Antibacterial Activity from Momordica charantia L. Leaves and Flavones Enriched Phase

Momordica charantia L. (Cucurbitaceae) is a plant known in Brazil as “melão de São Caetano”, which has been related to many therapeutic applications in folk medicine. Herein, we describe antibacterial activities and related metabolites for an extract and fractions obtained from the leaves of that species. An ethanolic extract and its three fractions were used to perform in vitro antibacterial assays. In addition, liquid chromatography coupled to mass spectrometry and the molecular networking approach were used for the metabolite annotation process. Overall, 25 compounds were annotated in the ethanolic extract from M. charantia leaves, including flavones, terpenes, organic acids, and inositol pyrophosphate derivatives. The ethanolic extract exhibited low activity against Proteus mirabilis (MIC 312.5 µg·mL−1) and Klebsiella pneumoniae (MIC 625 µg·mL−1). The ethyl acetate phase showed interesting antibacterial activity (MIC 156.2 µg·mL−1) against Klebsiella pneumoniae, and it was well justified by the high content of glycosylated flavones. Therefore, based on the ethyl acetate phase antibacterial result, we suggest that M. charantia leaves could be considered as an alternative antibacterial source against K. pneumoniae and can serve as a pillar for future studies as well as pharmacological application against the bacteria.


Introduction
Clinical infections caused by resistant bacteria have become a major public health concern worldwide, and approximately 700,000 deaths per year are caused by this type of bacteria [1]. It is estimated that by 2050, there will be more than 10 million deaths/year credited to 'superbugs', with the expectation of the highest rate being reported in developing countries [2]. The demand for new antimicrobial agents has been growing at the same time as the discovery and advancement of multi-resistant bacteria. Therefore, researchers have been focusing a great effort on the search for new therapeutic alternatives against multidrugresistant bacteria [3]. In this sense, plant derivatives appear as potential sources of new drugs that act in different ways to deactivate or block the growth of such pathogens [4]. The Cucurbitaceae family includes several species widely distributed throughout the tropical and subtropical regions. There are antimicrobial activities associated with crude extracts and isolated metabolites of species from this family [5]. Among the species, Momordica charantia L., popularly known as "melão de São Caetano", is considered an invasive plant in Brazil and can be found in different places in the country. It is frequently grown in orchards and coffee plantations, or even on fences and debris in abandoned land [6], and is composed of compound classes such as terpenoids, saponins, and flavonoids [5,7]. This species can be highlighted due to its antimicrobial [5], nutraceutical and inflammatory properties [8], as well as healing of gastric ulcers [9], rheumatism [10], and so on. In developing countries, M. charantia has been employed in folk medicine for several other pharmacological purposes, such as to treat toothache, diarrhea, furuncle, cancer, hypertension, obesity, bacterial and viral infections, diabetes, pneumonia, and even AIDS [11][12][13][14][15][16].
Although the chemical characterization of such plants can lead to new pharmaceuticals, it remains challenging due to the presence of several components with different physical-chemical properties, including many in relatively small quantities [17]. Nuclear magnetic resonance (NMR) techniques provide valuable structural information used for structure characterization and therefore, provide unequivocal identification. Liquid chromatography coupled to mass spectrometry (LC-MS) plays a crucial role as it is compatible with chromatographic techniques that allow the separation of compounds in the sample, leading to a deeper investigation of the whole chemical content due to the high sensitivity of MS that can detect metabolites of picomole to femtomole levels in some cases [18]. Furthermore, MS allows the detection of metabolites as ions in the form of mass to charge ratio (m/z) [19], and also allows the fragmentation of the ions at the gas phase by collisioninduced dissociation (CID), providing MS/MS spectra that contain valuable information to contribute to annotation of the chemical structure [20]. It is worth mentioning that the acquisition of scan spectra by mass spectrometry is fast, and a single chromatographic run can provide thousands of MS spectra. Therefore, methods that allow a fast and easy analysis of such data are welcome in this field.
In this context, molecular networking, a tool from the Global Natural Products Social Molecular Networking (GNPS) [21] infrastructure, allows the visualization of MS/MS data by organizing it by spectral similarity. This is a very effective strategy, especially because establishing a similarity relationship provides organization into molecular families (usually related chemical structures), and, therefore, finds related metabolites in the dataset even in small amounts. GNPS also contains a spectral library of known compounds, and the molecular networking allows automatic searches in the spectral library, which contributes to speed up the dereplication of known compounds, a phytochemical approach that has been widely used in recent decades, including the modest contribution of these authors [22][23][24][25][26], and which is also unknown by recognition of the analogs into the molecular families.
Based on the antimicrobial potential already reported for M. charantia, this study describes the evaluation of in vitro antibacterial activity for the ethanolic extract and fractions from the leaves of M. charantia. In addition, liquid chromatography coupled to highresolution mass spectrometry (LC-HRMS) analyses allowed the chemical characterization of the samples. The leaves were washed with water from the Direct-Q5 system (Millipore, Darmstadt, Germany) and decontaminated with sodium hypochlorite solution (NaOCl, 0.1%) acquired from Dinâmica (Jaraguá do Sul, SC, Brazil). The samples were dried in an air circulation oven (Quimis, Brazil) at 45 • C until constant weight. The dry material was crushed in ball mills up to a granulometry of 60 to 100 µm, obtaining 146.54 g. The mass was subjected to extraction with 1 L of ethanol (Tedia, Fairfield, OH, USA) at room temperature for 24 h (2×), and 35.30 g of crude extract was obtained after the solvent evaporation process. Thus, 1 g of ethanolic extract (EE) was mixed with a hydroalcoholic solution, consisting of 60 mL of ultra-pure water, 20 mL of ethanol, and 1 mL of hydrochloric acid (Dinâmica, Indaiatuba, SP, Brazil). The resulting solution was subjected to liquid -liquid partition (LLP) to obtain the hexane (PhHex), ethyl acetate (PhEA), and hydroalcoholic (PhWOH) phases, respectively.

