α‑Glucosidase Inhibitors from the Leaves of Cannabis sativa: Structure–Activity Relationship, Kinetic Investigation, and Molecular Docking

Abstract

Cannabis sativa L. is a valuable agricultural crop, extensively utilized in various fields and comprising diverse chemical constituents. In preliminary experiments for rat intestinal α-glucosidase inhibition, the methanolic extract of Cannabis sativa demonstrated potential for inhibiting maltase and sucrase. Bioassay-guided isolation led to 30 metabolites, including five new cannabinoids (1, 2, 3, 4, 30) and four new spiroindans (5, 6, 10, 11). Their structures were elucidated using spectroscopic techniques such as NMR, and absolute configurations were determined by Mosher’s method. Cannabinoids were the main contributors to inhibitory potency (IC₅₀: 0.09–0.8 mM), while spiroindans and simple phenolics showed lower activity (IC₅₀: 1.0–2.2 mM). 8-Hydroxycannabinol (18) was the most potent inhibitor retarding the enzymes through a noncompetitive mechanism. Molecular dynamics simulations of compounds 15 and 18 showed that hydrogen bonding between phenolic hydroxyl groups and specific amino acid residues at the allosteric site was essential for strong α-glucosidase binding.

Introduction

Type 2 diabetes is defined by an inadequate insulin supply due to impaired pancreatic β-cell function and insulin resistance. Due to the considerable annual rise in the number of patients, it imposes significant socioeconomic pressures on global health economies, with costs estimated at USD 850 billion in 2017. , The current pharmacological treatments for type 2 diabetes mellitus primarily aim to regulate and reduce blood glucose levels by using oral antihyperglycemic agents, specifically α-glucosidase inhibitors. However, the recent commercial drugs such as acarbose, miglitol, and voglibose are associated with numerous gastrointestinal side effects including bloating, nausea, constipation, and diarrhea during the treatment period. To avoid these adverse effects, alternative natural products have been employed to support for the treatment of diabetes. ,

Cannabis sativa L., a member of the Cannabaceae family derived from Western Asia, has been grown for centuries for food, fiber, and medicine. Its adaptability to diverse climatic conditions and broad geographic distribution reinforces its status as a globally important cash crop. The commercial significance of C. sativa is further evidenced by the rapid growth of the worldwide cannabis market, which was valued at around USD 27.7 billion in 2022 and is anticipated to grow to USD 82.3 billion by 2027, reflecting a compound annual growth rate of 24%. In addition, edible cannabis or edible marijuana products have become extremely popular with highly economic value in regions which have legalized cannabis for recreational use or medicinal purposes. , This crop is currently cultivated in at least 47 countries and is widely recognized as a multifunctional species utilized across nutritional, pharmaceutical, and industrial sectors. , In addition to its agronomic and economic benefits, the leaves of C. sativa contain diverse bioactive compounds, including cannabinoids, flavonoids, and phenolic constituents known for their therapeutic potential. Repurposing these typically underutilized leaves as a source of antioxidants and other phytochemicals offers a sustainable and economically viable valorization strategy. Taken together, the dual role of C. sativa as both a high-yielding crop and a reservoir of valuable bioactive agents underscores its growing importance in agricultural and food chemistry research. Among more than 500 identified chemical constituents, at least 125 compounds belong to the cannabinoid class, which are largely responsible for the plant’s pharmacological potential. These include a wide spectrum of bioactivities such as analgesic, antidiabetic, antioxidant, anti-inflammatory, and antimicrobial effects. , Research on C. sativa consumption indicated a reduced prevalence of diabetes among those who use it, compared to nonusers. Recently, Δ⁹-THC and CBD have been reported to show the promising biological activity against yeast-α glucosidase. However, they revealed lower inhibitory effects than the ethanolic leaf and flower extracts. In addition, the ethanolic and aqueous extracts from the roots of C. sativa also reduced blood glucose levels in diabetic mice. Moreover, based on our preliminary experiments, the methanolic leaf extract of C. sativa demonstrated better α-glucosidase inhibition compared to extracts from seeds and flowers (Table S3). To date, there is no experiment conducted to identify the key components responsible for antidiabetic activity of C. sativa.

The aim of this study was to identify α-glucosidase inhibitors present in the methanolic extract of C. sativa leaves through a bioassay-guided isolation approach. A total of 30 compounds were isolated, including five novel cannabinoids and four previously unreported spiroindans, all of which were fully characterized by spectroscopic methods. To explore the inhibitory potential of these bioactive constituents, structure–activity relationship (SAR) analysis was conducted to evaluate the impact of structural features on α-glucosidase inhibition. Enzyme kinetic assays were subsequently performed to determine the type of inhibition and to quantify kinetic parameters such as K_i and K_i′ in the presence of active compounds. Furthermore, molecular docking simulations were carried out to predict the binding affinities and molecular interactions between the selected active compounds and the α-glucosidase active site.

Materials and Methods

General Experiment Procedures

The NMR spectra were obtained using JEOL JNM-ECZ500R/S1 NMR spectrometers, which was operated at frequencies of 500 MHz for ¹H and 125 MHz for ¹³C nuclei. Analytical thin-layer chromatography (TLC) was carried out on precoated Merck silica gel 60 F254 plates or silica gel 60 RP-18 F254S (Merck). The spots on the TLC profile were visualized under a UV 254 nm followed by spraying with a 3% p-anisaldehyde reagent and heating. Open column chromatography was carried out employing Merck silica gel 60 (70–230 mesh) and Sephadex LH-20. Flash chromatography was conducted using an Isolera One system from Biotage, equipped with SiliaSepTM flash cartridges (C18, 230–400 mesh, 60 Å, 12 g).

Plant Material

Cannabis sativa leaves were collected in Kanchanaburi province, Thailand, in 2022. The voucher specimen (PP021–2022) was deposited at the Center of Excellence in Natural Products Chemistry.

Extraction and Isolation

The air-dried leaf powder of Cannabis sativa (2.5 kg) was subjected to two successive macerations with MeOH (15 L each) at room temperature, each lasting 3 days. After evaporation under reduced pressure, the concentrated methanolic extract (450 g) was successively fractionated by using silica gel column chromatography (silica gel CC) using gradient elution of ethyl acetate (EtOAc)-hexane (0:100 to 100:0) to obtain 6 fractions; named fractions CF1 (60 g), CF2 (45 g), CF3 (50 g), CF4 (52 g), CF5 (60 g), and CF6 (120 g). Based on the α-glucosidase inhibitory evaluation, the active fractions were further purified (See the Supporting Information).

Fraction CF1 (60 g) was separated through a Sephadex LH-20 column, eluted with a MeOH/DCM mixture (1:1, v/v), to give two subfractions (1.1 and 1.2). Subfraction 1.2 (17.8 g) was purified again using the same Sephadex LH-20 column and an elution mixture of MeOH/DCM (1:1, v/v), producing two subfractions (1.2.1 and 1.2.2). Subfraction 1.2.2 (6.8 g) was then processed through C-18 reverse-phase flash column chromatography (FCC) using MeOH/water (10:1, v/v), to afford three subfractions (1.2.2.1–1.2.2.3). Δ⁹-Tetrahydrocannabinol (19, 25 mg) and cannabichromene (20, 20 mg) were isolated from subfraction 1.2.2.2 (1.3 g) using normal-phase silica gel column chromatography (CC) eluted with n-hexane/ethyl acetate (EtOAc)/diethylamine (DEA) (18:1:1, v/v/v). Subfraction 1.2.2.3 (1.6 g) was subjected to silica gel CC eluted with n-hexane/DCM (19:1, v/v), yielding compound 15 (25 mg). Fraction CF2 (45 g) was purified using a Sephadex LH-20 column eluted with MeOH/DCM (1:1, v/v), yielding two subfractions (2.1 and 2.2). Subfraction 2.2 (6.8 g) was further separated by the same Sephadex LH-20 column with a MeOH/DCM mixture (6:4, v/v), yielding three subfractions (2.2.2.1–2.2.2.3). Compounds 5 (6.5 mg) and 6 (3.5 mg), cannabigerol (21, 24 mg), and cannabichromanone (22, 3.2 mg) were obtained from subfraction 2.2.2.2 (300 mg) using C-18 reverse-phase FCC with MeOH/water (17:2, v/v). Subfraction 2.2.2.3 (250 mg) was further purified by C-18 reverse-phase FCC eluted with MeOH/water (8:1, v/v), to get compounds 1 (10 mg), 2 (4.2 mg), and 3 (4.5 mg). Fraction CF3 (50 g) was chromatographed on Sephadex LH-20 using MeOH/DCM (1:1, v/v), yielding two subfractions (3.1 and 3.2). Subfraction 3.2 was further subjected to a Sephadex LH-20 column eluted with MeOH/DCM (6:4, v/v), to obtain four subfractions (3.2.2.1–3.2.2.4). Compounds 4 (3.8 mg) and 30 (6.0 mg) were isolated from subfraction 3.2.2.1 (200 mg) using silica gel CC with n-hexane/EtOAc (17:3, v/v). Subfraction 3.2.2.3 (150 mg) was further purified by reverse phase C-18 FCC (MeOH/water, 15:2, v/v) to afford cannabinol (16, 15.1 mg), 17 (7.8 mg), and 18 (32.3 mg). Fraction CF6 (120 g) was subjected to a Sephadex LH-20 column eluted with MeOH/DCM (1:1, v/v), to afford two subfractions (6.1 and 6.2). Subfraction 6.2 (28.5 g) was further fractionated using Sephadex LH-20 eluted with a MeOH/DCM mixture (7:3, v/v), yielding two subfractions (6.2.1 and 6.2.2). Subfraction 6.2.2 (6.8 g) was purified using C-18 reverse-phase FCC eluted with MeOH/water (3:1, v/v), resulting in three subfractions (6.2.2.1–6.2.2.3). Canniprene (26, 50 mg), vanillin (27, 20 mg), syringaldehyde (28, 10 mg), and vanillic acid (29, 15 mg) were isolated from subfraction 6.2.2.1 (800 mg) using C-18 reverse-phase FCC eluted with MeOH/water (2:1, v/v). Subfraction 6.2.2.2 (1.2 g) was purified through reverse phase C-18 FCC eluted with MeOH/water (3:2, v/v), to yield α-cannabispiranol (12, 14 mg), β-cannabispiranol (13, 15 mg), and acetylcannabispirol (14, 9.8 mg).

