Profiling and cheminformatics bioprospection of curcurbitacin I and momordin Ic from Momordica balsamina on α-amylase and α-glucosidase

Abstract

Momordica spp. has been traditionally used to manage type 2 diabetes mellitus, but the mechanisms and metabolites remain unclear. This study evaluated the inhibitory potential of Momordica balsamina extracts on α-amylase and α-glucosidase in vitro, identifying cucurbitacin I and momordin Ic via high-performance liquid chromatography-photo diode array, and their inhibitory potential in silico. Ethyl acetate seed extract (14.46 µg/ml) and hexane fruit flesh extract (16.79 µg/ml) exhibited lower IC₅₀ values against α-amylase and α-glucosidase, respectively, compared to acarbose (reference standard). Comparatively, momordin Ic concentrations (36.57–605.98 µg/ml) were higher than cucurbitacin I (17.08–44.34 µg/ml). A 140 ns simulation showed that cucurbitacin I (−63.06 kcal/mol) and momordin Ic (−66.53 kcal/mol) exhibited stronger binding to α-amylase than acarbose (−36.46 kcal/mol), whereas cucurbitacin I (−38.08 kcal/mol) and momordin Ic (−54.87 kcal/mol) displayed weaker binding to α-glucosidase, relative to acarbose (−63.73 kcal/mol). Generally, momordin Ic demonstrated better thermodynamic properties, hence further in vitro and in vivo studies are needed to validate their antidiabetic potential.

GRAPHICAL ABSTRACT

Introduction

Diabetes mellitus is a major contributor of global morbidity and mortality with neuropathy, nephropathy, retinopathy and cardiovascular diseases identified as chronic complications¹. In 2021, the disease was reported to affect 537 million adults worldwide, with an estimated increase of 643 million by 2045². This metabolic disorder occurs from elevated amounts of glucose in the bloodstream (hyperglycaemia) caused by insufficient insulin production by pancreatic β-cells and a reduced sensitivity of insulin-responsive tissues to insulin³. The glycaemic index and digestibility of carbohydrates impacts the human glycaemic response; hence, dietary regimens low in sugars/starch are recommended to prevent the dangers of glucose metabolism. Alternatively, the inhibition of carbohydrate metabolising enzymes can be used to reduce high postprandial glucose levels^4,5.

Notably, α-amylase and α-glucosidase have been recognised as the main enzymes implicated in postprandial hyperglycaemia⁶. Carbohydrate digestion begins in the mouth, where saliva containing α-amylase acts as a starch-hydrolyzing enzyme, catalysing the breakdown of polymeric substrates into short oligomers during the hydrolysis of α-1,4-glucan linkages^7,8. Thereafter, the disaccharide carbohydrate units are further broken down into simple monosaccharides by α-glucosidase located at the intestinal mucosal cells brush border. This enzyme facilitates the breakdown of dietary starches into smaller, absorbable glucose units^6,7,9, leading to a high glycaemic index due to rapid intestinal absorption. Consequently, α-amylase and α-glucosidase inhibition can slow carbohydrate digestion and reduce glucose absorption¹⁰.

Unfortunately, the use of conventional α-amylase and α-glucosidase inhibitors (acarbose, miglitol, and voglibose) to manage postprandial hyperglycaemia has been undermined by their drawbacks viz. high cost, long-term regimens accompanying side effects (gastrointestinal discomfort, cardiovascular complications, and weight gain/loss)^11,12, hence the need for more effective therapeutics. Interestingly, medicinal plants have gained global acceptance as complementary therapeutics with conventional drugs due to their availability, affordability and safety profiles¹³. Specifically, Momordica charantia, Momordica balsamina, and Momordica foetida Schumach. are the most common M. spp. used in traditional medicine, with antidiabetic potentials supported by various scientific studies^14–16. Furthermore, M. balsamina and M. foetida Schumach extracts exhibited a potential antidiabetic effect via the inactivation of enzymes responsible for carbohydrate digestion and the prevention of oxidative stress¹⁷. Unfortunately, the mechanism of action responsible for this activity remains unknown despite numerous studies reporting their antidiabetic potential.

Cucurbitacin I, characterised as a tetracyclic triterpenoid, has been noted to exhibit antipyretic, analgesic, anti-inflammatory, antimicrobial, anti-tumour, and antidiabetic properties^18,19. Furthermore, cucurbitacin I has been detected in M. balsamina L. and M. charantia; which have exhibited anti-inflammatory, antioxidative, and antidiabetic properties^20,21. Specifically, momordin Ic categorised as pentacyclic triterpenoid has been found to inhibit ethanol-induced gastric mucosal lesions, exhibit an antioxidant capacity, accelerate gastrointestinal transit, alleviate carbon tetrachloride-induced hepatotoxicity, prevent an increase in blood sugar levels and has also been identified as an effective anticancer candidate²². Despite being present in M. spp., the exact contributions of cucurbitacin I and momordin Ic to the plant’s antidiabetic effects are not well-documented. Hence, this study examined the antidiabetic potential of M. balsamina in vitro while the presence of cucurbitacin I and momordin Ic in the plant extracts was quantified using high-performance liquid chromatography-photo diode array (HPLC-PDA). Thereafter, molecular docking was used to examine the interactions between cucurbitacin I/momordin Ic and the investigated enzymes while MD simulation was employed to offer insight into the protein-ligand interactions and the mechanism of action of cucurbitacin I and momordin Ic towards the discovery and development of therapeutic agents.

Materials and methods

Plant preparation and extraction

Leaves and fruits of M. balsamina were obtained from local markets in Chatsworth and Ottawa, Durban, KwaZulu-Natal, South Africa. Prof Himansu Baijnath authenticated the plant materials while Dr Syd Ramdhani deposited the voucher specimen (V. Jaichand and H. Baijnath, No 1) at the Ward Herbarium of the University of KwaZulu-Natal. The fruit was separated into different components (pericarp, seeds, and fruit flesh) while damaged and discoloured leaves and stems were removed. Thereafter, the plant materials were rinsed with tap water, dried in an incubator (30 °C) for seven days, then milled into a powder with an electrical grinder and stored in an airtight container. The Soxhlet extraction method as per Redfern et al.²³ was slightly modified by binding samples in Whatman no 1 filter paper and running the extraction for 4 h. This was adopted for all solvents to avoid bias while examining phytochemical quality across extracts. The resulting extracts (hexane, ethyl acetate and aqueous) were obtained and concentrated accordingly. A Buchi rotary evaporator (60 °C) was used to concentrate hexane and ethyl acetate extracts, followed by air-drying in a dark cupboard while the aqueous extracts were kept overnight at −80 °C and freeze-dried (VirTis SP Scientific).