Liquid Chromatography-High Resolution Mass Spectrometry Analysis
The analyses were performed on a Xevo G2-S QqTOF mass spectrometer (Waters Corp., Milford, MA, USA) equipped with a LockSpray source. The instrument was calibrated with a mass of reference (leucine-enkephalin) utilized for accurate mass measurements. MassLynx 4.1 software was used for system control and data acquisition. The samples were analyzed in a BEH C18 column (Waters Corp.; 50 mm; 2.1 mm; 1.7 µm particle size) using ultra-pure water (solvent A) and acetonitrile (solvent B), both containing 0.1% formic acid. The column temperature was maintained at 40 • C. Linear gradient elution was performed with a flow of 300 µL/min and 5 -95% of solvent B in 20 min. The injection volume of the samples was 5 µL. The mass spectra data were recorded in a negative ionization mode (ESI) for a mass range from m/z 50 to 1200. The source temperature was set to 120 • C with a cone gas flow of 50 L/h. The desolvation gas flow was set to 600 L/h at a temperature of 150 • C. The capillary was set at 3.0 kV with cone voltage at 40 V. The settings of the data-dependent acquisition (DDA) experiments were: centroid format, number of ions selected 5 (Top5 experiment), the normalized collision energy (NCE) was set to 10, 20, 30, 40 and 50, scan rate of 0.5 sec, charge states of +1 and +2, tolerance window of ±0.2 Da and peak extract window of 2 Da, tolerance of deisotope ± 3 Da, extraction tolerance of deisotope 6 Da.

Mass Spectrum Data Treatment
The raw files of the EE and PhEA acquired in the Xevo G2-S QTof mass spectrometer (Waters Corp., Milford, MA, USA) were converted into mzML format using the software MS Convert of the ProteoWizard package [27] and were processed with the software MZmine 2.53 version [28]. The limit for the detection of ions in negative mode at the MS 1 level was set at 1.0 × 10 3 and MS 2 at 5.0 × 10 1 . Chromatograms were constructed using ADAP with a minimum group number of 3 and a minimum group intensity limit of 1.0 × 10 3 , a min highest intensity of 3.0 × 10 3, and an m/z tolerance of 10.0 ppm. The local minimum search algorithm was used to deconvolve the chromatogram, with an m/z tolerance of 0.5 for the pairing of MS 2 and 0.2 min for RT. Isotopes were detected using a peak window with a tolerance of 10.0 ppm, an RT tolerance of 0.5 min, a maximum charge of 1. For peak alignment, the tolerance of m/z 10 ppm was used, scores for m/z of 75 and 25 for RT with a tolerance of 0.2 min. The resulting list was filtered to remove duplicates and lines with no associated MS 2 spectrum. Then, gap filling was used to fill in the gaps in the peak list. The resultant files were exported using FBMN-GNPS.