Fraction CF4 (52 g) was fractionated through Sephadex LH-20 eluted with MeOH/DCM (1:1, v/v), to afford three subfractions (4.1–4.3). Fraction 4.2 (10.5 g) was fractionated using Sephadex LH-20 eluted with MeOH/DCM (6:4, v/v), yielding two subfractions (4.2.1 and 4.2.2). Subfraction 4.2.2 (5.9 g) was purified using C-18 reverse phase FCC eluted with MeOH/water (5:1, v/v), yielding four subfractions (4.2.2.1–4.2.2.4). Subfraction 4.2.2.1 (1.2 g) was subjected to silica gel CC and eluted with n-hexane/DCM (9:1, v/v), affording compound 10-methoxy-β-cannabispiranol (10, 15.1 mg). 10-Methoxy-α-cannabispiranol (11, 9.8 mg), cannabitriol (23, 25 mg), and cannabitriol-C3 (24, 3.5 mg) were obtained from subfraction 4.2.2.3 (1.6 g) using C-18 reverse phase FCC eluted with MeOH/water (5:1, v/v). Fraction CF5 (60 g) was separated using a Sephadex LH-20 column and eluted with MeOH/DCM (1:1, v/v), to yield two subfractions (5.1 and 5.2). Subfraction 5.2 (6.6 g) underwent further purification using the same Sephadex LH-20 column and a MeOH/DCM mixture (7:3, v/v), affording two subfractions (5.2.1 and 5.2.2). Subfraction 5.2.2 (2.5 g) was purified using C-18 reverse phase FCC eluted with MeOH/water (4:1, v/v), resulting in four subfractions (5.2.2.1–5.2.2.4). Subfraction 5.2.2.1 (800 mg) was purified through C-18 reverse phase FCC eluted with MeOH/water (7:2, v/v), to obtain cannabispirone (7, 25 mg), isocannabispirone (8, 4.2 mg), and cannabispirenone (9, 6.5 mg). Compound 25 (25 mg) was isolated from subfraction 5.2.2.3 using silica gel CC eluted with n-hexane/EtOAc (9:1, v/v).

1-O-Prenylcannabitriol (CBT-1P, 1): An amorphous powder; [α]_D −40° (c 1.0, MeOH); UV (MeOH) λ_max (log ε) 205 (3.40), 230 (3.47), 280 (3.15) nm; ¹H (500 MHz) and ¹³C (125 MHz) NMR data (CDCl₃), see Table ; HRESI-MS m/z 415.2843 [M + H]⁺ (calcd for C₂₆H₃₉O₄ ⁺, 415.2842).

2. ¹H NMR and ¹³C NMR Data of 1 and 30 in CDCl₃ .

	(1)
no.	δ H (mult, J  in Hz)	δ C (ppm)
1		154.0
2	6.36 (d, 1.4)	106.2
3		144.0
4	6.40 (d, 1.4)	110.96
4a		154.2
6		77.1
6a		134.8
7	2.30 (m)	23.2
	2.17 (m)
8	1.93 (m)	29.8
	1.68 (m)
9		71.1
10	4.23 (d, 3.6)	72.1
10a		124.9
10b		111.04
11	1.34 (s)	24.7
12	1.39 (s)	24.9
13	1.27 (s)	24.1
1′	2.51 (t, 7.5)	36.1
2′	1.59 (m)	30.8
3′	1.32 (m)	31.6
4′	1.31 (m)	22.7
5′	0.89 (t, 6.9)	14.2
1′′	4.62 (m)	65.8
	4.53 (m)
2′′	5.53 (t, 7.0)	118.4
3′′		140.6
4′′	1.81 (s)	26.0
5′′	1.77 (s)	18.4
9-OH	1.87 (s)
10-OH	4.08 (d, 3.6)

1-O-Prenyl-2,4-diprenylcannabitriol (CBT-1,2,4TP, 2): A yellow oil; [α]_D −43° (c 1.0, MeOH); UV (MeOH) λ_max (log ε) 214 (3.44), 226 (3.40), 270 (3.06) nm; ¹H (500 MHz) and ¹³C (125 MHz) NMR data (CDCl₃), see Table ; HRESI-MS m/z 551.4103 [M + H]⁺ (calcd for C₃₆H₅₅O₄ ⁺, 551.4089).

3. ¹H NMR and ¹³C NMR Data for Compounds 2-4 in CDCl₃ .

	(2)		(3)		(4)
	δ H (mult, J  in Hz)	δ C (ppm)	δH(mult, J  in Hz)	δ C (ppm)	δH(mult, J  in Hz)	δ C (ppm)
1		149.4		151.2		151.8
2		126.6		126.5	6.36 (brs)	106.5
2		141.2		143.1		141.6
4		126.1	6.55 (brs)	114.6		121.7
4a		150.4		152.2		152.0
6		76.9		77.2		77.0
6a		136.4		136.3		134.9
7	2.29 (m)	24.7	2.29 (m)	24.8	2.30 (m)	23.1
	2.14 (m)		2.15 (m)		2.16 (m)
8	1.85 (m)	33.2	1.86 (m)	33.1	1.94 (m)	29.8
	1.78 (m)		1.74 (m)		1.67 (m)
9		71.9		71.8		71.1
10	4.53 (brs)	74.5	4.53 (brs)	74.4	4.22 (d, 3.2)	72.2
10a		125.3		126.4		125.2
10b		113.9		114.1		111.2
11	1.22 (s)	20.4	1.23 (s)	20.5	1.34 (s)	24.5
12	1.44 (s)	25.4	1.19 (s)	22.7	1.38 (s)	24.8
13	1.13 (s)	22.7	1.44 (s)	25.7	1.24 (s)	24.0
14	3.42 (m)	25.8	3.40 (dd, 15.5, 6.7)	25.2	3.28 (dd, 14.8, 6.9)	24.8
	3.35 (m)		3.22 (dd, 15.5, 5.8)		3.24 (dd, 14.8, 7.0)
15	5.09 (t, 6.7)	125.2	5.06 (t, 6.0)	124.5	5.04 (t, 6.9)	124.0
16		130.8		131.2		130.2
17	1.73 (s)	25.8	1.66 (s)	25.8	1.66 (s)	25.9
18	1.67 (s)	18.1	1.73 (s)	18.1	1.77 (s)	18.4
19	3.22 (m)	25.5
	3.18 (m)
20	5.06 (t, 7.0)	124.3
21		130.1
22	1.67 (s)	26.0
23	1.76 (s)	18.3
1′	2.51 (m)	29.8	2.49 (m)	32.9	2.54 (m)	33.6
2′	1.41 (m)	30.8	1.51 (m)	30.5	1.53 (m)	31.1
3′	1.36 (m)	32.9	1.24 (m)	29.8	1.34 (m)	32.1
4′	1.34 (m)	22.8	1.31 (m)	32.1	1.35 (m)	22.7
5′	0.89 (t, 7.0)	14.3	0.89 (t, 7.0)	14.2	0.90 (t, 7.0)	14.2
1′′	4.32 (m)	73.0	4.34 (m)	73.0	4.61 (m)	65.7
	4.20 (m)		4.24 (m)		4.51 (m)
2′′	5.47 (t, 7.0)	119.8	5.48 (t, 7.1)	119.6	5.52 (t, 7.1)	118.6
3′′		138.0		138.4		140.2
4′′	1.63 (s)	25.9	1.76 (s)	26.0	1.76 (s)	25.9
5′′	1.76 (s)	18.2	1.64 (s)	18.3	1.80 (s)	18.1
9-OH	1.60 (s)		1.64 (s)		1.64 (s)
10-OH	4.30 (d, 4.1)		4.31 (d, 4.0)		4.15 (d, 3.4)

^aExchangeable within the column.