In vitro antidiabetic inhibition analyses

α-amylase inhibition assay

The methods of Ranilla et al.²⁴ and Moodley et al.²⁵, were adopted for this assay with slight modifications. Crude extracts were suspended in sodium phosphate buffer (SPB) (pH 6.93, 20 mM) with 6 mM NaCl, at varied concentrations of 200, 600, and 1000 µg/ml, after which 1 ml, 1% starch solution (20 mM SPB) was added to the extracts and incubated for 5 min at 25°C. Thereafter, 1 ml α-amylase solution from Aspergillus oryzae, obtained from Sigma-Aldrich (86250), was added and incubated at 25°C for 3 min. A 1 ml aliquot of DNS solution was introduced into the reaction mixture and subsequently heated at 95.5°C for 15 min, cooled to 25°C and followed with the addition of distilled water (9 ml). The mixture was vortexed, and absorbance was read at 540 nm with SPB as the control, while acarbose was used as the positive control. The equation below was used to calculate the inhibition percentage of α-amylase (Supplementary Figure S1 and Figure S2) followed by the estimation of the half-maximal inhibitory concentration (IC₅₀) values from the nonlinear regression curve.

$$\mathrm{\%} \mathit{\text{inhibition}}=\frac{\mathit{\text{Absorbance}}540\left(\mathit{\text{control}}\right)-\mathit{\text{Absorbance}}540\left(\mathit{\text{sample}}\right)}{\mathit{\text{Absorbance}}540\left(\mathit{\text{control}}\right)}\times 100$$

α-glucosidase inhibition assay

The assay was conducted as previously described by Ranilla et al.²⁴ and Moodley et al.²⁵, using acarbose as the positive control. Potassium phosphate buffer (PPB) (0.1 M, pH 6.93, 50 µl) was combined with extracts at various concentrations (200, 600, and 1000 µg/ml, 50 µl) and incubated (10 min at 25°C) with the α-glucosidase enzyme solution from Saccharomyces cerevisiae which was obtained from Sigma-Aldrich (G5003) (100 µl, 1 unit/ml). After preincubation, 50 µl of 5 mM PNPG (0.1 M PPB) was added and incubated for 5 min at 25°C. Absorbance was measured at 405 nm, and the percentage inhibition of α-glucosidase (Figure S3 and Supplementary Figure S4) was calculated using the equation below while the nonlinear regression curve was used to determine the IC₅₀ values.

$$\mathrm{\%} \mathit{\text{inhibition}}=\frac{\mathit{\text{Absorbance}}405\mathrm{ }\left(\mathit{\text{control}}\right)-\mathit{\text{Absorbance}}405\left(\mathit{\text{sample}}\right)}{\mathit{\text{Absorbance}}405\left(\mathit{\text{control}}\right)}\times 100$$

HPLC-PDA for the determination of cucurbitacin I and momordin Ic

The approach of Vilkickyte and Raudone²⁶, was slightly modified for this analysis. A UFLC Shimadzu system with an XBridge C₁₈ column (5 µl 4.6 × 150 mm) was used. Acetonitrile and methanol (10:90 v/v) were used as the mobile phase, with a 15 min isocratic elution. The injection volume was set to 10 µL at a flow rate of 0.2 ml min⁻¹, and the wavelength of the PDA detector remained at 205 nm. The mobile phase was used to dissolve extracts (1 mg/ml) (Supplementary Figure S5), and a standard curve of the pure compounds obtained from Sigma-Aldrich [momordin Ic (PHL83281) and cucurbitacin I (PHL89464)] at different concentrations (50, 100, 250, 500 µg/ml) was conducted (Supplementary Figure S6 and Figure S7). Subsequently, compound concentrations were determined by plotting a standard curve of known concentrations against the area under the graph (Supplementary Figure S8 and Figure S9).

Computational analyses

Metabolites acquisition and preparation

The 3D structures of cucurbitacin I, momordin Ic, and acarbose (reference standard) were downloaded from PubChem [https://pubchem.ncbi.nlm.nih.gov/ (accessed on 01 May 2023)]. The ligands were prepared for molecular docking through the addition of gasteiger charges via the Open Babel plugin on Python Prescription (PyRx) v 0.9.5²⁷.

Protein acquisition and optimisation

The α-amylase (PDB: 4W93) and α-glucosidase (PBD: 2ZQ0) X-ray crystal structures were downloaded from RCSB Protein Data Bank [https://www.rcsb.org/ (accessed on 01 May 2023)], where water molecules and non-standard amino acids were removed²⁸, using the UCSF Chimera software package V1.14²⁹.

Molecular docking and validation

Using the PyRx V 0.9.5, docking of the optimised ligands and prepared protein was conducted by selecting the amino acid residues at the active site of α-amylase³⁰ and α-glucosidase³¹, using the x-y-z coordinates [centre (X: −9.71; Y: 7.99; Z: −21.91)] and [centre (X: 31.20; Y: 49.68; Z: 37.68)], respectively. Thereafter, the redocking technique was applied for the verification of the docking procedure accuracy³², because false-positive binding configurations can be generated when identifying the position with the lowest energy. After optimal superimposition, the root mean square deviation (RMSD) of the docked ligands from the native inhibitor’s reference position in the experimental co-crystal structure of α-amylase (4W93) and α-glucosidase (2ZQ0) was determined. The native inhibitor and docked ligands showed a low RMSD value of 0.50 Å in the crystal structures of α-amylase (Figure 1) and α-glucosidase (Figure 2), indicating a similar alignment in the binding orientation, thus supporting the accuracy and validity of the docking procedure.

Figure 1.

Docking technique validation was conducted via the superimposition approach of the α-amylase (4W93) co-crystal structure where the orientation of (a) the superimposed docked compounds cucurbitacin I (blue) and momordin Ic (yellow) closely matched the native inhibitor (black) of 4W93 with 0.50 Å as the RMSD value (b) the visualisation of the investigated compounds, reference standard and the native inhibitor at the binding pocket of 4W93 while displaying the active site amino acids (Gln63, His101, Arg197, Arg195, Asp300, His305, Asn298, His201, Glu233, Tyr151, Lys200, Glu240, and Ile235)³⁰.

Figure 2.

Docking technique validation were conducted via the superimposition approach of the α-glucosidase (2ZQ0) co-crystal structure where the orientation of (a) the superimposed docked compounds cucurbitacin I (blue) and momordin Ic (yellow) closely matched the native inhibitor (black) of 2ZQ0 with 0.50 Å as the RMSD value (b) the visualisation of the investigated compounds, reference standard and the native inhibitor at the binding pocket of 2ZQ0 while displaying the active site amino acids (His437, Lys467, Glu391, His507, Glu439, Trp331, Glu508, Glu526, Trp341, Glu194, Glu532, Trp400, Ser217, Phe401, Ile335, Tyr533, and Phe536)³¹.

Molecular dynamics (MD) simulation

AMBER 18 suite (graphical processing unit version) of the Centre for High Power Computing, Cape Town, South Africa was used for the MD simulation in which for protein description, the AMBER force field (FF18SB variant) was utilised³³. ANTECHAMBER was used to generate atomic partial charges for the ligands by utilising the Restrained Electrostatic Potential (RESP) and the General Amber Force Field (GAFF) processes. Thereafter, neutralisation of the system was achieved by adding hydrogen atoms, Na⁺ and Cl⁻ counter ions via the Leap module of AMBER 18. Residue numbering was then carried out for α-amylase (1 – 496) and α-glucosidase (1 – 1476). To ensure that the atoms were inside 10 Å of the box edges, the systems were subjected to suspension in an TIP3P orthorhombic box which had been filled with water molecules. Using the SPFP precision model at a step size of 2 fs, the SHAKE algorithm was used to constrict the hydrogen bond atoms in every simulation. After, subjecting the steepest descent technique to both solutes via constraint potential (500 kcal/mol) and minimisation (2000-step), conjugate gradients (1000 steps) were applied to the system. An additional full minimisation (990 steps) without restraint of the conjugate gradient procedure was conducted.

