Three New Minor Compounds from the Roots of Rehmannia glutinosa and Their In Silico and In Vitro AMPK Activation

1. Introduction

Rehmannia glutinosa Libosch is an herb in the Orobanchaceae family and found in China, Japan, and Korea [1]. The chemical classes of the compounds typically found in this plant are iridoids, ionones, phenylethanoids, and phenylpropanoids [2]. R. glutinosa is used in traditional medicine to treat excessive heat from the body and unbalanced water metabolism in the kidney [3]. In modern medicine, it has been reported to have hypoglycemic, anti-inflammatory, angiogenetic, and hepatoprotective effects [1]. As anti-diabetic and anti-obese effects of R. glutinosa have drawn attention among researchers [4,5,6,7,8], finding biologically active ingredients and mechanism of actions have become an fascinating areas of interest.

AMP-activated protein kinase (AMPK) is a cellular energy sensor and activated by energy-lowering changes in the body, such as malnutrition and hypoxia [9]. AMPK plays a critical role in maintaining energy homeostasis during metabolic stress. As the functional aspects of AMPK have been revealed to be related to glucose/lipid homeostasis, body weight, food intake, insulin signaling and mitochondrial biogenesis, AMPK has become a major therapeutic target for the treatment of metabolic diseases, including type 2 diabetes and obesity.

In silico simulations help predict the potential of a constituent of a drug by calculating the interaction of a ligand with its receptor, and the absorption, distribution, metabolism, excretion, and toxicity (ADMET) of a drug. Molecular dynamics can also be applied through in silico docking to calculate the behavior of molecules over time by predicting system energy and behavior. These approaches can predict a drug target, such as receptors and enzymes, and behavior of a drug in the body. Appropriate in vitro or in vivo investigations should be requested to support the results from in silico simulations.

In this study, compounds in R. glutinosa were isolated to identify active ingredients. Structures were identified using 1D/2D nuclear magnetic resonance (NMR) and high-resolution electrospray ionization mass spectrometry (HR-ESIMS). Binding energies of the isolated compounds and AMPK receptor were calculated using in silico docking simulations. Moreover, a full set of molecular dynamics and ADMET predictions were performed on compound 1 to predict its AMPK-activating effect. The results from the simulation were also explored by an in vitro AMPK-activating assay.

2. Results

2.1. Structure Elucidation

Compound 1 was obtained as a brown gum. HR-ESIMS [M + 2H₃O]²⁺ at m/z 236.1232 and ¹³C NMR data suggested a molecular formula of C₁₇H₂₆N₂O₁₁. Compound 1 showed characteristic signals for uridine in the ¹H- and ¹³C-NMR spectra as follows: δ_C 152.5 (C-2); 166.3 (C-4); δ_H 5.69 (1H, d, J = 8.0 Hz)/102.7 (CH-5); 8.00 (1H, d, J = 8.0 Hz)/142.8 (CH-6); 5.90 (1H, d, J_1',2' = 4.7 Hz)/90.8 (CH-1'); 4.17 (1H, ddd, J_1',2' = 4.7 Hz, J_2',3' = 4.6 Hz)/75.8 (CH-2'); 4.15 (1H, ddd, J_2',3' = 4.6 Hz)/71.4 (CH-3'); 4.00 (1H, dt, J = 4.5, 3.0 Hz)/86.4 (CH-4'); 3.83 (1H, dd, J = 12.2, 3.0 Hz) and 3.73 (1H, dd, J = 12.2, 3.0 Hz)/62.3 (CH₂-5') [10,11]. H-2' and H-3' included peaks that skewed toward each other in the ¹H-NMR spectrum, and their J value, along with the dihedral angles (−33.1°) predicted by Chem 3D, also supported a moiety associated with uridine. The presence of β-glucopyranoside was supported by ¹H- and ¹³C-NMR data and the attachment of this moiety was determined to be the 3'-OH position of uridine ribofuranose by the HMBC correlation between H-6''a (δ_H 3.66 ppm) and C-3' (δ_C 71.4 ppm) (Figure 2) [12]. The three-bond correlation between CH₃CH₂O-1" and C-1" in the HMBC spectrum supported the assignment of the ethyl group in the structure to be an anomeric hydroxyl group of glucopyranoside (Figure 2). In the NOESY spectrum, the correlation between H-1' and H-4' supported the aglycone group as uridine, and the cross-peaks between H-1'', H-3'', and H-5'' suggested the presence of glucopyranoside (Figure 3). The circular dichroism (CD) spectrum of compound 1 exhibited a positive Cotton effect at 273 nm and a negative Cotton effect at 220 and 243 nm, which was similar with the CD spectroscopy trend of uridine [13]. Compound 1 was therefore determined to be 1-O-ethyl-β-D-glucopyranosyl-(6→3')-uridine (Figure 1).