1-O-Prenyl-2-prenylcannabitriol (CBT-1,2DP, 3): A yellow oil; [α]_D −42° (c 1.0, MeOH); UV (MeOH) λ_max (log ε) 209 (3.50), 226 (3.43), 275 (3.09) nm; ¹H (500 MHz) and ¹³C (125 MHz) NMR data (CDCl₃), see Table ; HRESI-MS m/z 483.3477 [M + H]⁺ (calcd for C₃₁H₄₆O₄ ⁺, 483.3472).

1-O-Prenyl-4-prenylcannabitriol (CBT-1,4DP, 4): A yellow oil; [α]_D −38° (c 1.0, MeOH); UV (MeOH) λ_max (log ε) 210 (3.40), 230 (3.33), 278 (3.01) nm; ¹H (500 MHz) and ¹³C (125 MHz) NMR data (CDCl₃), see Table ; HRESI-MS m/z 483.3477 [M + H]⁺ (calcd for C₃₁H₄₆O₄ ⁺, 483.3472).

10-O-Prenylcannabispirone (5): A yellow oil; [α]_D 46° (c 1.0, MeOH); UV (MeOH) λ_max (log ε) 213 (4.24), 232 (3.77), 285 (3.44) nm; ¹H (500 MHz) and ¹³C (125 MHz) NMR data (CDCl₃), see Table ; HRESI-MS m/z 315.1955 [M + H]⁺ (calcd for C₂₀H₂₇O₃ ⁺, 315.1955).

4. ¹H NMR and ¹³C NMR Data of Compounds 5 and 6 in CDCl₃ .

	(5)		(6)
no.	δ H (mult, J  in Hz)	δ C (ppm)	δ H (mult, J  in Hz)	δ C (ppm)
1		213.3		213.5
2	2.49 (td, 14.1, 5.7)	39.1	2.49 (td, 14.0, 4.6)	39.1
	2.41 (m)		2.41 (d, 15.0)
3	2.64 (td, 13.3, 4.8)	34.2	2.64 (td, 13.3, 4.7)	34.3
	1.80 (m)		1.80 (m)
4		48.0		48.2
5	2.22 (t, 7.5)	35.5	2.20 (t, 7.5)	35.3
6	2.93 (t, 7.5)	31.0	2.88 (t, 7.4)	29.3
6a		145.7		144.5
7	6.36 (d, 1.9)	100.8		118.0
8		160.6		157.5
9	6.29 (d, 1.9)	98.2	6.32 (1H, s)	95.4
10		156.6		154.4
10a		128.4		127.9
11			3.23 (d, 7.1)	25.8
12			5.13 (t, 7.1)	122.9
13				131.0
14			1.72 (s)	17.9
15			1.80 (3H, s)	25.9
1′	4.47 (d, 6.5)	64.8	4.49 (d, 6.6)	64.9
2′	5.41 (t, 6.5)	119.9	5.41 (t, 6.6)	120.2
3′		137.6		137.3
4′	1.70 (s)	18.4	1.75 (s)	18.4
5′	1.77 (s)	25.9	1.66 (s)	25.9
8-OMe	3.79 (s)	55.5	3.80 (s)	56.1

10-O-Prenyl-7-prenylcannabispirone (6): A yellow oil; [α]_D 47° (c 0.3, MeOH); UV (MeOH) λ_max (log ε) 210 (4.36), 227 (3.81), 278 (3.17) nm; ¹H (500 MHz) and ¹³C (125 MHz) NMR data (CDCl₃), see Table ; HRESI-MS m/z 383.2581 [M + H]⁺ (calcd for C₂₅H₃₅O₃ ⁺, 383.2581).

1β-Hydroxy-10-methoxy-cannabispiranol (10): A yellow oil; [α]_D 189° (c 1.2, MeOH); UV (MeOH) λ_max (log ε) 208 (4.30), 226 (3.73), 280 (3.12) nm; ¹H (500 MHz) and ¹³C (125 MHz) NMR data (CDCl₃), see Table ; HRESI-MS m/z 263.1641 [M + H]⁺ (calcd for C₁₆H₂₃O₃ ⁺, 263.1642).

5. ¹H NMR and ¹³C NMR Data of Compounds 10 and 11 in CDCl₃ .

	(10)		(11)
	δ H (mult, J  in Hz)	δ C (ppm)	δ H (mult, J  in Hz)	δ C (ppm)
1	4.06 (t, 3.0)	65.9	3.71 (m)	71.0
2	1.73 (m)	30.0	1.46 (m)	32.9
3	2.53 (td, 13.7, 4.6, H3ax)	28.0	2.25 (td, 14.6, 4.9, H3ax)	33.0
	1.26 (dt, 15.9, 2.4, H3eq)		1.99 (m)
4		48.6		48.4
5	2.0 (t, 7.6)	34.7	2.02 (t, 7.6)	35.5
6	2.84 (t, 7.6)	31.2	2.84 (t, 7.6)	31.2
6a		145.6		145.9
7	6.34 (d, 1.8)	100.7	6.34 (d, 2.0)	100.7
8		160.3		160.4
9	6.29 (d, 1.8)	97.3	6.27 (d, 2.0)	97.3
10		157.6		157.5
10a		130.5		129.7
8-OMe	3.79 (s)	55.5	3.77 (s)	55.5
10-OMe	3.77 (s)	55.3	3.76 (s)	55.1

1α-Hydroxy-10-methoxy-cannabispiranol (11): A yellow oil; [α]_D 125° (c 2.0, MeOH); UV (MeOH) λ_max (log ε) 205 (4.40), 225 (3.86), 275 (3.17) nm; ¹H (500 MHz) and ¹³C (125 MHz) NMR data (CDCl₃), see Table ; HRESI-MS m/z 263.1646 [M + H]⁺ (calcd for C₁₆H₂₃O₃ ⁺, 263.1642).

8-Hydroxy-1-methoxycannabinol (CBN-8OH-10OMe, 30): A yellow oil; UV (MeOH) λ_max (log ε) 215 (3.40), 290 (3.15), 305 (3.04) nm; ¹H (500 MHz) and ¹³C (125 MHz) NMR data (CDCl₃), see Table ; HRESI-MS m/z 341.2112 [M + H]⁺ (calcd for C₂₂H₂₉O₃ ⁺, 341.2105).

1. Summary of MD Systems.

Model system	Total water	Total atoms	MD length (ns) × repeats
rat-ntMGAM-cpd 15	31,863	106,800	100 × 3
rat-ntMGAM-cpd 18	31,862	106,798	100 × 3

Determination of Absolute Configuration

Preparation of R- and S-MPA ester of 1 was performed using our previously described protocol. A solution of compound 1 (1 mg, 0.0024 mmol) in anhydrous CH₂Cl₂ (DCM) (1 mL) was mixed with R-MPA acid (0.4 mg, 0.0024 mmol), EDC (0.56 mg, 0.0036 mmol), and DMAP (0.3 mg, 0.0024 mmol). The reaction mixture was stirred in an ice bath for 2 h before being quenched with water and extracted with DCM. The combined organic phases were dried over anhydrous Na₂SO₄, and the solvent was evaporated under reduced pressure. The R-MPA ester product (1a) was purified via silica gel column chromatography using a mixture of EtOAc-Hexane (1:4) as an eluent. The S-MPA ester (1b) was synthesized by a similar procedure. The ¹H NMR spectra of R-MPA ester (1a) and S-MPA ester (1b) are presented in Figures S11 and S13, respectively.

α-Glucosidase Inhibitory Activities

The inhibition against α-glucosidase was determined by a colorimetric method using our previous procedures. − One gram of rat intestinal acetone powder (Sigma-Aldrich,St. Louis, MO, USA) was homogenized in 30 mL of a 0.9% NaCl solution. After centrifuging at 12,000 rpm for 30 min, an aliquot of the supernatant was used as an enzyme stock solution. Isolated compounds which were dissolved in DMSO at three concentration levels ranging between 0.01 and 1 mg/mL (10 μL) were added to a mixture containing 0.1 M phosphate buffer (pH 6.9, 30 μL) and a substrate solution (20 μL of 10 mM maltose or 100 mM sucrose). Subsequently, 80 μL of a glucose assay kit (Human Gesellschaft für Biochemica and Diagnostica mbH, Germany) and 20 μL of the enzyme stock solution were added in order. Then, the reaction mixture was incubated at 37 °C for 10 min (for maltase) or 40 min (for sucrase). The amount of glucose liberated after incubation, which is proportional to the amount of quinoneimine, was determined by its absorbance at 520 nm using a Bio-Rad 3550 microplate reader. The inhibition percentage was calculated using the formula [(A₀ - A₁)/A₀] * 100, where A₁ represents the absorbance with the samples, and A₀ represents the absorbance without the samples. The IC₅₀ value was derived from a plotting of inhibition percentage (Y axis) against sample concentration (X axis). Acarbose (Bayer, Germany) served as a positive control, and each experiment was conducted in triplicate.