Molecular Networks
The molecular network was built from the mgf and CSV files exported from MZmine. We used metadata to organize metabolite information according to the online workflow (available online: https://ccms-ucsd.github.io/GNPSDocumentation/, accessed on 30 January 2022) available on the GNPS website (available online: http://gnps.ucsd.edu, accessed on 8 July 2022) [21]. The tolerance of m/z for the precursor ion was adjusted to 0.02 Da and for fragment ion to 0.02 Da. Minimum cosine score above 0.5 and the minimum number of fragment ions were fixed on 4. The spectra on the network were then searched in the GNPS spectral libraries. The database spectra were filtered with a minimum cosine score above 0.6 and a minimum of 4 fragment ions correspondence. The job is available online at the link: https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=4890e08934bb433 4b8738a346c10e7c4, accessed on 8 July 2022. The results were visualized and organized using Cytoscape version 3.8.2 (Seattle, WA, USA) [29]. Lastly, the correlation between PhEA and bioactivity score of the metabolites was obtained from NPAnalyst [30].

In Vitro Antibacterial Assay
Three strains of bacteria were used to evaluate the potential of the extracts: Staphylococcus aureus ATCC 25923, and Klebsiella pneumoniae ATCC 700603, provided by Instituto Evandro Chagas, Pará State (Brazil) and the bacteria Proteus mirabilis LACEN 8/7 (human isolated) provided by the Central Laboratory of Pará State collection. These microorganisms were selected for being common pathogens that can infect humans, animals or plants.
The pure cultures were maintained by routine sub-culturing at one-week intervals in BHI broth (Brain Heart Infusion, Kasvi, Spain), incubated at 37 • C, and spiked for 24 h for their metabolic activation. The minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) were conducted using a method approved by the Clinical and Laboratory Standards Institute in 96-well microtitration plates [31].

Determination of Minimum Inhibitory Concentration (MIC) and Minimum Bactericidal Concentration (MBC)
The antimicrobial susceptibility test was conducted using a method approved by the Clinical and Laboratory Standards Institute [31]. The tests were carried out with a stock solution of 5 mg·mL −1 of the crude extract/phases using the successive dilution method to obtain the concentrations from 2500 to 78.1 µg·mL −1 . Ciprofloxacin (Medley, Brazil) and vancomycin (1 mg·mL −1 each) were used as positive controls and BHI (Brain Heart Infusion) culture medium was used as negative control.
A total of 5 mg of extract/phases was dissolved in 100 µL of DMSO (Neon, Brazil) contained in Eppendorf tubes. Then, 900 µL of sterile BHI was added and stirred for better homogenization. In a 96-well microtitration plate (KASVI, Brazil), 100 µL of BHI broth was added to each well. Then, 100 µL of solution containing the samples was added to the first well of each column.
In each well, 5 µL of the bacterial suspension (10 4 Colony Forming Unit CFU/mL as required by CLSI) was inoculated and adjusted to 0.5 McFarland Standard scale and then the plates were incubated at 37 • C for 24 h. The results were read by adding 10 µL of TTC (2,3,5-triphenyltetrazolic chloride, NEON, Brazil). To prepare the dye, 0.2 g of TTC was added to the penicillin type flask containing 10 mL of sterile distilled water. The final solution showed a translucent color, and when in contact with environments where there are microorganisms presents a red color.
The type of activity presented in each concentration (bacteriostatic or bactericidal) was checked. In the cavities where there was no red color caused by the reaction of TTC with the bacteria, 5 µL of the volume contained in the wells was (re) inoculated in a Petri dish containing BHI agar culture medium and incubated at 37 • C for 24 h. Wells where bacteria grew indicated a bacteriostatic effect at that concentration and wells without bacteria indicated a bactericidal effect.