While maintaining the number and volume of atoms in the system, the temperature of the MD simulations was increased gradually (from 0K to 300K) for 50ps. Once heated, the working temperature was kept at 300K and equilibration (500ps) was applied to the systems. Notably, all simulations followed the isobaric-isothermal ensemble (NPT); viz. 2 ps pressure-coupling constant, random seeding, Langevin thermostat (1.0 ps collision frequency), 300 K temperature and Berendsen barostat (1 bar constant pressure)³⁴. Using AMBER 14 (PTRAJ system), the system coordinates obtained from the MD simulations were saved, and at intervals of 1 ps trajectory analysis was conducted. This was followed by the analysis of root mean square deviation (RMSD), root mean square fluctuation (RMSF), surface area solvent accessibility (SASA) and radius of gyration (ROG). Similarly, using the Molecular Mechanics/Generalised Born Surface Area method (MM/GBSA)³⁵, the binding free energy was calculated using an average among 140 000 snapshots extracted from the 140 ns MD simulation trajectory. Thereafter, the binding free energy (ΔG) of each molecular species (complex, ligand, and receptor) was computed as detailed in the corresponding equations below:

$$\Delta {\mathrm{G}}_{\mathrm{\text{bind}}}={\mathrm{G}}_{\mathrm{\text{complex}}}-{\mathrm{G}}_{\mathrm{\text{receptor}}}-{\mathrm{G}}_{\mathrm{\text{ligand}}}$$ (1)

$$\Delta {\mathrm{G}}_{\mathrm{\text{bind}}}={\mathrm{E}}_{\mathrm{\text{gas}}}+{\mathrm{G}}_{\mathrm{\text{sol}}}‐\mathrm{\text{TS}}$$ (2)

$${\mathrm{E}}_{\mathrm{\text{gas}}}={\mathrm{E}}_{\mathrm{\text{int}}}+{\mathrm{E}}_{\mathrm{\text{vdw}}}+{\mathrm{E}}_{\mathrm{\text{ele}}}$$ (3)

$${\mathrm{G}}_{\mathrm{\text{sol}}}={\mathrm{G}}_{\mathrm{\text{GB}}}+{\mathrm{G}}_{\mathrm{\text{SA}}}$$ (4)

$${\mathrm{G}}_{\mathrm{\text{SA}}}{=}_{\text{ϓ}}\mathrm{\text{SASA}}$$ (5)

Pharmacokinetics and toxicity studies

The drug-likeness, pharmacokinetics and physicochemical features of the selected triterpenoids were predicted through the SwissADME web (http://swissadme.ch/index.php, accessed on 06 April 2024), while their toxicological properties were predicted using the Protox II webserver (https://tox-new.charite.de/protox_II/, accessed on 06 April 2024).

Results and discussion

In vitro antidiabetic inhibition analyses

Alpha-amylase and α-glucosidase break down carbohydrates into absorbable monosaccharides, raising post-prandial blood glucose levels which contributes significantly to the development of type 2 diabetes mellitus (T2DM)³⁶. Therefore, inhibiting these enzymes can reduce postprandial blood glucose by impeding the breakdown of carbohydrates, delaying glucose absorption and ultimately lowering blood sugar levels^37,38, making them therapeutic targets for drug development towards diabetes management. In this study, the observed IC₅₀ values of hexane and ethyl acetate for each plant part were similar across both solvent systems, for the inhibition of α-amylase. This is evident in the IC₅₀ values obtained for hexane (16.79–17.97 µg/ml) and ethyl acetate (17.62–19.09 µg/ml) extracts; with aqueous extracts (25.13–66.10 µg/ml) having the highest values. In addition, the IC₅₀ values obtained for fruit flesh (ethyl acetate), pericarp (hexane and ethyl acetate), seeds (hexane and ethyl acetate), and leaves (hexane) extracts were lower than acarbose (26.35 µg/ml). Notably, the IC₅₀ values of ethyl acetate seed (17.46 µg/ml) and hexane leaves (17.86 µg/ml) extracts can be considered the most promising, as these extracts had the lowest inhibitory concentration (50%) values for α-amylase (Table 1).

Table 1.

IC₅₀ values for the inhibitory activity of various Momordica balsamina extracts and acarbose against α-amylase and α-glucosidase.

IC50 (µg/ml)
	Inhibitor	α-amylase	α-glucosidase
Positive control	Acarbose	26.35	41.41
Plant part	Solvent
Fruit flesh
	Hex	37.78	16.79
	EA	24.56	18.24
	AQ	37.14	54.52
Pericarp
	Hex	19.04	17.97
	EA	18.61	17.62
	AQ	135.20	25.13
Seeds
	Hex	20.43	17.11
	EA	17.46	18.74
	AQ	79.22	66.10
Leaves
	Hex	17.86	17.81
	EA	32.18	19.09
	AQ	94.32	25.58

Note: Hex: hexane; EA: ethyl acetate; AQ: aqueous.

For the inhibition of α-glucosidase, the lowest IC₅₀ value (16.79 µg/ml) was observed in the hexane fruit flesh extracts of M. balsamina, indicating that, compared to the other extracts, it exhibited the strongest inhibitory effect. Also, all hexane and ethyl acetate extracts had similar IC₅₀ values which were lower than that of acarbose (41.41 µg/ml), for α-glucosidase and further corroborate the antidiabetic properties of M. balsamina. Conversely, the aqueous extracts had higher IC₅₀ values relative to acarbose, hexane, and ethyl acetate extracts for both α-amylase and α-glucosidase; hence, aqueous extracts would not be ideal inhibitors of the investigated enzymes. The observed inhibitory activity of these extracts could be attributed to their constituent phytochemicals which is a function of different factors such as the plant part (leaves, flower, fruit, bark, stem, seed, root), its origin (moist or arid region), solvent polarity and the method of extraction, which may altogether influence the chemical constituents of each extract. Agu et al.³⁹, reported a higher IC₅₀ value for Annona muricata methanolic (1.85–2.18 mg/dL), dichloromethane (2.13–8.53 mg/dL) and ethyl acetate (2.20–4.41 mg/dL) plant extracts against α-amylase, relative to acarbose (1.72 mg/dL). However, the IC₅₀ value of Adenanthera pavonina methanolic leaf extracts (16.16 μg/ml) was on par with M. balsamina pericarp (hexane and ethyl ­acetate), seeds (ethyl acetate), and leaves (hexane) for the inhibition of α-amylase⁴⁰. On the other hand, Alqahtani et al.¹⁰ and Chokki et al.⁴¹ reported that both dichloromethane (moderately polar) and ethyl acetate (moderately polar) extracts, and a n-hexane (non-polar) fraction effectively inhibited α-amylase and α-glucosidase at lower concentrations than methanol and ethanol-water extracts (polar). Thus, the traditional administration of Momordica leaf as a tea or decoction^42,43 may not be optimal for the inhibition of α-amylase and α-glucosidase. This suggests that the selection of an extraction solvent is crucial, as it determines the identified phytochemicals and their corresponding inhibitory potential. Given the vast ­number of plant metabolites, understanding the synergistic antidiabetic effect of these compounds could be complex and time-consuming, thus necessitating individual ­compound analysis⁴⁴.