Kinetic Study of α-Glucosidase Inhibition

The enzyme kinetics study was conducted following the protocol described in our previous reports. − The inhibition type was determined by constructing Lineweaver–Burk plots, which involved varying the concentrations of the substrates (maltose and sucrose) and rat intestinal α-glucosidase in the presence and absence of the tested compounds. The active compounds (10 μL) were mixed with phosphate buffer (pH 6.9, 30 μL) and varying concentrations of maltose (2–10 mM) and sucrose (20–100 mM). An enzyme solution (20 μL) was then added, and the mixture was incubated at 37 °C for 10 min (for maltase) and 40 min (for sucrase). The activity of α-glucosidase was determined at 520 nm using a microplate reader, and each assay was performed in triplicate with acarbose as the positive control. Inhibition types were deduced through a series of Lineweaver–Burk plots. To calculate the K_i and K_i′ values, secondary plots were generated from the slopes and intercepts of the Lineweaver–Burk plots versus the concentrations of the active compounds. The data from the kinetic study provided insights into the inhibition mechanism.

Molecular Docking and Binding Energy Calculation

Molecular docking was employed to predict the potential binding poses of cannabis compounds and their derivatives within the target protein. The target protein is a homology model of the rat intestinal N-terminal catalytic domain of maltase-glucoamylase (rat-ntMGAM) which was developed in our previous study. , This model was constructed based on the 3D crystal structure of the human intestinal N-terminal domain of maltase-glucoamylase (PDB code 3L4W), sharing 92% sequence similarity and 82% sequence identity between the rat and human proteins. Similar to human-ntMGAM, the rat-ntMGAM model consists of five major structural domains: (1) a Trefoil Type-P domain; (2) an N-terminal β-sandwich domain; (3) a catalytic (β/α) barrel domain (4) a proximal C-terminal domain; and (5) a distal C-terminal domain (Figure S76).

To investigate the binding interactions between cannabis compounds and the predicted binding pocket of the target protein, molecular docking was performed. The binding pocket was selected based on its ranking from MetaPocket and Fpocket scores, ensuring that the highest-ranked site was chosen. Importantly, this pocket was distinct from the substrate-binding pocket where miglitol binds, which is known for competitive inhibition. This distinction was essential, as the ligands under investigation exhibit either uncompetitive or noncompetitive inhibition mechanisms. Small ligand molecules present in the template crystal structure such as glycerol (GOL), N-acetyl-d-glucosamine (NAG), and miglitol (MIG) (Figure S76) were excluded from the docking study. The rat-ntMGAM model comprises residues 68 to 928. To explore the potential binding interactions of cannabis compounds 15, 18, 23, 2, 6, and 7 with the residues of rat-ntMGAM, flexible docking was performed using AutoDock Vina. Compounds 15, 23, and 18 were selected based on their high binding affinity (low K_i values), while compounds 6 and 7, characterized by low affinity or inactivity, were included as negative controls. The 3D structures of these compounds were constructed and optimized at the AM1 or B3LYP level and with 6–31+G­(d,p) using Gaussian09. Convergence was achieved when the maximum force acting on any atom fell below 0.00045 hartree/Bohr and the maximum allowed displacement was set at 0.0018 Å.

To set up the docking parameters, polar hydrogens were added, and partial atomic charges for both the protein and inhibitors were calculated using AutoDockTools, with Gasteiger charges applied. All rotatable bonds in the inhibitors, as identified by the program, were treated as active torsional bonds to allow flexible docking. A grid box was defined around the binding site predicted by MetaPocket and Fpocket, ensuring that it encompassed the entire binding cavity, with dimensions of 20 Å × 20 Å × 20 Å. The inhibitor was then placed within this grid. The exhaustiveness parameter, which determines the thoroughness of the search, was set to 32. Vina utilizes a scoring function to evaluate the energy of each inhibitor pose within the grid. The scoring function considers factors such as van der Waals interactions, hydrogen bonding, electrostatic interactions, and desolvation energy. Docking simulations were performed, generating multiple binding poses for each inhibitor. These docking conformations were ranked based on their binding affinities, and the top-ranked poses were further analyzed for their interactions with the protein using VMD (Visual Molecular Dynamics). Additionally, clustering analysis was performed to identify the most frequently occurring binding poses, and molecular dynamics simulations were employed to assess the stability of the docked complexes.

The final models of the complex structures, extracted from MD snapshots, included the ligand and amino acid residues within 3 Å of the ligand. These models were re-evaluated using binding energy calculations based on the single-point energy method at the AM1 level

$$\mathrm{\Delta }{E}_{bind}=E_{complex}-(E_{protein}+E_{ligand})$$ 1

where E_complex represents the total energy of the protein–ligand complex, E_protein denotes the energy of the isolated protein, and E_ligand corresponds to the energy of the unbound ligand. The binding energy (ΔE_bind ) is determined as the difference between the energy of the complex and the sum of the individual energies of the protein and ligand. A more negative ΔE_bind value indicates a stronger binding affinity between the ligand and the target protein.

Molecular Dynamics Simulation

Molecular dynamics (MD) simulations were conducted to evaluate the binding stability of the rat-ntMGAM in complex with compounds 15 and 18, as identified from the docking results. Missing hydrogen atoms were added to the structures using the PSFGEN module in VMD. The protonation states of ionizable residues, such as Lys, Arg, Asp, and Glu, were assigned based on pK _a predictions conducted using PROPKA at neutral pH. The protein was then solvated in a simulation box filled with TIP3P water molecules. To neutralize the overall charge of the system, counterions were added, and a physiological salt concentration of 0.1 M NaCl was maintained using VMD’s Autoionize plugin.

MD simulations of all systems were performed using NAMD software, employing the CHARMM36 all-atom force field to model protein interactions. , Force field parameters for the cannabis compounds 15 and 18 were generated with the CHARMM Generalized Force Field (CGenFF) via the CHARMM-GUI Web server. The TIP3P force field parameters were used to represent solvent interactions. Simulations were conducted under periodic boundary conditions with a simulation box of approximately 110 Å × 100 Å × 100 Å, containing ∼106,800 atoms. A 12 Å cutoff was applied for nonbonded interactions, with electrostatic interactions computed using the particle mesh Ewald (PME) summation method. van der Waals interactions were smoothly switched off at a 10 Å distance. The system temperature was maintained at 298 K using Langevin dynamics with a damping coefficient of 1 ps^–1. Pressure was regulated at 1 atm using the Nose–Hoover Langevin piston method, with a piston period of 200 fs and a damping time of 50 fs.

Energy minimization was performed to eliminate unfavorable atomic contacts. The system was then equilibrated through three sequential steps. In the first step, the protein and inhibitor were restrained, while water molecules and counterions were allowed to adjust, forming a stable solvation shell around the protein. In the second step, water molecules, counterions, and protein side chains were relaxed, while the protein backbone and inhibitor remained fixed. In the final equilibration step, all atoms were allowed to move freely. Following equilibration, production MD simulations were conducted for 100 ns using a 2 fs time step. To enhance statistical significance and reproducibility, three independent simulation runs were performed. All simulations were based on the model systems outlined in Table .

MD Trajectory Analysis

Structural and dynamic properties of the simulated systems were evaluated using root-mean-square deviation (RMSD), hydrophobic interaction analysis, and hydrogen bond analysis. Backbone RMSD values relative to the initial structures were computed using the ’measure’ commands in VMD, assessing both the protein backbone and CBN inhibitors. For protein backbone RMSD calculations, particular focus was given to residues within 10 Å of the inhibitors, as identified from the docking results, specifically residues 273–284, 407–417, 429–434, 494–530, 556–559, 630–638, and 761–773. Hydrogen bond occupancy was quantified as the percentage of simulation time during which a hydrogen bond was maintained, applying geometric criteria of a donor–acceptor distance ≤ 3.5 Å and a donor-hydrogen-acceptor angle ≥ 120°. Structural property analyses were conducted using TCL scripting in VMD, with a particular emphasis on the 80–100 ns simulation time frame. Complex structures extracted from MD snapshots were selected based on ligand and amino acid residues within 3 Å of the ligand. These structures were then used for binding energy calculations via single-point energy calculations using both the AM1 and PM6 semiempirical methods and DFT calculations at the B3LYP/6–31G level of theory, following Equation . Additionally, binding free energies were estimated using the MMGBSA method to provide a more refined evaluation of ligand-protein interactions.

Results and Discussion

Extraction and Isolation

To explore bioactive components responsible for suppressing glucose levels, we applied α-glucosidase inhibitory guided isolation. This strategy led to the discovery of nine new compounds (1, 2, 3, 4, 5, 6, 10, 11, 30) and 21 previously known compounds as illustrated in Figure . The known compounds were identified through spectroscopic analysis and literature data comparison as cannabispirone (7), isocannabispirone (8), cannabispirenone (9), − α-cannabispiranol (12), β-cannabispiranol (13), acetylcannabispirol (14), cannabinol (15), cannabinol-C3 (16), , cannabinol-C1 (17), , 8-hydroxycannabinol (18), Δ⁹-tetrahydrocannabinol (19), , cannabichromene (20), , cannabigerol (21), cannabichromanone (22), cannabitriol (23), − cannabitriol-C3 (24), 10-methoxycannabitriol (25), canniprene (26), vanillin (27), syringaldehyde (28), and vanillic acid (29).

