition, tandem mass spectra provide valuable structural information on the detected metabolites. Hence, we analyzed ethanolic and aqueous extracts with untargeted tandem mass spectrometry in data-dependent acquisition (DDA) mode. The generated LC-MS/MS data were processed by MZmine [15] and visualized as a molecular network using the GNPS environment [14]. In molecular networking analysis, each signal (feature) is represented as a node with a specific m/z and retention time range. The metabolites are grouped into several clusters based on similarities of their MS/MS fragmentation patterns, indicating related chemical structures. With feature-based molecular networking (FBMN) [13], improved spectral annotation is possible compared to classical GNPS [11]. In our experiment, data processing revealed a total of 2565 features in all 12 samples, i.e., two cultivated plants, each extracted with two solvents, measured in triplicate. After filtering by the MZmine algorithm, 864 features could be identified that are present in at least three samples. Subsequently, we evaluated the distribution of the detected features across the different extraction methods (Figure 3A,B). A prediction of the corresponding compound class of several nodes was possible, and the nodes from the GNPS network were annotated (see Supplementary File Table S2).

In ethanolic extracts, we detected 217, and in aqueous extracts, 53 exclusive features. Approximately two-thirds of the features (594) are present in both ethanolic and aqueous extracts. Next, the detected metabolites were subjected to automated chemical classification by SIRIUS [18] using the ClassyFire [20] and the Natural Product Classifier (NPC) algorithms [21]. Both algorithms rely on CANOPUS (class assignment and ontology prediction using mass spectrometry), a deep neural network, which is trained for chemical structure identification from MS/MS spectra [19]. CANOPUS uses a machine learning algorithm (CSI:FingerID) to compute fragmentation trees and predict molecular fingerprints, which can be compared to structure databases [39]. With the ClassyFire algorithm, 177 features were classified into four taxonomic chemical levels, from superclass, class, and subclass to the most specific class (Figure 4). With the NPC algorithm, 200 compounds were assigned to pathway, superclass, and class levels (see Figure S5).

With the ClassyFire algorithm, approximately 41% of the plant metabolites from the LC-MS/MS analysis were ranked as lipids and lipid-like molecules, including terpenoids, glycerolipids, steroids, and phospholipids. Of the predicted terpenes, most were classified as terpene glycosides or saponins and further as di-, tri-, sester-, and tetraterpenes. This nicely complements our GC-MS data, which showed an abundance of volatile mono- and sesquiterpenes [28] (Figure 2B and Table 2) and reveals C. sumatrensis as particularly rich in terpenoid chemistry. Moreover, this also confirms initial data obtained from TLC UV/Vis. Organic acids and derivatives were predicted to be the second most abundant class with 17%, including amino acids and analogues, carboxylic acids, and derivatives. Organic oxygen compounds with, e.g., alcohols, polyols, and carbohydrates, account for 13% of the identified compounds and phenylpropanoids with flavonoids account for 8%. The remaining major chemical classes are benzenoids and organoheterocyclic compounds (each 7%). Organic nitrogen compounds, lignans and related compounds, nucleotides, and alkaloids sum up to a total of 7%. Similar to the previously presented TLC and GC-MS data, the LC-MS measurement also indicated the presence of steroids and other terpenes, indicating a rich terpenoid metabolism in C. sumatrensis.

3.6. Compound Prediction with In Silico Techniques and Database Matches

In addition to SIRIUS compound class prediction, we performed an automated search against the spectral libraries from GNPS and MassBank of North America (MoNA) (as described in Section 2.8) and mapped the hits in the network (see

Table 3. Library matches with the corresponding name, database, and URL.