The molecular formula of compound 2 was determined to be C₁₁H₁₈O₅ based on the HR-ESIMS [M + Na]⁺ ion at m/z 253.1051. The ¹H-NMR spectrum of compound 2 showed three methyls [δ_H 1.52 (3H, s, H-10), 1.05 (3H, s, H-12), and 1.04 (3H, s, H-11)], two methylenes [δ_H 1.93 (1H, m, H-3b), 1.90 (1H, m, H-4b), 1.82 (1H, m, H-4a), and 1.37 (1H, dddd, J = 11.4, 6.7, 2.1 Hz, H-3a)], and two methines [δ_H 4.20 (1H, s, H-7) and 4.06 (1H, d, J = 5.6 Hz, H-5)]. A ¹³C-NMR spectrum revealed signals of 11 carbons, including a carbonyl carbon at δ_C 174.8 (C-9) and four oxygenated carbons — two quaternary carbons at δ_C 92.1 (C-8) and 85.8 (C-1) and two methine carbons at δ_C 81.9 (C-5) and 81.4 (C-7). Key correlations of COSY were shown between H-3a/H-4a and H-4b/H-5 (Figure 2). In addition, the correlations of H-3a/C-1, H-4b/C-2 and C-8, H-5/C-1, C-3, C-7, and C-8, and H-7/C-8 in the HMBC spectrum provided crucial information to elucidate the structure of compound 2 as an O-bridged bicyclic skeleton (Figure 2). The HMBC correlation from H-7 to C-9 is indicative of the attachment of a carbonyl group at C-7. Correlations of H-11/C-12, H-12/C-1, C-2, C-3, and C-11, and H-10/C-1, C-5, and C-8 in the HMBC spectrum suggested the attachment of three methyl groups at C-2 and C-8. The key nuclear Overhauser effect (NOE) correlations between H-7/H-3b and H-12, H-4a/H-10 and H-11, H-5/H-10, and H-10/H-11 helped to assign the relative configuration of compound 2, confirming H-3a, H-4a, H-5, H-10, and H-11 oriented in the same direction (Figure 3). In a further study, the energy-minimization of compound 2, which was determined using Chem3D software, corresponded with those of the experimental results. Compound 2 was named 1,8-dihydroxy-2,2,8-trimethyl-6-oxabicyclo[3.2.1]octane-7-carboxylic acid.

A molecular formula of C₁₀H₁₄O₄ was established for compound 8 based on the molecular ion peak at m/z 100.0519 [M + 2H]²⁺ derived from HR-ESIMS. The ¹H- and ¹³C-NMR signals exhibited two methyls [δ_H 2.33/δ_C 27.6 (CH₃-7) and 2.11/13.8 (CH₃-8)], two methylenes [δ_H 3.39/δ_C 43.5 (CH₂-1') and 3.64/61.5 (CH₂-2')], and three olefinic groups [δ_H 7.56/ δ_C 139.1 (H-4), 6.92/131.7 (H-3), and 6.38/135.1 (H-5)]. This suggests that three protons (H-3, 4, and 5) are connected in a trans-form diene based on the J values of the three olefinic protons (J_3,4 = 11.5 Hz and J_4,5 = 15.6 Hz) and NOESY correlations between H-3/H-5 (Figure 2 and 3). In the HMBC spectrum, the attachments of carbonyl carbons at C-2 and C-5 were supported by the correlations from H-4, H-5, and H-7 to C-6 and from H-3 and H-8 to C-1, respectively (Figure 2). The position of an ethyl moiety was assigned by the HMBC correlations between H-1' and C-1, and the downfield-shifted chemical shift value of C-2' proved the existence of a hydroxyl group at C-2'. This was also supported by HR-ESIMS fragmentation traces. A thorough analysis of 1D and 2D NMR spectra of compound 8, revealed it to be 2'-hydroxyethyl (2E,4E)-2-methyl-6-oxohepta-2,4-dienoate, which is a structure similar to organic compounds from the reaction mechanisms and synthetic pathways in the previous study (Figure 1) [14].