HPLC-PDA profiling of cucurbitacin I and momordin Ic in Momordica balsamina

The standards obtained from Sigma-Aldrich were used to establish retention times for the identification and quantification of cucurbitacin I (7.20 min) and momordin Ic (8.00 min). The chromatograms representing all triterpenoids present in M. balsamina specific for the hexane extracts, are shown in Supplementary Figure S3. Ethyl acetate fruit flesh (44.34 µg/ml) and ethyl acetate leaf (17.08 µg/ml) extracts contained the highest and lowest concentrations of cucurbitacin I, respectively; with hexane fruit flesh (39.56 µg/ml) and pericarp ethyl acetate (34.80 µg/ml) extracts having a relatively small difference. Notably, hexane pericarp and seeds had similar cucurbitacin I concentrations (0.87 µg/ml), while the fruit flesh extracts contained a higher concentration of cucurbitacin I than other extracts. The non-detection cucurbitacin I in the aqueous and ethyl acetate seed extracts (Table 2) may be because of the nucleus’s hydrophobic nature, impairing its water solubility²⁰. On the contrary, the cucurbitacin I concentration (5150.00 mg/kg) of Citrullus colocynthis⁴⁵ was lower than those obtained for all the extracts examined in this study; however, concentrations of cucurbitacin I in ripe and unripe fruits of Blastania cerasiformis and B. garcinii leaves and unripe fruits⁴⁶ were higher than those quantified in all the extracts investigated in this study (Table 2).

Table 2.

Concentration of cucurbitacin I in Momordica balsamina plant portal extracts.

Plant part	Extracts	Area under graph (mV*min)	Concentration (µg/ml)
Fruit flesh	Hex	2 303 428	39.56
	EA	2 695 451.50	44.34
	AQ	–	–
Pericarp	Hex	772 151.67	20.86
	EA	1 913 644.67	34.80
	AQ	–	–
Seeds	Hex	843 705.50	21.73
	EA	–	–
	AQ	–	–
Leaves	Hex	669 438.00	19.60
	EA	462 819.00	17.08
	AQ	–	–

Note: Hex: hexane; EA: ethyl acetate; AQ: aqueous; – : absence of cucurbitacin I in crude extracts.

On the other hand, momordin Ic was consistently detected across all crude extracts of M. balsamina, with the highest concentrations observed in aqueous pericarp (605.98 µg/ml), ethyl acetate seed (583 µg/ml) and aqueous seed (514.61 µg/ml) extracts. Comparatively, hexane fruit flesh (110.55 µg/ml) and ethyl acetate pericarp (110.99 µg/ml) extracts showed nearly identical concentrations, differing by only 0.44. However, hexane pericarp (36.57 µg/ml) and hexane leave (39.67 µg/ml) extracts had the lowest momordin Ic concentration (Table 3). Generally, momordin Ic was found in all M. balsamina extracts and was present in higher concentrations, when compared to cucurbitacin I. (Table 3), likely due to its moderately non-polar nature⁴⁷. Specifically, the greatest concentration of momordin Ic was obtained in aqueous pericarp extracts (605.98 µg/ml), whereas the lowest concentration was observed in hexane pericarp extracts (36.57 µg/ml). Notably, all aqueous (polar) and ethyl acetate (moderately polar) extracts had the highest concentrations of momordin Ic; validating momordin Ic from M. balsamina extracts as characteristically polar. However, both hexane fruit flesh (110.55 µg/ml) and ethyl acetate pericarp (110.99 µg/ml) extracts had similar concentrations of momordin Ic compared to the lowest concentration obtained in the aqueous extracts (fruit flesh 137.76 µg/ml). Interestingly, despite its potential, charantin (more potent than a standard oral hypoglycaemic drug, tolbutamide)⁴⁸ obtained from ethanolic M. charantia fruit extract (29.17 µg/ml) was lower than momordin Ic concentrations obtained in all M. balsamina extracts elucidated in this study^49,50. Also, Liu et al.⁵¹, reported concentrations of momordicine I (458.78 μg/g) and momordicoside L (420.94 μg/g) from M. charantia L. to be slightly lower than aqueous and ethyl acetate seed extracts (Table 3). Consequently, the absence of cucurbitacin I in all aqueous extracts relative to the higher IC₅₀ values (Table 1), suggests cucurbitacin I may contribute to the antidiabetic potential of M. balsamina extracts, further validating the need to isolate and verify the inhibitory potential of cucurbitacin I and momordin Ic against α-amylase and α-glucosidase. To curb the time-consuming and expensive process of isolating and purifying compounds, computational techniques were employed to assess the potential of momordin Ic and cucurbitacin I as inhibitors of α-amylase and α-glucosidase. Hence, to provide insights into the binding affinity and compound-enzyme interactions, molecular docking and MD simulations were conducted.

Table 3.

Concentration of momordin Ic in Momordica balsamina plant extracts.

Plant part	Extracts	Area under graph (mV*min)	Concentration (µg/ml)
Fruit flesh	Hex	2 693 824.00	110.55
	EA	5 503 888.00	218.55
	AQ	3 401 770.00	137.76
Pericarp	Hex	768 929.50	36.57
	EA	2 705 195.00	110.99
	AQ	15 584 799.50	605.98
Seeds	Hex	1 213 058.67	53.64
	EA	15 005 034.50	583.69
	AQ	13 207 448.67	514.61
Leaves	Hex	849 556.50	39.67
	EA	1 411 600.33	61.27
	AQ	5 768 197.50	228.70

Note: Hex: hexane; EA: ethyl acetate; AQ: aqueous.

Molecular docking

Molecular docking facilitates the determination of the ligand’s conformation, orientation and position, with a higher negative docking score indicating a better pose⁵². The selected cucurbitacin I and momordin Ic among the identified metabolites had notably higher negative docking scores than acarbose against both enzymes (Table 4). The observed higher negative docking scores could indicate better interactions and orientation of both compounds at the active site of the investigated enzymes⁵³, relative to acarbose and could suggest their better potential as inhibitors of both α-amylase and α-glucosidase.

Table 4.

Molecular docking scores (kcal/mol) of cucurbitacin I and momordin Ic identified in Momordica balsamina with carbohydrate metabolising enzymes.