Cluster	Name	Database	GNPS Library URL
1	Lithocholic acid-C12:1	ECG-ACYL-ESTERS-C4-C24-LIBRARY	SpectrumID=CCMSLIB00010012466 https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00010012466 (accessed on 19 March 2025)
	Lithocholic acid-C20:3	ECG-ACYL-ESTERS-C4-C24-LIBRARY	SpectrumID=CCMSLIB00010012642 https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00010012642 (accessed on 19 March 2025)
	Cholic acid-C16:0	ECG-ACYL-ESTERS-C4-C24-LIBRARY	SpectrumID=CCMSLIB00010012382 https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00010012382 (accessed on 19 March 2025)
2	10S-Hydroxypheophorbide a	GNPS-LIBRARY	SpectrumID=CCMSLIB00010128701 http://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00010128701 (accessed on 19 March 2025)
	Phaeophorbide a	GNPS-LIBRARY	SpectrumID=CCMSLIB00010128702 https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00010128702 (accessed on 19 March 2025)
3	3,5-Dicaffeoylquinic acid	GNPS-NIH-NATURALPRODUCTSLIBRARY_ROUND2_POSITIVE	SpectrumID=CCMSLIB00000847510 http://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00000847510 (accessed on 19 March 2025)
	Cynarin	BMDMS-NP	SpectrumID=CCMSLIB00006366878 https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00006366878 (accessed on 19 March 2025)
	Loganoside + caffeic acid (Compound NP-013263)	GNPS-NIH-NATURALPRODUCTSLIBRARY_ROUND2_POSITIVE	SpectrumID=CCMSLIB00000854335 http://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00000854335 (accessed on 19 March 2025)
4	No library hit
5	Flavone base + 3O, O-HexA-HexA	MASSBANK	SpectrumID=CCMSLIB00005741225 https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00005741225 (accessed on 19 March 2025)
6	Luteolin 3′,4′-diglucoside	GNPS-NIH-NATURALPRODUCTSLIBRARY_ROUND2_POSITIVE	SpectrumID=CCMSLIB00000847749 http://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00000847749 (accessed on 19 March 2025)
7	Gaultherin	GNPS-NIH-NATURALPRODUCTSLIBRARY_ROUND2_POSITIVE	SpectrumID=CCMSLIB00000855202 https://gnps.ucsd.edu/ProteoSAFe/gnpslibraryspectrum.jsp?SpectrumID=CCMSLIB00000855202 (accessed on 19 March 2025)

for the database hits). SMILES codes of the library hit structures were provided by GNPS databases, and based on these SMILES, the structures were generated with ChemDraw (version 22.0.0) (Figure 3C). The structure predictions were validated by MS2LDA analysis (see supplementary file Table S3). Therefore, an unsupervised substructure annotation of the fragmentation spectra was performed with the MS2LDA Web app [22]. This approach uses the Latent Dirichlet Allocation (LDA) algorithm to discover co-occurring fragments and neutral losses in MS/MS spectra (Mass2Motifs) that are indicative of specific structural features. We found 15 clusters with at least three connected nodes, also referred to as molecular families, as well as 17 clusters with two nodes and 607 singletons.

In the dominating Cluster 1 (Figure 3C), the prediction showed lipids and lipid-like molecules, including terpenoids and fatty acids. Three spectral matches showed acylated phytosterols in the ethanolic extracts. This correlates with the in silico prediction discussed above (Figure 4), suggesting the presence of lipids and lipid-like molecules in C. sumatrensis. Phytosterols are cholesterol-like compounds and include sterols and stanols, which differ in the structure of the alkyl side chain and the degree of saturation. It has been reported that a diet rich in phytosterols may reduce the risk of developing certain cancers in humans and animals [40]; the ability of phytosterols to lower cholesterol levels in humans was already demonstrated in the 1950s [41].

Cluster 2 was predicted to contain chlorins, mainly porphyrin metabolites, and two spectral library hits (pheophorbide A and 10 S-hydroxy pheophorbide) support these findings. Cluster 3 was assigned to quinic acid derivatives with three spectral matches, including cynarin. MS2LDA analysis showed a cumaric acid-related pattern (rhamn_motif 28) for four nodes, which supports the findings. Most of the nodes in this cluster are present in both aqueous and ethanolic extracts. However, the node predicted as cynarin and another node with m/z 585.1238 are present only in aqueous extracts.