Along with three newly identified compounds, 16 previously known compounds were isolated (Figure 1): dihydroxy-β-ionone (3) [15], rehmaionoside C (4) [15], rehmaionoside A (5) [16], frehmaglutin J (6) [17], frehmaglutin I (7) [17], tianshic acid (9) [18], catalpol (10) [19], geniposide (11) [20], frehmaglutoside G (12) [21], 6-O-sec-hydroxyaeginetoyl ajugol (13) [22], salidroside (14) [12], acteoside (15) [23], jionoside A₁ (16) [24], leucosceptoside A (17) [25], martynoside (18) [26], jionoside B₁ (19) [24], isomartynoside (20) [27], and sucrose (21) [28]. Compound 9 was found for the first time in the R. glutinosa species.

2.2. In Silico Simulation

As various approaches to in silico docking simulations were explored, the binding energies of compounds 1–21 were predicted against AMPK by three software programs (Autodock Vina, Autodock 4, and Dock 6). Among the compounds, 1 (−11.4 Kcal/mol) showed stronger affinities with AMPK than did those of the control (5-amino-4-imidazolecarboxamide ribonucleoside, AICAR), as shown in Table 1. Compound 1 formed a higher number of hydrogen bonds with AMPK than did the control, indicating more stable interactions of compound 1 with AMPK. It also exhibited hydrophobic/van der Waals and aromatic interactions with AMPK. In other words, compound 1 has the potential for strong activity against AMPK in the molecular interaction models (Figure 4).

In molecular dynamics, the root mean square deviation (RMSD) is an indicator of the stability of a protein complex. The smaller the RMSD , the more stable the interaction [29]. Compound 1 usually showed smaller RMSD values compared with those of AICAR during the simulation (10 ns), as shown in Figure 5a, which means that compound 1 forms a more stable conformation with AMPK. Root mean square fluctuation (RMSF) quantifies how much a ligand moves from its average position over time in a molecular dynamics simulation [30]. As compound 1 shared similar fluctuation patterns (red circle) in the several residue index, it can have a similar activity of the AICAR against AMPK (Figure 5b). The number of hydrogen bonds of AICAR and compound 1 against AMPK were predicted. Although AICAR usually forms more hydrogen bonds with AMPK for 10 ns, compound 1 established the same or greater number of H-bonds (Figure 5c). This indicates that AICAR is a strong activator of AMPK and compound 1 also demonstrates strong interactions and stability when it interacts with AMPK. In silico physicochemical and ADMET profiling of compound 1 is shown in Table S1.

2.3. In Vitro Assay

Serial concentrations of compound 1 (5, 10, and 20 μM) were treated in HepG2 cells to investigate its effects against AMPK. Compound 1 increased the protein level of AMPK (Figure 6a). It also stimulated the production of AMPK at concentrations of 2, 4, and 8 μM, as shown in Figure 6b.

3. Discussion

In this study, three new compounds were isolated from R. glutinosa: 1-O-ethyl-β-D-glucopyranosyl-(6→3')-uridine (compound 1), 1,8-dihydroxy-2,2,8-trimethyl-6-oxabicyclo[3.2.1]octane-7-carboxylic acid (compound 2), and 2'-hydroxyethyl-(2E,4E)-2-methyl-6-oxohepta-2,4-dienoate (compound 8). Both compounds 1, a glycosylated uridine, and compound 2, an O-bridged bicyclo-octane, are rare in nature. Iridoid glycosides (compounds 10–13) and phenylpropanoid glycosides (14–20), which are found widely in this species, were also isolated in this study. Among the known compounds, compound 9 was isolated for the first time in this study. Catalpol (compound 10) reportedly exhibits anti-inflammatory and antioxidant activities [31], and geniposide (11) has shown protective effects on cerebral diseases [32]. Moreover, salidroside (compound 14) has shown beneficial effects against ischemic diseases [33], and acteoside (compound 15) has been reported to have cardioprotective, anti-diabetic, and neuroprotective effects [34].