Ligands	Docking score (kcal/mol)
α-amylase
Cucurbitacin I	−9.50
Momordin Ic	−10.10
Acarbose	−7.20
α-glucosidase
Cucurbitacin I	−10.40
Momordin Ic	−10.00
Acarbose	−7.80

Thermodynamic energy components of cucurbitacin I and momordin Ic against the target enzymes

The MMGBSA method is often used to determine the binding affinity between ligands and protein. This approach calculates the overall energy of a protein, ligand, and their combined structure⁵⁴. Thermodynamic binding free energy represents the energy gap between a ligand-protein complex and its unbound parts, where values that are more negative points to a more robust interaction and a higher affinity of the ligand for the protein⁵⁵. In this study, a higher negative binding free energy was observed in the momordin Ic-α-amylase complex (−66.53 kcal/mol) and cucurbitacin I-α-amylase complex (−63.06 kcal/mol) relative to the acarbose-α-amylase complex (−36.46 kcal/mol) (Table 5), suggesting that cucurbitacin I and momordin Ic may exhibit a stronger inhibitory activity against α-amylase compared to acarbose. In contrast, a higher negative binding free energy was observed in the α-glucosidase-acarbose complex (−63.73 kcal/mol) compared to α-glucosidase-momordin Ic (−54.87 kcal/mol) and α-glucosidase-cucurbitacin I (−38.08 kcal/mol) complexes (Table 5); suggesting the higher inhibitory potential of the standard in comparison to the investigated compounds. The results of this study are partially similar to that of Ahmed et al.⁵⁶, where apigenin-7-O-glucoside had a higher negative free binding energy than acarbose on binding to both α-amylase (−45.02 kcal/mol) and α-glucosidase (−38.28 kcal/mol). Conversely, Imtiaz et al.⁵⁷, reported that acarbose-α-glucosidase had a higher negative free binding energy (−50.91 kcal/mol) than catechin-α-glucosidase (−31.070 kcal/mol) and syringic acid-α-glucosidase (−31.070 kcal/mol) for α-glucosidase which is in tandem to this study.

Table 5.

MMGBSA-based binding free energy outline (kcal/mol) of cucurbitacin I and momordin Ic against α-amylase and α-glucosidase.

Energy components (kcal/mol)
Complex	ΔEvdW	ΔEelec	ΔGgas	ΔGsolv	ΔGbind
α-amylase
Cucurbitacin I	−52.89 ± 4.42	−56.51 ± 5.97	−109.41 ± 7.68	46.34 ± 3.93	−63.06 ± 5.58
Momordin Ic	−70.39 ± 4.34	−46.27 ± 12.14	−119.51 ± 10.64	52.98 ± 4.58	−66.53 ± 9.63
Acarbose	−47.71 ± 4.04	−157.26 ± 26.70	−204.97 ± 26.73	168.51 ± 19.47	−36.46 ± 9.15
α-glucosidase
Cucurbitacin I	−45.50 ± 3.63	−45.15 ± 7.51	−90.64 ± 7.17	52.56 ± 6.42	−38.08 ± 4.45
Momordin Ic	−61.31 ± 4.32	−32.33 ± 9.94	−97.18 ± 10.29	42.31 ± 7.02	−54.87 ± 5.30
Acarbose	−45.09 ± 7.06	−457.84 ± 20.23	−502.93 ± 19.21	439.20 ± 15.97	−63.73 ± 7.63

Note: ΔEvdW: van der Waals energy; ΔEele: electrostatic energy; ΔEgas: gas-phase free energy; ΔGsol: solvation free energy; ΔGbind: total binding free energy.

Stability, flexibility, compactness, and H-bonds analysis of the carbohydrate metabolising enzymes bound systems

Enzyme-ligand complexes often undergo conformational changes due to ligand binding, leading to potential alterations in the enzyme’s biological activities²⁸. To assess the stability, flexibility, compactness and H-bonds formed in the enzyme-ligand complex, parameters such as RMSD, RMSF, ROG, SASA, and number of H-bonds were evaluated after a 140 ns simulation. The RMSD quantifies the deviation over time between a bound structure and its unbound protein, while lower RMSD values suggest better complex stability⁵⁸. The ligand-α-amylase complexes and the apo-protein remained relatively stable after reaching equilibrium within the first 10 ns of the simulation. However, the apo-protein, α-amylase exhibited significant swaying between 25 and 65 ns (Figure 3(a)). Remarkably, a lower mean RMSD value relative to the apo-protein, α-amylase (1.70 Å) was observed in momordin Ic-α-amylase (1.30 Å), cucurbitacin I-α-amylase (1.39 Å) and the reference standard, acarbose-α-amylase (1.31 Å) (Table 6). This suggests that ligand binding leads to increased structural stability⁵⁹, of the α-amylase-complexes relative to the unbound α-amylase.

Figure 3.

Comparative root mean square deviation (RMSD) plots of α-carbon, cucurbitacin I, momordin Ic, and standard drug (acarbose) against (a) α-amylase and (b) α-glucosidase after 140 ns MD simulation period.

Table 6.

Post-MD simulation parameters of cucurbitacin I and momordin Ic against α-amylase and α-glucosidase.

Ligands	Mean RMSD (Å)	Mean ROG (Å)	Mean RMSF (Å)	Mean SASA (Å)	Mean number of H-bonds
α-amylase
Cucurbitacin I	1.39 ± 0.17	23.22 ± 0.07	0.96 ± 0.47	17 143.70 ± 481.06	266.71 ± 11.14
Momordin Ic	1.30 ± 0.09	23.12 ± 0.09	0.93 ± 0.44	16 635.97 ± 400.66	272.74 ± 10.67
Acarbose	1.31 ± 0.08	23.16 ± 0.06	0.88 ± 0.39	17 174.93 ± 434.43	267.84 ± 10.63
α-amylase	1.70 ± 0.18	23.44 ± 0.08	0.97 ± 0.51	17 712.42 ± 458.99	258.02 ± 10.63
α-glucosidase
Cucurbitacin I	1.42 ± 0.19	26.19 ± 0.09	0.96 ± 0.64	22 803.64 ± 457.31	403.36 ± 13.06
Momordin Ic	1.49 ± 0.14	26.16 ± 0.06	0.91 ± 0.56	22 266.15 ± 359.35	410.30 ± 12.63
Acarbose	1.53 ± 0.15	26.19 ± 0.07	0.95 ± 0.58	22 950.90 ± 432.39	404.78 ± 12.49
α-glucosidase	1.52 ± 0.19	26.19 ± 0.07	0.97 ± 0.57	22 935.98 ± 464.66	401.46 ± 12.66

Note: RMSD: root mean square deviation; RMSF: root mean square fluctuation; ROG: radius of gyration; SASA: solvent accessible surface area; H-bonds: hydrogen bonds.

On the contrary, a few deviations from the apo-protein (α-glucosidase) were noted in the bound α-glucosidase-complexes for both compounds and the standard throughout the simulation (Figure 3(b)). Notably, the cucurbitacin I-α-glucosidase complex exhibited superior structural stability over the first 70 ns, compared to the momordin Ic-α-glucosidase and acarbose-α-glucosidase complexes. This could have contributed to the lower mean RMSD value obtained for the cucurbitacin I-α-glucosidase complex (1.42 Å) compared to momordin Ic-α-glucosidase (1.49 Å), acarbose-α-glucosidase (1.53 Å) and the apo-protein, α-glucosidase (1.52 Å). Ultimately, the mean RMSD values for the bound ­complexes suggest that ligand binding resulted in better structural stability. Sabiu et al.²⁸, reported a similar trend where procyanidin-α-amylase and rutin-α-amylase exhibited lower RMSD values than the apo-protein (α-amylase). Overall, this suggests enhanced stability of bound compounds, compared to the apo-proteins. Remarkably, in this study, the RMSD values for all systems (bound and unbound) (Table 6) were within 3 Å (acceptable limit)⁶⁰, indicating that the bound complexes formed with momordin Ic and cucurbitacin I were stable after binding with both α-amylase and α-glucosidase.