Cluster 4 has been classified as acetylated polyamines. Natural polyamines such as N-acetylspermine or spermidine can play a major role in plant stress tolerance. They can form conjugates with cinnamic acid derivatives, serve as reserve polyamines for proliferative tissue, and act against pathogens or for the detoxification of phenolic compounds [42]. Polyamines are also used as polymers in a variety of ways, e.g., as a carrier for drug delivery systems, as a polyethylene glycol alternative, and as a food additive [43]. Through accumulation in the food chain and via wastewater, these polymers can enter the environment and then be taken up by plants [44]. In the environment, they can have the same effect on plants as nitrogenous fertilizers [45].

Cluster 5 comprises three nodes representing flavonoids with spectral matches supporting this finding. Cluster 6 is also classified as flavonoids with one node assigned as a phenylpropanoid related Mass2Motif (rhabdus_motif_91) by MS2LDA. Overall, with MS2LDA, some molecular class predictions were consistent with those of the molecular network and spectral library hits. Flavonoids are secondary metabolites with a polyphenolic structure and are ubiquitous in plants [46]. They are, e.g., responsible for the color, smell, and taste of plants and have several important physiological functions, such as regulating cell growth, regulating UV filters, scavenging reactive oxygen species, and attracting pollinators [27]. Flavonoids are produced by plants against microbial infestation, among other conditions, and have been shown to have positive effects on human health. The positive impact is explained by antioxidant properties, which are mediated by substituted hydroxyl groups that can scavenge free radicals. Studies have also shown effects against cardiovascular diseases and cancer, as well as hepatoprotective, antibacterial, antiviral, and anti-inflammatory effects [47]. All signals in cluster 5 were only present in the aqueous extracts, and since flavonoids as aglycons are usually non-polar and poorly water-soluble, it is likely that these putative flavonoids would be glycosylated, improving water solubility [48]. And indeed, the library match for (ID 55 within this cluster) showed a glycosylated apigenin derivative. Cluster 7 showed a spectral match to a structure of a benzoic acid methyl ester called gaultherin that is present in both ethanolic and water extracts. This compound is of scientific interest due to its analgesic and anti-inflammatory effects similar to those of salicylic acid, which was shown by Zhang et al. [49]. Notably, gaultherin has mainly been found in plants of the genus heather family (Ericaceae). Further studies are thus necessary to evaluate the presence of gaultherin in C. sumatrensis and the possibility of analgesic properties of this plant.

The use of in silico tools for compound class prediction has become increasingly important and can support classical library and database searches [18]. However, it is essential to note the limitations of our approach: although we were able to predict natural product classes, we were mainly unable to assign the features to specific compounds. The purification and structural characterization of these compounds are essential to facilitate their availability for bioassays, enabling a more comprehensive exploration of the plant’s bioactivity profile. Notably, more than half of the nodes could not be predicted using these tools. Similar outcomes in other studies have been explained by poor ionization, fragmentation, and separation of analytes during the LC-MS/MS run [18,19].

Overall, this study demonstrates the potency and limitations of a multi-technique analytical approach in the field of plant metabolomics. Database searches and comparisons of the literature enabled the preliminary identification of over 40 substances present in the plant that have not been reported before from C. sumatrensis. Our evaluation strategy has successfully provided interesting insights, demonstrating the potential of advanced analytical techniques in phytochemical research. When comparing the aqueous and ethanolic extracts, the HPLC fingerprints and the LC-MS measurements revealed a high degree of similarity between the two extracts, with the majority of features being present in both.

Most of the results we obtained by different methods can be regarded as complementary. The in-depth analysis of the phytochemical profile provided, for the first time, valuable insights into the volatile metabolite profile of C. sumatrensis and has revealed a diverse array of potentially bioactive compounds, highlighting the plant’s potential for therapeutic applications. These compounds exhibit unique properties that merit further investigation to elucidate their specific biological activities and mechanisms of action.