To the best of our knowledge, this is the first report of in silico docking simulations of compounds 1–21 against AMPK as a receptor. Among the ingredients in R. glutinosa, catalpol (compound 10) has been studied using in silico docking simulations as a ligand against the glutathione peroxidase receptor due to its neuroprotective effect [35], and the involvement of rhein in the absorption of rehmannioside D was investigated using in silico approaches [36]. This study provides a full set of in silico molecular-docking simulations, along with the binding energies between the isolated compounds and AMPK, the binding interactions of compound 1 and AMPK, the molecular dynamics of compound 1, and the ADMET of compound 1. In molecular dynamics, compound 1 was suggested as an effective AMPK activator because its RMSD and RMSF values and its number of hydrogen bonds were similar or higher than those of AICAR, a widely known AMPK activator.

The in vitro assay was performed to confirm the AMPK-producing activity of compound 1, as measured in the in silico docking simulation results. As expected, compound 1 increased the protein level of AMPK at 5, 10, and 20 μM. It also stimulated the protein expression of AMPK at 2, 4, and 8 μM. In a previous study, catalpol (10) decreased blood glucose levels and mitigated insulin resistance by recruiting the AMPK signaling pathway [37,38]. A more detailed activity study of compound 1 will therefore be needed to meet the goal as a treatment for metabolic diseases by identifying the mechanisms of action and structure-activity relationships, along with in vivo AMPK-activating effects.

Although R. glutinosa has been widely investigated, its chemical library can be expanded by isolating structures and chemical profiling of the extract. Various receptor targets can be selected to determine the therapeutic potential of R. glutinosa. Well-organized in vitro and in vivo assays will also help determine the potential of the plant to serve as a therapeutic agent for the treatment of metabolic diseases, such as diabetes mellitus and obesity.

4. Materials and Methods

4.1. General Experimental Procedures

Optical rotation data and ultraviolet (UV) spectra were obtained on a JASCO P-2000 polarimeter (Tokyo, Japan) and a Hitachi U-3000 UV/visible-light spectrophotometer (Tokyo, Japan), respectively. Infrared spectra were recorded on a Nicolet iS10 FT-IR spectrometer (Thermo Fisher, MA, USA). NMR spectra were recorded on an Agilent 400 MHz Fourier-transform NMR instrument (Agilent Technologies, Santa Clara, CA, USA) and analyzed in MestreNova 9.0.0 software (Mestrelab Research S.L., Santiago de Compostela, Spain). HR-ESIMS was acquired on an Agilent 6230 time-of-flight liquid chromatography (LC)/mass spectrometer (Agilent Technologies). Adsorption column chromatography was performed using a silica gel (63–200 μm, Merck, Darmstadt, Germany). Medium-pressure LC was chromatographed on a CombiFlash Rf-200 instrument (Teledyne Isco, Lincoln, NE, USA) and RediSep Silver Silica Gel Disposable Flash Columns of 330.0 g and 24.0 g (Teledyne Isco) were used for separations. Preparative high-performance liquid chromatography (MPLC) was carried out on a YMC-Pack Pro C 18 column (20 × 250 mm, 5 μm, YMC Co., Kyoto, Japan) using a Waters 600 pump and a Waters 996 photodiode array detector (Waters, MA, USA).

4.2. Plant Material

Dried roots of R. glutinosa were purchased from the Nonglim Saengyak Company in Seoul, South Korea in January 2020. A voucher specimen (no. EA388) was deposited at the Natural Product Chemistry Laboratory, College of Pharmacy, Ewha Womans University.

4.3. Extraction and Isolation

The ground root of R. glutinosa (9 kg) was extracted with MeOH (13 × 5 L) by maceration and dried under reduced pressure to produce 987 g of crude extract. This extract was partitioned with hexanes (10 L, 16 g), EtOAc (10 L, 9 g), and n-BuOH (80 L, 90 g) sequentially.