1.

Isolated compounds from the leaves of Cannabis sativa.

Structural Elucidation of New Compounds (1, 2, 3, 4, 5, 6, 10, 11, and 30)

1-O-Prenylcannabitriol (CBT-1P, 1) was obtained as a white powder. HRESI-MS displayed a pseudomolecular ion [M + H]⁺ peak at m/z 415.2843 (calcd. for C₂₆H₃₉O₄, 415.2842). The UV spectrum showed absorption maxima at 205, 230, and 280, indicating a conjugated chromophore corresponded with a styrene core structure. The ¹H NMR spectrum indicated the presence of two meta-aromatic protons at δ_H 6.36 (d, J = 1.4 Hz, 1H, H-2) and 6.40 (d, J = 1.4 Hz, 1H, H-4) and an n-pentyl side chain at 2.51 (t, J = 7.5 Hz, 2H, H-1′), 1.59 (m, 2H, H-2′), 1.32 (m, 2H, H-3′), 1.31 (m, 2H, H-4′), and 0.89 (t, J = 6.9 Hz, 3H, H-5′). These signals suggested the presence of a phenolic core structure comprising an alkyl residue, which is typical of cannabinoid type compounds. The three singlet methyls at δ_H 1.34 (H-11), 1.39 (H-12), and 1.27 (H-13), two methylenes (2.30 and 2.17 ppm), and one oxygenated methine at δ_H 4.23 (d, J = 3.6 Hz, 1H, H-10) suggested the presence of a monoterpenyl portion, which was subsequently verified by ¹³C NMR together with 2D NMR data analysis (Figure ). In the HMBC spectrum, the correlations of H-7/C-9, H-11/C-8, H-11/C-9, H-10/C-9, and H-10/C-6a confirmed the presence of a bicyclic monoterpenyl portion. In addition, a signal found at δ_H 4.23 (d, J = 3.6 Hz, 1H, H-10) was assigned to oxygenated methine. Therefore, the cannabitriol (CBT) core structure was established definitely. ,, The presence of the prenyl unit was suggested by the characteristic resonances at δ_H 4.62 (m, 1H, H-1′′a), 4.53 (m, 1H, H-1′′b), 5.53 (t, J = 7.0 Hz, 1H, H-2′′), 1.81 (s, 3H), and 1.77 (s, 3H). According to the above-mentioned data, compound 1 possessed a CBT core structure comprising one isoprenyl group, which was further accommodated at the benzene ring by HMBC. The diagnostic correlation of H-1′′/C-1 (154.0 ppm) indicated the connectivity of the prenyl group to C-1 of the benzene ring through the ether bond. Once the whole structure of compound 1 was established, the relative configuration was further deduced by NOESY and coupling constant analysis. The through space correlations of H-10/CH₃-11 and 9-OH/10-OH (Figure S10) suggested the cis-orientation of vicinal dihydroxy groups. To gain insight into the absolute configuration of contiguous two chiral centers, C-9 and C-10, Mosher’s method was applied. Esterification of compound 1 with (R)- and (S)-MPA acids in the presence of N-ethyl-N′-(3-(dimethylamino)­propyl)­carbodiimide (EDC) afforded the corresponding MPA esters 1a and 1b, respectively (Figure A). The distribution of the ¹H chemical shift difference between 1b and 1a (Δδ ^RS , Figure B) indicated that C-10 possessed the R-configuration, while C-9 was inevitably assigned as the S-configuration. Thus, compound 1 was identified as 1-O-prenylcannabitriol (CBT-1P).

2.

Selected COSY, HMBC, and NOESY correlations of 1-prenylcannabitriol (1).

3.

(A) Synthesis of MPA esters of 1, reagent, and conditions: (a) (R)-MPA, EDC, DMAP, CH₂Cl₂ and (b) (S)-MPA, EDC, DMAP, CH₂Cl₂; (B) Δδ ^RS distribution.

1-O-Prenyl-2,4-diprenylcannabitriol (CBT-1,2,4TP, 2) was isolated as a yellow oil. The molecular formula of 2 was established as C₃₆H₅₄O₄ based on the pseudomolecular ion [M + H]⁺ at m/z 551.4103 (calcd. for C₃₆H₅₅O₄, 551.4089). The ¹H NMR spectrum of 2 demonstrated particular signals which essentially resemble those of compound 1 such as the pentyl group, bicyclic tepenyl portion, and oxygenated prenyl group. The lack of two aromatic protons H-2 and H-4 together with the 136 Da (C₁₀H₁₆) larger molecular formula suggested that C-2 and C-4 were possibly connected to two prenyl groups as substituents. This assumption was verified by the characteristic signals of two prenyl groups (Table ) and HMBC cross peaks. The correlations of H-14/C-2 (126.6 ppm) and H-19/C-4 (126.1 ppm) indicated that two prenyl groups accommodated the aromatic ring through the C–C bond at C-2 and C-4, respectively. We further addressed the relative configuration of two chiral centers, C-9 and C-10, by NOESY data analysis (Figure S22). A pair of correlations between 9-OH/10-OH was indicative of vicinal diol cis-orientation, while the absolute stereochemistry was proposed to be identical to those described in compound 1, based on comparable specific rotation (−40°). Therefore, compound 2 was assigned as 1-O-prenyl-2,4-diprenylcannabitriol (CBT-1,2,4TP).

1-O-Prenyl-2-prenylcannabitriol (CBT-1,2DP, 3) and 1-O-prenyl-4-prenylcannabitriol (CBT-1,4DP, 4) were obtained as a white powder and yellow oil, respectively. The molecular formulas of 3 and 4 were identically determined to be C₃₁H₄₆O₄ from their HRESIMS [M + H]⁺ ions at m/z 483.3477 (calcd. for C₃₁H₄₇O₄, 483.3472). The ¹H and ¹³C NMR spectroscopic data of 3 and 4 were similar to those of compound 1, except for the presence of one additional prenyl group, which was supported by a 68 Da (C₅H₈) larger molecular mass. Thus, compounds 3 and 4 were structural isomers possessing the CBT-1P core structure. Careful comparison of ¹H NMR of both 3 and 4 preliminarily suggested that one additional prenyl group substituted at the aromatic ring. The presence of H-4 (6.55 ppm) in compound 3 and H-2 (6.36 ppm) in compound 4 indicated that one additional prenyl group was accommodated at C-2 of compound 3 and C-4 of compound 4, respectively. These linkages were supported by the diagnostic HMBC correlation of H-14 (3.40 and 3.22 ppm)/C-2 (126.5 ppm) in 3 and that of H-14 (3.28 and 3.24 ppm)/C-4 (121.8 ppm) in 4 (Figure ).

4.

Selected HMBC correlations of compounds 2-4.

The relative configuration of two chiral centers, C-9 and C-10, was further determined by NOESY data analysis (Figure S30). The NOESY spectrum of compound 3 showed the correlation between H-10/CH₃-11 which indicates the vicinal diol cis-orientation, whereas the absolute stereochemistry was proposed to be identical to those described in compound 1, based on comparable specific rotation (−42°). Similarly, the correlation between 9-OH/10-OH (Figure S38) also supported the vicinal diol cis-orientation of compound 4. The absolute stereochemistry of 4 was determined to be similar to compound 1, evidenced by comparable specific rotation (−38°).

10-O-Prenylcannaspirone (5) was isolated as a yellow oil. The HRESIMS exhibited a pseudomolecular ion peak at m/z 315.1955 ([M + H]⁺), implying the molecular formula of C₂₀H₂₆O₃. The UV spectra (MeOH) displayed maximum absorption bands at 210, 220, and 275 nm, suggesting the presence of a conjugated chromophore of cannabispirone. The ¹H NMR spectrum revealed the presence of one methoxy group at δ_H 3.79 (s), meta-aromatic protons at δ_H 6.36 (d, J = 1.9 Hz, 1H, H-7) and 6.29 (d, J = 1.9 Hz, 1H, H-9), six methylene units at δ_H 2.49 (td, J = 14.1, 5.7 Hz, 2H, H-2a), 2.41 (d, J = 15.0 Hz, 2H, H-2b), 2.64 (td, J = 13.3, 4.8 Hz, 2H, H-3a), 1.80 (m, 2H, H-3b), 2.22 (t, J = 7.5 Hz, 2H, H-5), and 2.96 (t, J = 7.5 Hz, 2H, H-6). These observed data suggested the core structure of cannabispirone. ,, The COSY and HMBC data assisted in assigning partial substructures A, B, and C (Figure ). The identification of symmetrical ring A was accomplished through COSY correlation of H-2/H-3 and HMBC correlations from H-2 to C-4 along with H-3 to C-2, C-4, and C-5. On the other hand, rings B and C were constructed by the following HMBC correlations; H-5 to C-3, C-4, C-6, C-6a, and C-10a; H-6 to C-4, C-5, C-6a, C-7, and C-10a; H-7 to C-6, C-9, and C10a; together with H-9 to C-7, C-8, C-10, and C-10a.