Antibacterial Activity
The EE and the PhEA of M. charantia showed antimicrobial activity and, therefore, they were selected to be analyzed by LC-MS/MS. Both positive controls of vancomycin and ciprofloxacin had a MIC of 7.8 µg·mL −1 . The EE demonstrated a good bactericidal effect against Klebsiella pneumoniae and Proteus mirabilis. The PhEA stands out for presenting the best result with a MIC of 156.2 µg·mL −1 for the K. pneumoniae, suggesting that this presents bioactive substances, possibly the glycosylated flavones, responsible for such activity (Table 1). We highlighted the bactericidal effect of Klebsiella Pneumonie, and Proteus Mirabilis, both Gram-negative strains. The literature [32] reports that the cell wall of Gram-negative bacteria acts as a barrier to a number of substances, including antibiotics. However, recently it was confirmed that quercetin derivatives have strong antibacterial action against Gram-negative bacteria [33], and the ethyl acetate phase (PhEA, see Table 1) described in this study showed high content of quercetin (16) derivatives, i.e., quercetin-O-sambubioside (4), quercetin-O-glucoside (6), quercetin-O-glucosyl-6 -acetate (9), and quercetin-O-acetylpentoside (13). MIC values < 100 µg·mL −1 are considered significant antimicrobials; moderate inhibitors present MIC in the range of 100 to 625 µg·mL −1 ; and inhibitors with MIC > 625 µg·mL −1 are considered weak [34]. In this sense, the ethyl acetate phase showed moderate activity (Table 1) against K. pneumoniae (MIC of 156.2 µg·mL −1 ), while the ethanolic extract showed moderate activity against P. mirabilis (312.5 µg·mL −1 ) and weak activity against K. pneumoniae and S. aureus (625 µg·mL −1 ). We emphasize that the MIC value against K. pneumoniae is in the range moderate-significant, which characterize PhEA, a source of candidate inhibitors of important hospital bacteria. In previous reports, antimicrobial activity for leaves, fruits and seeds was reported against some clinically important bacteria [20]. The extract of the leaves showed the main results of being a potent inhibitor for Staphylococcus aureus, moderate for Staphylococcus epidermidis and weak for Candida albicans. Studies with extracts of the seed showed interesting activities against Escherichia coli, Salmonella typhi, and Staphylococcus aureus, but less activity against Pseudomonas aeruginosa [35].

Identification of Chemical Constituents
The total ion profiles of the EE and PhEA were recorded from 50 to 1200 Da in 20 min (Figure 1). The molecular formulas, main fragment ions, and putative names are shown in Table 2. A total of 32 major metabolites were detected in the EE. Of these, 25 compounds were annotated in level 2 and 3 of identification according to MSI [36] based on HRMS and MS/MS data and the literature; most of these compounds have well characterized fragmentation (MS/MS) profiles [37].
The total ion profiles of the EE and PhEA were recorded from 50 to 1200 Da in 20 min (Figure 1). The molecular formulas, main fragment ions, and putative names are shown in Table 2. A total of 32 major metabolites were detected in the EE. Of these, 25 compounds were annotated in level 2 and 3 of identification according to MSI [36] based on HRMS and MS/MS data and the literature; most of these compounds have well characterized fragmentation (MS/MS) profiles [37].

Molecular Networking (GNPS Annotation)
The molecular networking created with EE and PhEA showed 224 parent ions after removing the blank. Seven compounds (6, 7, 8, 10, 12, 23 and 32) including isomers were annotated based on MS 2 data available in the GNPS spectral libraries. A family of flavones was reported, and the structure of the compounds is shown in Figure 2.

Molecular Networking (GNPS Annotation)
The molecular networking created with EE and PhEA showed 224 parent ions after removing the blank. Seven compounds (6, 7, 8, 10, 12, 23 and 32) including isomers were annotated based on MS 2 data available in the GNPS spectral libraries. A family of flavones was reported, and the structure of the compounds is shown in Figure 2.  [21] and suspect list [38]. The arrows indicate the nodes that have a match in the spectral library and their respective structure. Node text indicates the parent ion, node color reports the extract of the plant (orange: ethanolic extract from leaves; blue: phase ethyl acetate from leaves).  [21] and suspect list [38]. The arrows indicate the nodes that have a match in the spectral library and their respective structure. Node text indicates the parent ion, node color reports the extract of the plant (orange: ethanolic extract from leaves; blue: phase ethyl acetate from leaves).  [43,44].
The peak 32 [M − H] − of m/z 571.2882 was characterized as 1-hexadecanoyl-sn-glycero-3-phospho-(1 -myo-inositol), an important inositol pyrophosphates derivative present in plants as a signaling metabolite [45][46][47]. While information about this new class of molecules in plants is still scarce, the enzymes responsible for their synthesis have recently been elucidated [48]. In this sense, 32 is being reported for the first time in the genus. Despite the report of some sulfur-derived compounds in the M. charantia sample [49], they were not observed in this work.
A molecular family of glycosylated terpene derivatives (Figure 3) was detected in the EE. The ion of m/z 385.1856 is referent from peak 3, identified as hydroxy-2,4,4-trimethyl-3-(3-oxobutyl)-2-cyclohexen-1-one glucoside, and the pathway of fragmentation has been shown in previous reports [50].  According to Figure 2, the compounds 3, 22, 25, and 27 showed the presence of glycoside. In this sense, following the chemosystematic from the genus, for example, the compound 25 was characterized as momordicoside L isomer, a cucurbitane-type triterpenoid already reported for the species [52]. According to the literature [52],