The ethyl acetate partition was fractionated on a silica-gel column (CH₂Cl₂-MeOH) to produce 10 fractions (E01–E10). Fractions E04 and E05 (530 mg) were loaded together on an MPLC column to produce 10 subfractions (E0401–E0410). Among these, E0403 (30 mg) was purified using a silica gel to produce 15 subfractions (E040301–E040315). Of these, subfraction E040305 (10.5 mg) was chromatographed using semi-preparative reverse phase (RP) C-18 columns with MeOH-H₂O (55:45) used as a solvent to yield compounds 2 (2.9 mg, t_R= 37.7 min), 6 (2.0 mg, t_R= 58.9 min), and 7 (0.7 mg, t_R= 63.4 min). Subfraction E040309 (125 mg) was purified in a semi-preparative a RP C-18 column using MeCN-H₂O (47:53) as a solvent to produce compound 3 (17.3 mg, t_R= 102.7 min). Subfraction E06 was subjected to a MPLC column (CH₂Cl₂-MeOH) to yield eight subfractions (E0601-E0608). Subfraction E0608 (991 mg) was separated on a silica-gel column (hexane-Acetone) to produce 10 fractions (E060801-E060810), followed by purification of E060807 (55 mg) over a semi-preparative RP C-18 column (MeOH-H₂O, 10 to 100) to produce compound 8 (0.9 mg, t_R= 19.6 min). Fraction E08 (1.8 g) was chromatographed on a MPLC column (CH₂Cl₂-MeOH) to yield 19 subfractions (E0801–E0819). Subfraction E0808 (249 mg) was eluted on an MPLC column (CH₂Cl₂-acetone, 100% to 0%) to afford nine subfractions (E0808701-E080809). Fraction E080806 was purified using a semi-preparative RP C-18 column, with MeOH-H₂O (75:25) to produce compound 9 (11.3 mg, t_R= 40.0 min). E0811 was fractionated using a semi-preparative RP C-18 column (MeOH-H₂O, 65:35) into four subfractions (E081101-E081104). Subfraction E081102 was purified on ODS-A and semi-preparative (MeOH-H₂O, 45:55) C-18 columns to afford compounds 4 (0.9 mg, t_R= 58.2 min), 18 (16.7 mg, t_R= 25.8 min), and 20 (17.6 mg, t_R= 32.7 min). Subfraction E0814 (260 mg) was chromatographed on RP C-18 column to produce 19 subfractions (E081401-E081419). Among these, E081401 was subjected to purification over a semi-preparative RP C-18 column using MeOH-H₂O (25:75) as a solvent to produce compounds 1 (1.7 mg, t_R= 20.3 min) and 14 (8.2 mg, t_R= 45.4 min). E081402 and E081404 were then subjected to a semi-preparative RP C-18 column (MeOH-H₂O, 25:75) to produce compounds 11 (1.7 mg, t_R= 45.7 min) and 5 (13.1 mg, t_R= 36.1 min), respectively. E081403 was chromatographed over a Sephadex-LH20 and compound 17 (30.5 mg) was precipitated from one of the subfractions. Fraction E09 (2.2 g) was chromatographed on a MPLC column to yield 12 subfractions E0901-E0912). One of these fractions, E0906 (330 mg), was then subjected to Sephadex LH-20 and semi-preparative RP C-18 column (CH₃CN-H₂O, 60:40) chromatography to afford compound 13 (35 mg, t_R= 37.2 min).

The n-butanol partition was loaded on a silica-gel column (CH₂Cl₂-MeOH) to fractionate into 12 fractions (B01-B12), and compounds 10 (445 mg) and 21 (586 mg) were precipitated from the fractions B08 and B09, respectively. B07 (39 g) was chromatographed on a silica gel (CH₂Cl₂-MeOH-H₂O and 100% MeOH) to produce 12 subfractions (B0701-B0712). Subfraction B0710 (9.1 g) was subjected to an octadecylsilyl (ODS) column (MeOH-H₂O, 0-100%) to yield 21 subfractions (B071001-B071021). Subsequent separation of B071015 (108 mg) over a semi-preparative RP C-18 column using MeOH-H₂O (32:68) as solvent yielded compounds 15 (20.5 mg, t_R= 97.4 min) and 16 (22.4 mg, t_R= 46.9 min). Compound 19 (10.2 mg, t_R= 68.6 min) was purified from B071018 on a semi-preparative RP C-18 column (MeOH-H₂O, 35:65). Subfraction B071011 (106 mg) was followed up by charging onto an ODS column (MeOH-H₂O, 10-100%) to afford 11 subfractions (B07101101- B07101111). Subfraction B07101106 (35.0 mg) was eluted on a silica gel to yield 10 subfractions (B0710110601–B0710110610), followed by further separation of B0710110608 (20.4 mg) over a semi-preparative RP C-18 column (MeOH-H₂O, 40 %), to yield compound 12 (6.2 mg, t_R= 164.2 min).