5.

Selected HMBC and COSY correlations of compounds 5 and 6.

The presence of the prenyl unit was suggested by the characteristic resonances at δ_H 4.47 (d, J = 6.5 Hz, 2H, H-1′), 5.41 (t, J = 6.5 Hz, 1H, H-2′), 1.70 (s, 3H, H-4′), and 1.77 (s, 3H, H-5′), which was linked through the ether bond at C-10 evidenced by the correlation of H-1′/C-10 (156.6 ppm). The methoxy group was thus located at C-8 (160.6 ppm) suggested by HMBC correlation of 8-OCH₃/C-8. According to the above-described evidence, compound 5 was identified as 10-prenylcannabispirone.

10-O-Prenyl-7-prenylcannabispirone (6) was acquired as a yellow oil. The HRESIMS exhibited a pseudomolecular ion peak at m/z 383.2581 ([M + H]⁺), establishing the molecular formula C₂₅H₃₄O₃. The ¹H and ¹³C NMR data of 6 were similar to those of compound 5, except for the presence of one additional prenyl group, which was supported by a 68 Da (C₅H₈) larger molecular mass. The comparison of ¹H NMR of 5 and 6 consequently suggested that one additional prenyl group substituted at ring C of 6 as evidenced by the remaining singlet proton of H-9 (6.32 ppm). The prenyl group was located at C-7 by a set of HMBC correlations from H-11 (3.23 ppm) to C-6a (144.5 ppm), C-7 (118.0 ppm), and C-8 (157.5 ppm). Therefore, compound 6 was determined as 7,10-diprenylcannabispirone.

Compounds 10 and 11 were isolated as yellow oils. Their HRESIMS exhibited pseudomolecular ion peaks at m/z 263.1641 and 263.1643 respectively ([M + H]⁺), implying the identical molecular formula of C₁₆H₂₂O₃. The ¹H NMR of compound 10 showed characteristic signals of cannabispiranol, ,, including two meta-aromatic protons at δ_H 6.34 (d, J = 1.8 Hz, 1H, H-7) and 6.29 (d, J = 1.8 Hz, 1H, H-9), two coupled methylene groups at δ_H 2.0 (t, J = 7.6 Hz, H-5) and 2.84 (t, J = 7.6 Hz, H-6), four methylene units at δ_H 1.73 (4H, m, H-2), 2.53 (2H, td, J = 13.7, 4.6 Hz, H-3ax), and 1.26 (2H, dt, J = 15.9, 2.4 Hz, H-3eq), one oxygenated methine unit at δ_H 4.06 (1H, t, J = 3.0 Hz, H-1), and one methoxy group at δ_H 3.79 (s). Careful comparison of ¹H NMR data with α-cannabispiranol (12) indicated one additional methoxy group δ_H 3.77 (s) in compound 10, which corresponded to its relatively less hydrophilicity observed on a silica gel TLC profile. Two methoxy groups were located at C-8 and C-10 based on HMBC correlations shown in Figure . Therefore, the whole structure of compound 10 was established. Once the gross structure of compound 10 was determined, we further determined the whole structure of compound 11. Side by side comparison of their ¹H and ¹³C NMR data revealed that they were nearly identical, except for particular positions, C-1, C-2, and C-3. The striking difference in ¹H (0.27–0.73 ppm) and ¹³C NMR (2.9–5.0 ppm) chemical shifts pointed out that they were possibly different in the orientation of 1-OH, which was addressed by NOESY correlations and coupling constant analysis. The diaxial correlation of H-1/H-3ax in the NOESY spectrum of compound 11 assisted in assigning H-1 as an axial proton. On the other hand, the doublet signal with a smaller coupling constant (3 Hz) of oxygenated methine H-1 in 10 in association with the absence of diaxial correlation of H-1 and H-3 indicated equatorial-oriented proton, H-1. Therefore, compounds 10 and 11 were rotational conformers which were assigned as 1β-hydroxy-10-methoxy-cannabispiranol and 1α-hydroxy-10-methoxy-cannabispiranol, respectively.

6.

Selected HMBC, COSY, and NOESY correlations of compounds 10 and 11.

1-Methoxy-8-hydroxycannabinol (CBN-1OMe-8OH, 30) was the last new cannabinoid isolated from leaves specimens. HRESIMS displayed a pseudomolecular ion [M + H]⁺ peak at m/z 341.2112 (calcd. for C₂₂H₂₉O₃, 341.2105) which was perfectly fitted with the molecular formula of C₂₂H₂₈O₃. The ¹H NMR spectrum indicated the presence of four aromatic protons at δ_H 6.42 (d, J = 1.7 Hz, 1H, H-2), 6.46 (d, J = 1.7 Hz, 1H, H-4), 6.67 (s, H-7), 8.18 (s, H-10), n-pentyl side chain at 2.55 (dd, J = 9.5, 7.5 Hz, 2H, H-1′), 1.62 (m, 2H, H-2′), 1.33 (m, 2H, H-3′), 1.32 (m, 2H, H-4′), and 0.89 (t, J = 6.9 Hz, 3H, H-5′) and three singlet methyls at δ_H 2.28 (H-11), 1.56 (H-12), and 1.56 (H-13). These signals suggested the presence of two benzene rings (A and C) possessing an alkyl residue and singlet methyl groups, which is typical of cannabinol (CBN) type compounds. However, the presence of one methoxy group (δ_H 3.92 ppm and δ_C 55.7 ppm) was located at C-1 based on HMBC correlation of 1-OCH₃/C-1 (156.8 ppm) (Figure ). A couple of more downfield aromatic singlets (6.67 and 8.18 ppm) indicated 1,2,4,5-tetrasubstituted benzene and para-oriented proton in ring C. One remaining substation group was verified to be hydroxy phenolic (−OH) based on the typical signal of oxygenated quarternary carbon at 152.8 ppm.

7.

Selected HMBC and COSY correlations of compound 30.

6. ¹H NMR and ¹³C NMR Data of 30 in CDCl₃ .

	(30)
no.	δ H (mult, J  in Hz)	δ C (ppm)
1		156.8
2	6.42 (d, 1.7)	105.2
3		143.5
4	6.46 (d, 1.7)	111.0
4a		153.6
6		76.9
6a		139.5
7	6.67 (s)	109.4
8		152.8
9		122.0
10	8.18 (s)	129.5
10a		120.8
10b		109.9
11	2.28 (s)	16.0
12	1.56 (s)	27.1
13	1.56 (s)	27.1
1′	2.55 (dd, 9.5, 7.5)	36.2
2′	1.62 (m)	30.8
3′	1.33 (m)	31.7
4′	1.32 (m)	22.7
5′	0.89 (t, 6.9)	14.2
1-OMe	3.92 (s)	55.7

Evaluation of α-Glucosidase Inhibition and the Structure–Activity Relationship

In 2022, Suttithumsatid and co-workers evaluated yeast α-glucosidase inhibition of Cannabis sativa extract along with two isolated cannabinoids, Δ⁹-THC and CBD. THC (IC₅₀ = 3.0 μg/mL) was slightly more potent than CBD (IC₅₀ = 5.5 μg/mL), whereas the extracts were more active (IC₅₀ = 0.16–1.2 μg/mL). However, due to the limited availability of pure compounds and the stronger inhibitory activity observed in the crude extracts, structure–activity relationship (SAR) analysis could not be established. To address this gap, the present study expanded the scope of compound isolation by purifying a broader range of secondary metabolites, with the diversity of cannabinoids and spiroindans. This extensive metabolite profiling enabled a more systematic evaluation of α-glucosidase inhibitory activity and allowed for detailed SAR analysis, thereby providing new insights into the structural features responsible for α-glucosidase inhibition.

In this study, the inhibitory effects of 30 isolated metabolites against rat intestinal α-glucosidases were assessed (Figure ). Most cannabinoids, which are key metabolites of Cannabis sativa, demonstrated stronger inhibition (IC₅₀ 0.09–0.8 mM), while spiroindans and simple phenolic compounds exhibited weaker (IC₅₀ 1.0–2.2 mM) or no inhibition.

8.

Summarization of α-glucosidase inhibition of isolated metabolites from Cannabis sativa.

Of 16 isolated cannabinoids, they could be categorized into 3 subclasses according to the core structures namely tricyclic cannabinoids comprising two aromatic rings (3C2Ar), tricyclic cannabinoids comprising one aromatic ring (3C1Ar), and miscellaneous cannabinoids. The 3C2Ar group (e.g., compounds 15, 16, 17, 18, 30) generally revealed stronger inhibition (IC₅₀ 0.09–0.47 mM) than the 3C1Ar group (IC₅₀ 0.31–0.7 mM) and miscellaneous cannabinoids. 8-Hydroxycannabinol (18), the most potent inhibitor of the 3C2Ar group, revealed 3 times significantly stronger inhibition than cannabitriol (23), the most leading inhibitor of the 3C1Ar group. This observation suggested that the occurrence of one additional aromatic ring in the 3C2Ar group played a vital role in retarding enzyme function.