Furthermore, ROG indicates the compactness or spatial distribution of a molecule or a group of atoms within a complex⁶¹, with lower ROG values demonstrating compactness of the complex. In this study, equilibrium was reached for all complexes between 0 and 10 ns; however, the α-amylase complex slightly unfolded between 20 and 80 ns compared to the other bound complexes (Figure 4(a)). This observation suggests that ligand binding resulted in increased compactness of the complexes relative to the unbound protein. Altogether, the marginal difference between the mean ROG values obtained for momordin Ic-α-amylase (23.12 Å), cucurbitacin I-α-amylase (23.22 Å), acarbose-α-amylase (23.16 Å) and the apo-protein, α-amylase (23.44 Å) (Table 6) indicates similar structural stability with little folding of the bound compounds relative to the apo-protein and consistent with their binding free energy which further signifies the potential of the investigated ligands at modulating α-amylase.

Figure 4.

Comparative radius of gyration (ROG) plots of α-carbon, cucurbitacin I, momordin Ic, and standard drug (acarbose) against (a) α-amylase and (b) α-glucosidase after 140 ns MD simulation period.

Overall, there were obvious minimal folding’s between the bound ligand systems and the unbound apo-protein, α-glucosidase. During the simulation, the cucurbitacin I-α-glucosidase system initially diverged to 26.36 Å, reaching its peak at 26.45 Å. In contrast, the momordin Ic-α-glucosidase complex had diverged at 90 ns, with no major folding thereafter (Figure 4(b)). An equal average ROG value of 26.19 Å was observed in all complexes against α-glucosidase except in momordin Ic-α-glucosidase (26.16 Å) with a lower value, but a negligible difference of 0.03 Å; indicating that the compactness of the examined target was unaffected by the ligand binding (Table 6). Similarly, Sharma et al.⁶² reported a similar trend where the resulting complexes of quercetin, ellagic acid and luteolin upon binding with α-amylase have similar ROG value with the apo-α-amylase. This disagrees with the momordin Ic-α-amylase and cucurbitacin I-α-amylase complexes, which exhibited higher compactness (lower ROG values) than the apo-protein (Table 6). Moreover, Balogun et al.⁶³ also observed that α-glucosidase systems except benzoic acid-α-glucosidase had slightly lower ROG values than the apo-α-glucosidase and reference standard complex.

Additionally, RMSF analysis was conducted to assess the flexibility of amino acid residues at the active site of the target, as a lower average RMSF value indicates restricted motion, thereby improving complex stability⁶⁴. In this study, fewer fluctuations were observed between residues 0–300 for all α-amylase bound systems compared to residues 300–500 (Figure 5(a)). Remarkably, the catalytic residues of α-amylase, Glu233, Asp197, and Asp300⁶⁵, were situated in the region of minor fluctuations in the α-amylase systems, indicating strong binding interactions at the catalytic site of the α-amylase complexes. However, the unbound α-amylase complex exhibited higher fluctuations relative to the bound systems. Similarly, the cucurbitacin I-α-amylase complex showed increased fluctuations at residues 0, 60, 130, 260, 370, 400, and 470 throughout the simulation, relative to the momordin Ic-α-amylase system. Conversely, lesser fluctuations were observed in the acarbose-α-amylase complex throughout the simulation (Figure 5(a)), hence the lowest mean RMSF value (0.88 Å) relative to momordin Ic-α-amylase (0.93 Å), cucurbitacin I-α-amylase (0.96 Å) and the apo-α-amylase (0.97 Å) (Table 6). This suggests that there is a stronger binding interaction between acarbose and α-amylase, leading to less movement or fluctuation in specific regions of the protein during the simulation.

Figure 5.

Comparative root mean square fluctuation (RMSF) plots of α-carbon, cucurbitacin I, momordin Ic, and standard drug (acarbose) against (a) α-amylase and (b) α-glucosidase after 140 ns MD simulation period.

Furthermore, higher fluctuations between residues 60–70, 340–360, 390–420, and 520–540 were observed in the α-glucosidase system (Figure 5(b)), indicating the reduced ability of the residues to form stable bonds. Of the catalytic residues in α-glucosidase (Glu532, Glu508, and Glu439)³¹, residues 439 and 508 are situated in the region of minor fluctuations; hence, they could be responsible for the intra and intermolecular bonding at the catalytic site of the protein during the simulation. The bound momordin Ic-α-glucosidase (0.91 Å) complex exhibited a lower RMSF value than both the cucurbitacin I-α-glucosidase (0.96 Å), acarbose-α-glucosidase (0.95 Å) complex and the apo-protein, α-glucosidase (0.97 Å) (Table 6), suggesting the momordin Ic-α-glucosidase complex to be more stable and thus exhibits a stronger and more effective binding interaction than the other bound complexes. Contrarily, Sabiu et al.²⁸, reported that all bound α-amylase and α-glucosidase complexes, except luteolin-7-O-beta-d-glucoside-α-glucosidase had higher average RMSF values than the apo-proteins. Hence, from the findings of this study, momordin Ic could be a better inhibitor against α-glucosidase relative to cucurbitacin I and acarbose.

Furthermore, SASA investigates surface area changes and protein folding⁶⁶, by determining the level of exposure of specific regions of the protein to the surrounding solvent. This facilitates the evaluation of protein-ligand binding and solvation effects on the interactions⁶⁷. Notably, throughout the simulation, the investigated compounds complexed with α-amylase resulted in lower SASA values compared to the apo-α-amylase (Figure 6(a)). Specifically, the momordin Ic-α-amylase complex (16 635.97 Å) exhibited the lowest SASA value, followed by the cucurbitacin I-α-amylase (17 143.70 Å) and acarbose-α-amylase (17 174.93 Å) complex for α-amylase (17 712.42 Å). This suggests that both bound compounds reduce protein exposure to the solvent than the reference standard complex (acarbose-α-amylase). This is significant because a lower SASA value implies reduced solvent exposure, suggesting a more stable and tightly bound interaction.

Figure 6.

Comparative solvent-accessible surface area (SASA) plots of α-carbon, cucurbitacin I, momordin Ic, and standard drug (acarbose) against (a) α-amylase and (b) α-glucosidase after 140 ns MD simulation period.