1-O-Ethyl-β-D-glucopyranosyl-(6→3')-uridine (compound 1): Brown gum; [α]²⁰_D -8.0 (c 0.5,MeOH); UV (MeOH) λ_max (log ε) 262 (2.78), 201 (2.88) nm; IR (KBr) ν_max 3587, 2962, 1647, 1260, 1031 cm^-1; ¹H NMR (CD₃OD, 400 MHz) δ_H 5.69 (1H, d, J=8.0 Hz, H-5), 8.00 (1H, d, J=8.0 Hz, H-6), 5.90 (1H, d, J=4.7 Hz, H-1'), 4.17 (1H, ddd, J_1',2'=4.7 Hz, J_2',3'=4.6 Hz, H-2'), 4.15 (1H, ddd, J_2',3'=4.6 Hz, H-3'), 4.00 (1H, dt, J=4.5, 3.0 Hz, H-4'), 3.73 (1H, dd, J=12.2, 3.0 Hz, H-5'a), 3.83 (1H, dd, J=12.2, 3.0 Hz, H-5'b), 4.26 (1H, d, J=7.8 Hz, H-1''), 3.16 (1H, dd, J=9.2, 7.8 Hz, H-2''), 3.34 (1H, d, J=2.0 Hz, H-3''), 3.27 (1H, d, J=2.0 Hz, H-4''), 3.26 (1H, dt, J=2.1, 0.8 Hz, H-5''), 3.66 (1H, dd, J=12.0, 5.4 Hz, H-6''a), 3.86 (1H, dd, J=12.0, 2.1 Hz, H-6''b), 3.61 (1H, dd, J=9.5, 6.9 Hz, 1''-O-CH₂CH₃), 3.96 (1H, dd, J=9.5, 6.9 Hz, 1''-O-CH₂CH₃), 1.23 (3H, t, J=6.9 Hz, 1''-O-CH₂CH₃);¹³C NMR (CD₃OD, 100 MHz) δ_C 152.5 (C-2), 166.3 (C-4), 102.7 (C-5), 142.8 (C-6), 90.8 (C-1'), 75.8 (C-2'), 71.4 (C-3'), 86.4 (C-4'), 62.3 (C-5'), 104.2 (C-1''), 75.2 (C-2''), 78.2 (C-3''), 71.7 (C-4''), 78.0 (C-5''), 62.8 (C-6''), 66.2 (1''-O-CH₂CH₃), 15.5 (1''-O-CH₂CH₃); HR-ESIMS m/z 236.1232 [M + 2H₃O]²⁺ (calculated for C₁₇H₃₂N₂O₁₃, 472.1893); CD (MeOH) λ (Δε) 220 (−2.99), 243 (−3.09), 273 (+5.47) nm.

1,8-Dihydroxy-2,2,8-trimethyl-6-oxabicyclo[3.2.1]octane-7-carboxylic acid (compound 2): amorphous white powder; [α]²⁰_D +52.0 (c 0.5,MeOH); UV (MeOH) λ_max (log ε) 196 (3.12) nm; IR (KBr) 3401, 2966, 1772, 997 cm^-1; ¹H NMR (CD₃OD, 400 MHz) δ_H 1.37(1H, dddd, J=11.4, 6.7, 2.1 Hz, H-3a), 1.93(1H, m, H-3b), 1.82(1H, m, H-4a), 1.90(1H, m, H-4b), 4.06(1H, d, J=5.6 Hz, H-5), 4.20(1H, s, H-7), 1.52(3H, s, H-10), 1.04(3H, s, H-11), 1.05(3H, s, H-12); ¹³C NMR (CD₃OD, 100 MHz) δ_C 85.8 (C-1), 38.6 (C-2), 33.6 (C-3), 25.6 (C-4), 81.9 (C-5), 81.4 (C-7), 92.1 (C-8), 174.8 (C-9), 13.3 (C-10), 24.4 (C-11), 27.9 (C-12); HR-ESIMS m/z 253.1051 [M + Na]⁺ (calculated for C₁₁H₁₈O₅Na, 253.1046); CD (MeOH) λ (Δε) 199 (+15.18), 228 (+33.33) nm.