3.7. Pharmacological Investigations (Antimicrobial Activity)

In our GC-MS analysis, we detected, amongst others, α-pinene and D-limonene, which are known for their antimicrobial properties [50]. As a first indicator of the bioactivity of the extracts, we conducted antimicrobial assays against the three different representative microorganisms. Bacillus subtilis DSM109511 was used as an indicator for activity against Gram-positive bacteria along with Escherichia coli DSM498 and Pseudomonas fluorescens DSM289 as indicators for activity against Gram-negative bacteria. The genus Pseudomonas is known for its nonsusceptibility against a variety of antibiotics and is thus of particular clinical interest. The use of these strains allows a comprehensive evaluation of the antimicrobial activity of the extracts against both Gram-positive and Gram-negative bacteria, as they exhibit different properties. For the assays, 0.5 mg (50 mg/mL) of the extracts was applied to agar plates inoculated with the target microorganisms. The inhibition zones (IZ), which indicate the extent to which the growth of the microorganisms was suppressed, were measured and recorded (see

Table 4. Antimicrobial testing of ethanolic and aqueous extracts. Diameter of the zones of inhibition (IZ), including the filter disc of 5 mm. The extracts were tested against E. coli, B. subtilis, and P. fluorescens in a concentration of 50 mg/mL. MICs were determined using extract concentrations ranging from 0.156 to 5 mg/mL. As positive controls, gentamycin (gent) for E. coli and P. fluorescens and erythromycin (ery) for B. subtilis were used in a concentration of 10 µg/mL.

	Positive Control	Ethanolic Extracts		Aqueous Extracts
	⌀ IZ	⌀ IZ	MIC	⌀ IZ	MIC
E. coli	20 mm (gent)	7 mm	0.313 mg/mL	-	-
B. subtilis	14 mm (ery)	9 mm	>2.5 mg/mL	-	-
P. fluorescens	17 mm (gent)	9 mm	0.078 mg/mL	-	-

). The minimal inhibition concentration (MIC) is described as the minimal concentration of the extract that inhibits cell growth. The MICs were determined in a range between 5 and 0.156 mg/mL extract.

The results of the antimicrobial assays revealed a notable distinction between the ethanolic and aqueous extracts and demonstrated antimicrobial activity of the ethanolic extracts while the aqueous extracts showed no impact (see Figure S7). Ethanolic extracts exhibited significant antimicrobial activity against all tested microorganisms, as evidenced by the formation of clear inhibition zones. The mean inhibition zone diameters for the ethanolic extracts ranged from seven to nine millimeters (including 5 mm of the filter disc), depending on the microorganism.

The minimal inhibition concentration (MIC) was determined as 0.313 mg/mL for the ethanolic extracts against E. coli and 5 mg/mL for B. subtilis (see Figures S8 and S9). The aqueous extracts showed no inhibitory effects on the growth of E. coli or B. subtilis.

These findings suggest that the active antimicrobial compounds within the plant extracts might be more effectively solubilized in ethanol, highlighting the importance of solvent selection in the extraction process. Further chemical analysis, accompanied by purification steps improving the bioactivity of the extracts, is warranted to identify the specific bioactive constituents responsible for the observed antimicrobial activity. These differences underscore the potential of specific plant extracts as natural antimicrobial agents and highlight the need for further investigation into their active compounds and mechanisms of action. To fully understand the biological variety of the phytochemical composition, additional replicate analyses are necessary. These can include samples from various geographical locations or parts of the plant. Such comprehensive sampling will provide a more robust understanding of the plant’s chemical diversity and help identify any regional variations that could impact its bioactivity.

4. Conclusions

We employed various analytical techniques to assess the phytochemical composition and capacity of C. sumatrensis. This included readily accessible and cost-effective methods, such as TLC and HPLC-UV, as well as more complex techniques, such as GC-MS and LC-MS/MS. Additionally, SIRIUS and CANOPUS were used as in silico tools to assign compounds or compound groups to recorded LC-MS/MS features. Moreover, we generated a first molecular network from metabolomic data of C. sumatrensis. The approach of combining classical analytical methods with molecular networking and structural class assignment allowed us to comprehensively investigate the plants’ metabolic profile. This integrated strategy enhanced the identification and characterization of metabolites, thereby substantially expanding our knowledge of the metabolic capacity of C. sumatrensis. While our study has laid a solid foundation for understanding the phytochemical profile of this plant, continued research and optimization are essential to fully harness its therapeutic potential.