The greater number of phenolic hydroxyls also obviously enhanced an inhibitory effect against α-glucosidases. This assumption was supported by a three-time more potent inhibition of 8-hydroxycannabinol (18), a 3C2Ar containing two phenolic hydroxyls, over the inhibitory effects of cannabinol (15) and 8-hydroxyl-1-methoxycannabinol (30), whose structures comprise one phenolic hydroxyl on ring A or C. The enhanced inhibitory activity caused by additional phenolic hydroxyls was also observed in our previous investigations. ,,− Although the structures of cannabinol (15) and 8-hydroxycannabinol (18) were nearly identical and they also contained one phenolic hydroxyl on ring A, the inhibition of 8-hydroxycannabinol (18) was three times more improved than that of 15, suggesting that the presence of one additional phenolic hydroxyl on ring C of 18 is significantly important. Conversely, the replacement of phenolic hydroxyl on ring A by a methoxy group resulted in three times less inhibition in 30. This finding suggested the equipotency of phenolic hydroxyls on rings A and C in inhibiting enzyme function, possibly through binding with different amino acid sequences of the active site.

3C1Ar was another cannabinoid family that was slightly less potent compared to the 3C2Ar group. Δ⁹-THC (19) is a key representative compound of this family, and it showed more potent inhibition against both rat intestinal maltase and sucrase. The presence of two additional hydroxyls on ring C did not improve the inhibition as expected. On the other hand, the presence of the propyl group in structure of 24 led to the slightly dropped inhibition. The inhibitory effect of 3C1Ar was clearly canceled in 1,2,4-triprenylcannabitriol (2) in which ring A was fully substituted by three prenyl groups. Although the functional groups that enhanced inhibitory activity of the 3C1Ar family could not be clearly pointed out, the occurrence of prenyl groups that led to the reduced inhibition was discovered.

Enzyme Kinetic Study

To elucidate the mechanism of inhibition, a kinetic study employing Lineweaver–Burk plots was performed. 8-Hydroxycannabinol (18) and cannabitriol (23) were the most potent inhibitors of the 3C2Ar and 3C1Ar families, whereas cannabinol (15) was the second most potent inhibitor of the 3C2Ar family distinguished from 18 by the lack of one additional phenolic hydroxyl.

8-Hydroxycannabinol (18) produced a series of Lineweaver–Burk plots, which intersected the X-axis for maltase and sucrase (Figure ). The decrease in maximum reaction velocity (V _max) and unchanged Michaelis constant (K_m) indicated that 18 inhibits maltase and sucrase through noncompetitive manners.

9.

Lineweaver–Burk plots of 15, 18, and 23 against maltase (A, C, and E) and sucrase (B, D, and F).

The secondary plot of inhibitor concentration versus Y intercept revealed the dissociation constant (K_i) of the enzyme–substrate-inhibitor (ESI) complex to be 0.1812 mM for maltase and 0.1258 mM for sucrase. Similarly, cannabitriol (23) inhibited maltase and sucrase in an noncompetitive manner using the same methodology.

In contrast to 8-hydroxycannabinol (18), cannabinol (15) demonstrated a different mechanism of action. The Lineweaver–Burk plots for compound 15 revealed a parallel series of straight lines associated with the decreases in both V _max and K_m as its concentration increased. This finding implied that compound 15 inhibits maltase and sucrase via an uncompetitive mechanism, with K_i ^′ values of 0.1964 and 0.4348 mM, respectively. The kinetic factors are summarized in Table .

7. Kinetic Parameters of 15, 18, and 23 against α-Glucosidases.

	Maltase			Sucrase
	Inhibition type	K i (mM)	K i ′ (mM)	Inhibition type	K i (mM)	K i ′ (mM)
15	Uncompetitive		0.1964	Uncompetitive		0.4348
18	Noncompetitive	0.1812	0.1812	Noncompetitive	0.1225	0.1225
23	Noncompetitive	0.2156	0.2156	Noncompetitive	0.3046	0.3046

The striking difference in the inhibitory effect between cannabinol (15) and 8-hydroxycannabinol (18) would be elaborated on in terms of the inhibition mechanism (Figure ). Once the enzyme–substrate (ES) complex was formed, the inhibitor could retard the generation of the product by associating the ES complex at the allosteric site to afford the ESI complex. This is a typical pathway in which the inhibitor retards the enzyme by noncompetitive, uncompetitive, and mix-manner inhibition. Due to a relatively weaker binding between the enzyme and inhibitor at the allosteric site, the ESI complex can be broken into the ES complex. Although the structures of compounds 15 and 18 were nearly similar, except for one additional phenolic hydroxyl at C-8 in compound 18, cannabinol (15) revealed a larger dissociation constant (K_i′ = 0.1964 for maltase and 0.4348 for sucrase), indicating the weaker complexation of ESI. This mechanism rationalizes the weaker inhibition of 15 compared with 18 and 23 (Figure ).

10.

Putative inhibitory mechanism against α-glucosidases. (A) Uncompetitive inhibition and (B) noncompetitive inhibition.

Molecular Docking and Binding Energy Calculations

To gain insight into interactions between two potent inhibitors (15 and 18) and active sites of α-glucosidase, their molecular docking was explored. The RMSD and standard deviation (SD) values for the protein backbone, amino acids within 10 Å of the binding site, and ligand atoms analyzed across three independent MD simulation runs for the rat-ntMGAM–compound 15 and rat-ntMGAM–compound 18 systems are summarized in Table . The table also lists hydrophobic residues involved in ligand interactions, extracted from the three MD runs. For compound 15, the backbone RMSD ranged from 1.80 ± 0.11 Å to 2.37 ± 0.08 Å, while the RMSD of amino acids within the binding site varied between 0.83 ± 0.08 Å and 1.13 ± 0.10 Å. The ligand RMSD values were between 1.11 ± 0.18 Å and 1.18 ± 0.18 Å, indicating overall stability in the binding pocket. For compound 18, the backbone RMSD ranged from 1.90 ± 0.09 Å to 2.40 ± 0.08 Å, while binding site residues showed RMSD values between 0.80 ± 0.06 Å and 1.07 ± 0.08 Å. Ligand RMSD values varied between 1.27 ± 0.13 Å and 1.82 ± 0.22 Å, suggesting relatively stable ligand-protein interactions.

8. RMSD and SD Values for the Protein Backbone, Amino Acids within 10 Å of the Binding Site, and Ligand Atoms from Three Independent MD Simulation Runs for the Rat-ntMGAM–Compound 15 and rat-ntMGAM–Compound 18 Systems .

		RMSD (Mean ± SD)
Systems		Backbone	Binding site	Ligand	Residues
rat-ntMGAM-cdp 15	RUN1	2.15 ± 0.08	0.88 ± 0.08	1.12 ± 0.16	A277, L278, P279, I498, A501, V504, I514, F515, F527, A528
	RUN2	1.80 ± 0.11	0.83 ± 0.08	1.18 ± 0.18	A277, I498, A501, V504, F514, I515, F527
	RUN3	2.37 ± 0.08	1.13 ± 0.10	1.11 ± 0.18	A277, I498, A501, V504, I515, F527
rat-ntMGAM-cdp 18	RUN1	1.90 ± 0.09	0.97 ± 0.07	1.81 ± 0.28	A277, L278, P279, I498, A501, V504, I515, F527
	RUN2	2.40 ± 0.08	1.07 ± 0.08	1.27 ± 0.13	A277, L278, P279, I498, A501, V504, F514, I515, F527, A528
	RUN3	1.93 ± 0.09	0.80 ± 0.06	1.82 ± 0.22	P276, A277, L278, I498, A501, V504, I515, F527

^aThe table also lists hydrophobic residues involved in ligand interactions, extracted from the three MD runs.

Figure A illustrates the MD snapshots of the rat-ntMGAM–compound 15 and rat-ntMGAM–compound 18 complexes, showing that both biologically active cannabis compounds bind stably within the same binding pocket. Notably, this binding site was distinct from the substrate-binding site, reinforcing the hypothesis that these compounds do not act through competitive inhibition. The stable binding observed in MD simulations was consistent with the experimentally determined uncompetitive or noncompetitive inhibition mechanisms, further supporting the proposed mode of action of these ligands.

11.

(A) Molecular dynamics (MD) snapshots of the rat-ntMGAM–compound 15 complex (left) and the rat-ntMGAM–compound 18 complex (right), illustrating ligand binding interactions. (B) Venn diagram displaying shared (orange) and unique amino acid residues within 10 Å of compound 15 (pale red) and compound 18 (green) in the protein. The numbers indicate the residue count in each category. (C) Venn diagram highlighting hydrophobic residues of the protein that are in close contact with compound 15 (left) and compound 18 (right).