2'-Hydroxyethyl-(2E,4E)-2-methyl-6-oxohepta-2,4-dienoate (compound 8): amorphous white powder; UV (MeOH) λ_max (log ε) 279 (3.07) nm; IR (KBr) ν_max 3361, 1647, 1260 cm^-1; ¹H NMR (CD₃OD, 400 MHz) δ_H 6.92 (1H, ddd, J=11.5,1.4,0.8 Hz, H-3), 7.56 (1H, dd, J=15.6, 11.5 Hz, H-4), 6.38 (1H, d, J=15.6 Hz, H-5), 2.33 (3H,s, H-7), 2.11 (3H, d, J=0.8 Hz, H-8), 3.39 (2H, t, J=6.0 Hz, H-1'), 3.64 (2H, t, J=6.0 Hz, H-2''); ¹³C NMR (CD₃OD, 100 MHz) δ_C 171.4 (C-1), 141.4 (C-2), 131.7 (C-3), 139.1 (C-4), 135.1 (C-5), 201.1 (C-6), 27.6 (C-7), 13.8 (C-8), 43.5 (C-1'), 61.5 (C-2'); HR-ESIMS m/z 100.0519 [M + 2H]²⁺ (calculated for C₁₀H₁₆O₄, 200.1038).

4.4. In Silico Simulation

4.4.1. Molecular Docking

Molecular-docking analysis began by obtaining the 3D structure of AMPK from the RCSB Protein Data Bank database (https://www.rcsb.org/, accessed on November 15, 2024) using the identifier 5ISO. The 3D structures of 1-21 were drawn using ChemSketch (ACD/Labs, Toronto, ON, Canada) and optimized with its integrated 3D optimization tool. AICAR, serving as a positive control, was obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov, accessed on November 15, 2024). All ligands were subsequently converted to appropriate file formats (pdb and mol2) using Open Babel version 3.1.1 (https://openbabel.org/index.html, accessed on November 15, 2024). Using UCSF Chimera (UCSF, CA, USA), both AMPK and the ligands underwent preparation processes, including hydrogen addition and charge assignment. The docking simulation employed four different programs: Autodock Vina 1.1.2 (Scripps Research, CA, USA), Autodock 4.2.6 (Scripps Research, CA, USA), and Dock6 (UCSF, CA, USA). Binding-site and grid-box dimensions were determined based on the co-crystallized ligand’s coordinates (x: −3.465, y: −56.795, z: 35.109), following each software’s specific guidelines. Following the docking simulations, pharmacophore analysis was conducted using LigandScout 4.0 (inte:Ligand, Vienna, Austria) to investigate potential interactions between the receptors and ligands.

4.4.2. Molecular Dynamics

Molecular dynamics (MD) simulations were performed using GROMACS 2024.3 (https://www.gromacs.org, accessed on October 24, 2024). The preparation of the AMPK–compound 1 complex involved separate preprocessing of protein and ligand structures, with ligand topology generated through the ATB server (https://atb.uq.edu.au). The ligand topology was then merged with the protein topology processed using the 54a7 force field. The complex was situated in a cubic box with a 1.0 nm buffer and solvated using simple point-charge water. After energy minimization using the steepest descent algorithm, the system underwent equilibration in two stages: 100 ps NVT at 300 K followed by 100 ps NPT at 1 bar. The production MD simulations ran for 10 ns using a 2 fs time step, incorporating periodic boundary conditions and the particle-mesh Ewald method for long-range electrostatics calculations. The resulting trajectories were analyzed using GROMACS tools to calculate RMSD, RMSF, and hydrogen bond interactions.