Furthermore, fewer fluctuations were observed in the α-glucosidase bound complexes and might account for the lower mean SASA values in the momordin Ic-α-glucosidase (22 266.15 Å), cucurbitacin I-α-glucosidase (22 803.64 Å) complexes, and the unbound apo-protein (22 935.98 Å), as opposed to the acarbose-α-glucosidase complex (22 950.90 Å) (Table 6, Figure 6(b)). This suggests that the binding of momordin Ic and cucurbitacin I to α-glucosidase results in less solvent interaction than the acarbose-α-glucosidase complex. Conversely, Sharma et al.⁶², reported that the acarbose-α-glucosidase complex had the lowest mean SASA value, which contradicted the data obtained in this study. Therefore, the lower SASA values observed in the study upon the binding of momordin Ic and cucurbitacin I to the investigated enzymes suggest lower exposure of the protein to the surrounding solvent and improved stability than the unbound apo-proteins⁶¹. Moreover, the assessment of H-bonds in MD simulations is vital since it determines the strength, specificity, and stability of the H-bond interactions, formed during the protein-ligand binding⁶⁸. The momordin Ic-α-amylase and acarbose-α-amylase complexes remained relatively consistent through the 140 ns simulation however, there were multiple fluctuations observed in the cucurbitacin I-α-amylase complex (Figure 7(a)). The highest mean number of H-bonds was observed in the momordin Ic-α-amylase complex (272.74), which was higher than the cucurbitacin I-α-amylase (266.71) and acarbose-α-amylase (267.84) complexes. Also, all bound α-amylase complexes exhibited more H-bonds compared to the apo-α-amylase (258.02) complex, implying an improved stability of all bound complexes (Table 6).

Figure 7.

Time progression of the number of hydrogen bonds in (a) α-amylase and (b) α-glucosidase upon binding with cucurbitacin I, momordin Ic, and standard drug (acarbose) during the 140 ns MD simulation period.

Interestingly, a similar trend was observed in the α-glucosidase systems, in which the highest amount of H-bonds was observed in momordin Ic-α-glucosidase (410.30), followed by acarbose-α-glucosidase complex and the cucurbitacin I-α-glucosidase complex (403.36), with the apo-α-glucosidase complex (404.78) having the least number of hydrogen bonds (Table 6, Figure 7(b)). Taken together, the momordin Ic-α-glucosidase and momordin Ic-α-amylase complexes exhibited the highest number of hydrogen bonds against the investigated enzymes. Hence, this suggests that momordin Ic could be a potential drug candidate against the enzymes studied, as the higher number of H-bonds formed with both enzymes reflects a more specific and stronger binding interaction than that of acarbose. In contrast, Ogunyemi et al.⁶⁹ observed fewer mean H-bonds in four steroidal pregnane phytochemicals bound to α-amylase, compared to the unbound protein. Similarly, the top-scoring compound complexes investigated by Kumari et al.⁷⁰ indicated fewer H-bonds than the acarbose-α-amylase complex. However, Ogunyemi et al.⁷¹, reported higher number of H-bonds were observed upon ligand binding relative to the unbound-α-glucosidase. Therefore, momordin Ic and cucurbitacin I complexed with either α-amylase or α-glucosidase enhanced protein stability and conformation by promoting hydrogen bond formation to maintain the geometry of the protein.

Bond analysis of the interaction plots of cucurbitacin I, momordin Ic, and acarbose against α-amylase and α-glucosidase after 140 ns simulation

To examine the interactions responsible for the binding affinity between the ligands and the investigated targets, the interaction plots at 0, 70, and 140 ns of the simulation were elucidated⁷². The time-dependent plot interactions showed the occurrence of multiple bond interactions over the course of the 140 ns simulation such as hydrogen bonds, van der Waals, salt bridge, attractive charges, unfavourable donor-donor bonds, π-alkyl, alky, π-sigma, and π-anion interactions (Figures 8 and 9). Therefore, this study examined these bonds responsible for the binding of cucurbitacin I, momordin Ic, and acarbose to α-amylase and α-glucosidase.

Figure 8.

Interaction types and plots at 0 ns, 70 ns, and 140 ns between α-amylase and (a) acarbose (reference standard), (b) cucurbitacin I, and (c) momordin Ic.

Figure 9.

Interaction types and plots at 0 ns, 70 ns, and 140 ns between α-glucosidase and (a) acarbose (reference standard), (b) cucurbitacin I, and (c) momordin Ic.

Generally, the total number of interactions exhibited by momordin Ic-α-amylase complex, cucurbitacin I-α-amylase complex and acarbose-α-amylase were reflective of the total binding free energy and specifically, the number of Van der Waals interactions were also consistent with their binding free energy (Table 5). Although a longer bond length implying a weaker force between intermolecular or intramolecular systems⁷³, and lower number of hydrogen bonds was observed in the momordin Ic-α-amylase complex (Table 7), a higher number of alkyl/π-alkyl interactions was noted compared to cucurbitacin I-α-amylase and acarbose-α-amylase complexes after 140 ns and could have also been responsible for the observed higher negative binding free energy. Also, the cucurbitacin I-α-amylase complex had a higher number of hydrogen bonds, lower total bond length and no unfavourable donor-donor bonds; it exhibited lower negative binding free energy than the momordin Ic-α-amylase complex. This could be attributed to the higher average hydrogen bond length observed in cucurbitacin I-α-amylase in contrast to the momordin Ic-α-amylase complex. Specifically, the lower binding free energy observed in the acarbose-α-amylase complex as opposed to the momordin Ic-α-amylase and cucurbitacin I-α-amylase complexes might be explained by the lower number of interactions and an unfavourable donor-donor bond present at the beginning of the simulation (Figure 8); as unfavourable bonds indicate the presence of repulsive forces, leading to a lower free binding energy⁷⁴. Therefore, the total number of interactions could be responsible for a higher negative binding free energy observed in the momordin Ic-α-amylase complex relative to the acarbose-α-amylase and cucurbitacin I-α-amylase complex.

Table 7.

The total number of interaction types of cucurbitacin I, momordin Ic, and reference standard (acarbose) against α-amylase and α-glucosidase after 140 ns.

Ligands	Total number of interactions (average distance)	Number of hydrogen bonds (average distance) and interaction residues	Other important interactions (average distance) and interaction residues
α-amylase
Cucurbitacin I	26 (4.95 Å)	6 (4.70 Å) [Arg 194, Asp 299 (2), Asp 196, Hie 304, Hie 298]	6 (5.20 Å) [Phe255, Ile234 (2), Hie304 (2), Trp58]
Momordin Ic	34 (4.98 Å)	4 (4.43 Å) [Asn297, Asp196 (3)]	10 (5.19 Å) [Ile234 (3), Lys199, Hie200 (2), Leu161, Tyr150, Leu236 (2)]
Acarbose	18 (4.38 Å)	8 (4.09 Å) [Ala306 (2), Asp299 (3), Hie304 (2), Gly305]	2 (5.57 Å) [Leu161, Leu164]
α-glucosidase
Cucurbitacin I	15 (5.82 Å)	3 (5.42 Å) [Lys433, Trp297, Glu357]	3 (6.22 Å) [Ile301, Hie403, Trp297]
Momordin Ic	29 (4.30 Å)	11 (4.00 Å) [Thr310, Asp311 (2), Glu364, Trp366, Asn369 (2), Ser183, Thr314 (2), Phe367]	7 (4.76 Å) [Ile301, Phe502 (3), Tyr499 (2), Phe367]
Acarbose	26 (4.62 Å)	13 (4.54 Å) [Glu405 (2), Trp363, Ala308, Met300 (2), Glu474 (2), Glu364, Glu498 (3), Ser306]	2 (5.16 Å) [Phe367, Glu498]

Conversely, despite the momordin Ic-α-glucosidase complex having the highest number of interactions (Table 7), acarbose-α-glucosidase had a higher negative free binding energy (Table 5). This observation could be attributed to the higher number of hydrogen bonds present in the acarbose-α-glucosidase after 140 ns. Contrastingly, the lowest negative free binding energy observed in the cucurbitacin I-α-glucosidase complex could be attributed to its lower number of interactions and longer bond length exhibited in the α-glucosidase systems as a longer bond length suggests a weaker force between intermolecular or intramolecular systems, resulting in a more positive free binding energy as highlighted earlier.