A Venn diagram summarizes the total number of the amino acid residues within 10 Å of compound 15 and compound 18 in the rat-ntMGAM binding pocket (Figure B). A total of 70 residues (orange) were shared between both complexes, indicating a common interaction environment and suggesting a similar binding mode. However, 13 residues were unique to compound 15 (pale red), while 8 residues were unique to compound 18 (green), highlighting subtle differences in local interactions. These variations in local residue interactions may contribute to variations in binding affinity or inhibitory mechanisms. The residue count in each category suggested that while both compounds occupy the same binding site, they may stabilize distinct interaction networks within the protein.

Analysis of hydrophobic interactions from the three MD runs identified key residues involved in stabilizing compounds 15 and 18 within the binding pocket (Figure C and Table ). Residues A277, I498, A501, V504, I515, and F527 were consistently engaged in hydrophobic interactions across multiple MD runs in both ligand-bound systems, indicating their critical role in ligand stabilization. The overlap of these residues suggested a shared hydrophobic interaction environment, implying that both compounds adopt similar stabilizing interactions within the binding pocket. However, compound-specific differences were observed, with L278, P279, F514, and A528 interacting uniquely with compound 15, while P276, L278, P279, and F514 were specific to compound 18. These variations highlight subtle distinctions in how each compound engages with the surrounding protein environment, potentially influencing their binding affinities and inhibitory mechanisms. The persistence of these interactions across all three MD runs supports the proposed binding pocket as a viable cannabis-binding site. This finding is in strong agreement with experimentally observed uncompetitive and noncompetitive inhibition mechanisms.

Figure A illustrates the binding poses of compounds 15 and 18 within the predicted binding pocket of rat-ntMGAM. Both compounds stably occupied the same pocket, engaging in extensive hydrophobic interactions with surrounding residues. The molecular surface representation highlighted the spatial accommodation of the ligands within the binding cavity. Notably, hydrophobic residues identified from MD simulations, including A277, I498, A501, V504, I515, and F527, were consistently observed to interact with both ligands, further supporting a shared binding environment. Figure B presents residues involved in hydrogen bonding interactions between rat-ntMGAM and the two compounds. In the compound 15–rat-ntMGAM complex (Figure B, left), a single hydrogen bond was observed, where the phenolic hydroxyl (−OH) group of compound 15 formed a hydrogen bond with the backbone oxygen of A501. In contrast, in the compound 18–rat-ntMGAM complex (Figure B, right), two hydrogen bonds were presented: one between the phenolic hydroxyl group of compound 18 and the backbone oxygen of K526 and another between an additional 8-hydroxy group and the backbone oxygen of L278.

12.

(A) Binding poses of compounds 15 (left) and 18 (right) within the predicted binding pocket of rat-ntMGAM. Compounds are shown in CPK representation, while the binding pocket is displayed as a molecular surface. Labeled residues indicate hydrophobic interactions identified from MD simulations that are shared between both compounds. (B) Residues involved in hydrogen bonding interactions with compound 15 (left) and compound 18 (right), as identified from MD simulations. Both compounds in (B) are shown in the same orientation as in (A) for direct comparison.

Interestingly, based on the orientation of the ether oxygen (−O−) in the ring of CBN, it appeared that the aromatic ring of compound 15 is flipped by approximately 180° compared to compound 18. This inversion in orientation likely contributed to the observed differences in hydrogen bonding interactions and may influence the stability and inhibitory properties of each ligand within the binding pocket. These findings aligned with the MD-derived interaction profiles, reinforcing the role of hydrogen bonding in stabilizing ligand binding and providing structural insights into the differential inhibitory mechanisms of these compounds.

The results summarized in Table provide a comprehensive comparison of the experimental binding activities (IC₅₀ values) and calculated binding affinities across multiple computational approaches, including docking scores, semiempirical AM1 screening, and post-MD binding energy calculations using PM6, DFT, and MMGBSA methods.

9. Experimental IC₅₀ Values for Binding Activity and Computational Results for Docking and Binding Energy Calculations of Isolated Compounds .

			ΔE bind (kcal/mol)
				After MD
Compound	IC 50 (mM)	Docking score (kcal/mol)	Before MD AM1	PM6	DFT	MMGBSA
15	0.35	–7.41	1.41	–4.18	–3.57	–25.78 ± 4.69
18	0.09	–7.68	0.00	–4.79	–5.74	–29.62 ± 4.52
23	0.31	–7.16	3.16	-	-	-
2	na	–7.07	10.59	-	-	-
6	na	–8.19	21.61	1.44	8.16	-
7	2.2	–7.26	26.97	-	-	-

^aDocking scores were obtained from Autodock vina, while ΔE_bind values were calculated before and after MD simulations. The AM1 method was used for initial screening, and binding energy refinements were performed using PM6, DFT, and MMGBSA approaches after MD. Compounds 6 and 7 exhibited no measurable activity.

The IC₅₀ values indicate that compounds 15, 18, and 23 exhibit strong binding activities, (IC₅₀ values of 0.35 mM, 0.09 mM, and 0.31 mM, respectively). In contrast, compounds 6 and 7 showed no measurable activity, consistent with their inclusion as negative controls. Docking scores, calculated using AutoDock Vina, provided an initial estimate of binding affinity. These scores revealed relatively favorable binding poses for all compounds, with compound 6 achieving the lowest docking score (−8.19 kcal/mol). However, docking scores alone did not fully align with experimental IC₅₀ values, particularly for compound 6, whose strong docking score did not reflect its lack of observed activity. This discrepancy underscores the limitations of docking methodologies, which do not fully account for protein flexibility, solvent effects, and entropic contributions that influence ligand binding in dynamic environments.

The pre-MD AM1 energies provided a relative ranking of binding affinities that better correlated with experimental bioactivities, despite their generally positive values due to initial geometric inconsistencies from the docking step. Compounds 15, 18, and 23, which exhibited strong experimental activity, displayed small lower AM1 binding energies (1.41, 0.00, and 3.16 kcal/mol, respectively). In contrast, compounds with poor experimental activity, such as 6 and 7, showed substantially higher values (10.59 and 21.61 kcal/mol, respectively). This trend suggested that the pre-MD AM1 offers a more reliable ranking of relative activity trends compared to docking alone.

Binding energy calculations performed after MD simulations using PM6 and DFT methods showed improved agreement with experimental data. Compound 18, which demonstrated the highest bioactivity experimentally, also had the most negative ΔE_bind (−4.79 kcal/mol from PM6 and −5.74 kcal/mol from DFT), indicating strong and stable binding at the predicted CBN-binding site. Conversely, compound 6 exhibited highly positive binding energies, further reflecting its lack of interaction with the target and confirming its classification as inactive. The MMGBSA calculations provided additional validation, reinforcing the observed binding trends. Compounds 15 and 18, which were the most potent inhibitors, showed the lowest MMGBSA-derived binding energies (−25.78 ± 4.69 kcal/mol and −29.62 ± 4.52 kcal/mol, respectively), demonstrating strong and stable interactions within the binding pocket.

Overall, the integration of MD simulations and high-level binding energy calculations (PM6, DFT, MMGBSA) significantly enhanced the accuracy of binding affinity predictions compared to docking alone. These results underscore the importance of incorporating dynamic and solvation effects to better capture the complexity of protein–ligand interactions. The strong correlation between computational predictions and experimental data supports the effectiveness of this multistep computational approach for screening potential inhibitors and prioritizing candidates for further experimental validation.

This study reveals that the leaves of Cannabis sativa are a prolific source of α-glucosidase inhibitors with cannabinoids, particularly cannabinol derivatives exhibiting the most pronounced bioactivity. Among 30 isolated compounds, 8-hydroxycannabinol demonstrated the strongest inhibitory effect, outperforming other cannabinoids through an uncompetitive inhibition mechanism. Structure–activity relationship analysis indicated that additional phenolic hydroxyl groups and extended aromatic rings substantially enhance inhibitory potency. Molecular docking and molecular dynamics simulations confirmed that active cannabinol derivatives bind stably to an allosteric site on the α-glucosidase enzyme, supported by favorable binding energies and consistent hydrophobic interactions. However, the limited structural diversity of the isolated compounds restricts the full elucidation of SAR trends, underscoring the need for broader analog libraries. These findings not only highlight the therapeutic potential of C. sativa leaf cannabinoids as natural antidiabetic agents but also lay a foundation for future research. In particular, the semisynthesis of novel cannabinol derivatives via rational modifications such as halogenation or esterification represents a promising strategy to enhance α-glucosidase inhibitory activity and to systematically probe the structure–activity relationship of this pharmacophore class.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jafc.5c08443.

• Isolation schemes; Tables of α-glucosidase inhibition of crude extracts, subfractions, and all isolated compounds; HRESIMS, ¹H, ¹³C, COSY, HSQC, HMBC, NOESY, and UV–vis spectra of new compounds (1-6 and 10, 11, and 30); Enzyme model with allosteric site; Secondary plots from primary Lineweaver–Burk plots (PDF)

The authors declare no competing financial interest.