4.4.3. ADMET Prediction

Prediction of ADMET properties began by preparing the ligand structure of compound 1 in the SMILES format using ChemSketch, which was then uploaded to ADMETlab2.0 (https://admetmesh.scbdd.com, accessed on October 22, 2024) and SwissADME (http://www.swissadme.ch, accessed on October 22, 2024) web servers. The analysis encompassed various absorption properties, including water solubility, Caco-2 permeability, and gastrointestinal absorption. Distribution assessment focused on volume of distribution and blood-brain barrier penetration, while metabolism predictions examined cytochrome P450 enzyme inhibition profiles for five key enzymes: CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4. The excretion analysis used predicted clearance rates, and toxicity evaluations included assessments of hepatotoxicity, hERG inhibition, and mutagenicity through the AMES test. All predictions were conducted using default parameters provided by both platforms.

4.5. In Vitro Assay

4.5.1. Cell Treatment Experiments

HepG2 cells, a human hepatocellular carcinoma line, were sourced from the American Type Culture Collection (Manassas, VA, USA). The cells were grown in Dulbecco’s modified Eagle medium (Welgene, Gyeongsan, Korea), supplemented with 10% fetal bovine serum inactivated at 56°C for 30 min (Welgene), along with 100 U/mL penicillin and 100 μg/mL streptomycin (Welgene). Cultures were maintained in a humidified incubator at 37°C with 5% CO₂. The medium was replaced every two days to eliminate non-adherent cells and debris. To evaluate the effects of Rg 6 on activation of AMPK, HepG2 cells were treated with 2, 4, and 8 μM of Rg 6 and 500 μM of AICAR for 12 h.

4.5.2. Cell Lysis and Protein Extraction

Cells were washed with phosphate-buffered saline (PBS, Gibco, Grand Island, NY, USA) and collected using chilled PBS to lyse them and extract proteins. Protein extraction was performed with a cold RIPA buffer (Biosesang, Seongnam, Korea) supplemented with a protease inhibitor cocktail (GenDEPOT, Katy, TX, USA), following the manufacturer’s guidelines. Whole-cell lysates were prepared by heating 10–40 μg of total protein at 98°C for 5 min in a gel-loading buffer (0.3125 M Tris-HCl pH 6.8, 2% SDS, 5% 2-mercaptoethanol, 0.05% bromophenol blue, and 25% glycerol) at a 4:1 ratio. Protein quantification was achieved with Pierce BCA Protein Assay Kits (Thermo Scientific, Waltham, MA, USA). Sodium dodecyl sulfate–polyacrylamide gel electrophoresis was used to separate proteins, which were then transferred onto polyvinyl difluoride membranes (Millipore) using a Bio-Rad Western system. Membranes were blocked with 5% non-fat milk in TBS (50 mM Tris, 150 mM NaCl, pH 7.6) containing 0.1% Tween 20 (TBS-T) for 1 h, then washed five times for 10 min each in TBS-T. Membranes were incubated overnight at 4°C with primary antibodies diluted to 1:1,000. The following day, they were washed five times for 15 min each in TBS-T and then treated for 1 h at 25°C with secondary antibodies conjugated to horseradish peroxidase (HRP) (1:10,000). Finally, immunoblots were visualized using a chemiluminescent HRP substrate (Advansta, San Jose, CA, USA) and a ChemiDoc imaging system (Bio-Rad, Hercules, CA, USA), according to the manufacturer’s instructions.

4.5.3. Reagents

All antibodies were purchased from Santa Cruz Biotechnology (Dallas, TX, USA) and Cell Signaling Technology, Inc. (Beverly, MA, USA). The primary antibodies used were anti-GAPDH (sc-25778) and anti-AMPK (Thr172) (#2535). AICAR (A9978) and AMPK activator were obtained from Sigma-Aldrich (Milwaukee, WI, USA).

5. Conclusions

In this study, 21 compounds were isolated from the dried roots of R. glutinosa, and 3 of them (compounds 1, 2, and 8) were newly found structures in nature. These were uridine glycoside, O-bridged bicyclic skeleton, and hepta-dienoate skeletons, respectively. Compound 9 was found for the first time in this plant. As R. glutinosa has been reported to show protective effects against metabolic diseases, AMPK was selected as a therapeutic target. In silico docking simulations suggested that compound 1 can be a strong AMPK activator and this finding was supported by an in vitro assay. This study provided new structures that can be found in R. glutinosa, and we also found a potential ingredient (compound 1) that is active in AMPK activation. Future studies should investigate the mechanisms of action of compound 1 in the AMPK-signaling pathway and concentration-activity relationship of compound 1 with AMPK.