Bioactivity and Structure
Previous studies reported that flavonoids have antimicrobial activity [55,56]. These stand out even more because of their antibacterial properties, especially against strains of According to Figure 2, the compounds 3, 22, 25, and 27 showed the presence of glycoside. In this sense, following the chemosystematic from the genus, for example, the compound 25 was characterized as momordicoside L isomer, a cucurbitane-type triterpenoid already reported for the species [52]. According to the literature [52],

Bioactivity and Structure
Previous studies reported that flavonoids have antimicrobial activity [55,56]. These stand out even more because of their antibacterial properties, especially against strains of Gram-negative bacteria, which are responsible for serious opportunistic infections and are resistant to common therapies. In this sense, the study of plants with high flavonoid content should be highlighted [57]. Thus, our study focused on the ethyl acetate phase (MIC 156.2 µg·mL −1 ) against K. pneumoniae, which stands out for its ability to develop enzymatic resistance mechanisms and is considered to be largely responsible for several infectious diseases [58]. The PhEA proved to be rich in flavones, for instance quercetin-O-glucoside (6) and luteolin-O-rutinoside (8) (Figure 4), which confirms the correlation of the observed antimicrobial activity, as well the EE. Gram-negative bacteria, which are responsible for serious opportunistic infections and are resistant to common therapies. In this sense, the study of plants with high flavonoid content should be highlighted [57]. Thus, our study focused on the ethyl acetate phase (MIC 156.2 µg·mL −1 ) against K. pneumoniae, which stands out for its ability to develop enzymatic resistance mechanisms and is considered to be largely responsible for several infectious diseases [58]. The PhEA proved to be rich in flavones, for instance quercetin-Oglucoside (6) and luteolin-O-rutinoside (8) (Figure 4), which confirms the correlation of the observed antimicrobial activity, as well the EE. Our chemical prospecting data summarize chemical constituents that have in common a structural core of flavones, well known for a variety of activities [25]. Furthermore, isolated flavonols such as quercetin and kaempferol already show promising results as antimicrobials [59] as well against K. pneumoniae (MIC > 256 µg·mL −1 ) [60] and are the base structures of the main protagonists of the PhEA in this study. We emphasize carefully that the nominal results have a much more expressive value than many works given the same biological answer; perhaps the answer may be even more significant if the studies aim at obtaining isolated compounds. However, we emphasize that the rapid annotation provided using the LC-MS/MS technique does not require isolation, but shows an understanding of the active extension of the extract, directing more objective studies to that specific class. In addition, the literature [61] treats enriched phases as the main drivers for more specific studies, not to reach the main compounds  (6,8). The color of metabolite nodes is defined by cluster score (0.5-0.75). The size corresponds to the activity score (normalized data).
Our chemical prospecting data summarize chemical constituents that have in common a structural core of flavones, well known for a variety of activities [25]. Furthermore, isolated flavonols such as quercetin and kaempferol already show promising results as antimicrobials [59] as well against K. pneumoniae (MIC > 256 µg·mL −1 ) [60] and are the base structures of the main protagonists of the PhEA in this study. We emphasize carefully that the nominal results have a much more expressive value than many works given the same biological answer; perhaps the answer may be even more significant if the studies aim at obtaining isolated compounds. However, we emphasize that the rapid annotation provided using the LC-MS/MS technique does not require isolation, but shows an understanding of the active extension of the extract, directing more objective studies to that specific class. In addition, the literature [61] treats enriched phases as the main drivers for more specific studies, not to reach the main compounds responsible for the activity but to exclude those who may be acting as deterrents of the activity. In this sense, we are not being categorical in pointing only to these substances as active, but we are presenting a more interference-free sample.

Conclusions
This study showed the extract and phase ethyl acetate from M. Charantia leaves as an antibacterial agent. A total of 32 major compounds were detected, and, of these, 25 were annotated based on mass spectrometry data. Compounds including flavones, terpenes, organic acids, and inositol pyrophosphate derivatives are reported for the first time for the genus Momordica. The ethanolic extract exhibited low activity against Proteus mirabilis and Klebsiella pneumoniae. However, the phase ethyl acetate enriched with flavones showed interesting antibacterial inhibition against K. pneumoniae. Hence, we show that the leaves are a renewable antibacterial source and can serve as a pillar for future studies.