Additionally, the unfavourable donor-donor bond present at 0 ns (Figure 9) could be attributed to the lesser free binding energy of the cucurbitacin I-α-glucosidase and momordin Ic-α-glucosidase complex relative to the acarbose-α-glucosidase complex. On the other hand, even though both investigated compounds had an unfavourable donor-donor bond, a higher number of interactions, particularly hydrogen bonds in the momordin Ic-α-glucosidase complex could have contributed to its higher negative binding free energy relative to cucurbitacin I-α-glucosidase. Notably, hydrogen bonds have emerged as key players in drug discovery as non-covalent bonds and are essential for molecular recognition, by stabilising protein-ligand interactions, and ensuring the overall stability of drug-target complexes⁷⁵. Overall, momordin Ic was observed to exhibit the highest negative free binding energy when bound to α-amylase and α-glucosidase, as opposed to cucurbitacin I and was reflective of the interaction type and average bond lengths observed.

Pharmacokinetic and toxicity profiling of cucurbitacin I, momordin Ic, and the reference standard

The time, cost, and resources needed for drug discovery can be minimised by predicting the pharmacokinetics and relative toxicity of drug candidates through virtual prediction tools to assess pharmacokinetics and toxicity of drug candidates, which can minimise the risk of failure during preclinical and clinical stages of drug development⁷⁶. As per Lipinski’s rule of five (Ro5), bioactive compounds must adhere to the following criteria including molecular weight (≤500 g/mol), hydrogen bond donors (≤5), hydrogen bond acceptors (≤10), and octanol-water partition coefficient (≤5). Compounds that violate no more than two of these criteria are generally considered suitable for oral administration⁷⁷. Momordin Ic and acarbose failed with three violations, while cucurbitacin I passed the Ro5 with one violation (Table 8), suggesting its oral bioavailability status and ability to reach target tissues, organs or cells. Thus, momordin Ic may require structural modifications to reduce its molecular weight and adjust its hydrogen donor/acceptor properties for improved pharmacological activity. The CYP450 system is a distinct group of heme-bound enzymes which play a crucial role in catalytic oxidation and removal of foreign substances, including environmental pollutants, dietary and herbal molecules, and medications⁷⁸. Remarkably, momordin Ic and acarbose did not contribute to any CYP isoenzymes inhibition, however, cucurbitacin I inhibited just one of the iso-enzymes suggesting its tendency to instigate drug-drug interactions in the liver when co-administered with other drugs. Additionally, all compounds exhibited activity for at least one toxicity endpoint or the other (Table 8); meanwhile, momordin Ic and acarbose were classified under drug toxicity class 4, which is appropriate for drug development with a lethal dose of 1759 and 2000 mg/kg respectively. Conversely, cucurbitacin I was classified under drug toxicity class 1 which is unsuitable for drug development with a lethal dose of 5 mg/kg. The high toxicity profile of cucurbitacin might be linked to its reactive electrophilic centre (α,β-unsaturated carbonyl group) and may possibly establish unwanted covalent bonds. Therefore, modifying this functional group via partial saturation or by introducing electron-donating or polar substituents will reduce its reactivity while enhancing selectivity⁷⁹. Apart from structural modification, the toxicity of cucurbitacin I can be reduced via recent developments in nanotechnology⁸⁰, bioreductive prodrugs⁸¹, antibody-drug conjugates (ADCs)⁸², combination therapy⁸³ among others. Altogether, these techniques ensure that the therapeutic efficacy of the compound is preserved while reducing its toxicity. Summarily, the compound pharmacokinetics and toxicity profiles suggest the need for structural alterations to reduce their toxicity and improve their suitability for drug development.

Table 8.

ADMET properties of cucurbitacin I and momordin Ic docked against α-amylase and α-glucosidase.

Ligands	MW < 500 (g/mol)	HBA ≤ 10	HBD ≤ 5	Log P o/ w ≤ 5	WS	GI Abs	BBB P	Pgp	CYP1A2	CYP2C19	CYP2C9	CYP2D6	CYP3A4	LV (N)	BS	H	C	IM	M	CY	LD50 (mg/kg)	TC
Cucurbitacin I	514.65	7	4	2.88	S	Low	No	Yes	No	No	No	No	Yes	1	0.55	In	In	In	Ac	In	5	1
Momordin Ic	764.94	13	7	2.92	S	Low	No	Yes	No	No	No	No	No	3	0.11	In	In	Ac	Ac	Ac	1750	4
Acarbose	645.60	19	14	−6.67	S	Low	No	Yes	No	No	No	No	No	3	0.17	Ac	In	Ac	Ac	Ac	2000	4

Note: MW: molecular weight; BBB permeant: blood-brain barrier permeation; HB-D: hydrogen bond donor; Log P o/w: partition coefficient; HB-A: hydrogen bond acceptor; CY: cytotoxicity; Pgp: permeability glycoprotein substrate; WS: water solubility; CYP: cytochrome; LV: Lipinski violation; BS: bioavailability score; H: hepatotoxicity; C: carcinogenicity; IM: immunotoxicity; GI abs: gastrointestinal absorption; M: mutagenicity; LD: lethal dose; TC: toxicity class; S: soluble; PS: poorly soluble; Ac: active; In: inactive; N: number of violation.

Conclusion

Overall, ethyl acetate seed and hexane fruit flesh extracts of M. balsamina were found to exhibit the best inhibitory activity against α-amylase and α-glucosidase, respectively indicating that the plant constituents can be developed to regulate postprandial blood glucose levels while a significant concentration of both cucurbitacin I and momordin Ic was detected in all the investigated extracts of the plant. The in silico study revealed that both cucurbitacin I and momordin Ic had more stable, compact, and less flexible interactions than the apo-proteins. In addition, while both compounds modulated α-amylase effectively, they exhibited less modulatory ability against α-glucosidase compared to the standard drug. Despite the limitation of not being able to conduct a cytotoxicity assay for the tested extracts, the pharmacokinetic and toxicity features of the investigated compounds indicated the need for structural modifications required to improve their druggability. In general, while the findings of this study support the antidiabetic potential of M. balsamina via the inhibition of α-amylase and α-glucosidase, further in vitro and in vivo studies are imminent to elucidate their efficacy and evaluation of potential toxicity effects.

Funding Statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Disclosure statement

The authors report no conflicts of interest.

Data availability statement

The data are contained within the article or supplementary material.