Molecular Networking-Guided Isolation of Cycloartane-type Triterpenoids from Curculigo orchioides and Their Inhibitory Effect on Nitric Oxide Production

Abstract

The MolNetEnhancer workflow was applied to molecular networking analysis of the CH₂Cl₂-soluble fraction of the rhizomes of Curculigo orchioides, which showed a potent inhibitory effect on the lipopolysaccharide (LPS)-induced nitric oxide production. Among the molecular network, clusters of cycloartane-type triterpenoids were classified using the ClassyFire module of MolNetEnhancer, and their structures were predicted by the in silico fragment analysis tool, Network Annotation Propagation (NAP). Using mass spectrometry (MS)-guided isolation methods, six cycloartane-type triterpenoids (1–6) were isolated, and their structures were elucidated based on the interpretation of NMR, HRESIMS, and single-crystal X-ray diffraction. Among the isolates, compounds 1 and 4, which have an α,β-unsaturated carbonyl moiety on the A-ring, exhibited significant inhibitory effects on LPS-induced nitric oxide production in RAW264.7 cells with IC₅₀ values of 12.4 and 11.8 μM, respectively.

Introduction

The rhizomes of Curculigo orchioides Gaertn. (Amaryllidaceae) have been used as a traditional folk medicine in Asia for the treatment of impotence, jaundice, asthma, decline in physical strength, and arthritis.¹⁻³ In the previous phytochemical studies of the genus Curculigo, phenolic compounds including chlorophenolic glycosides,⁴⁻⁶ lignans,^7,8 and triterpenoidal saponins⁹⁻¹² have been reported as the main bioactive constituents. Various pharmacological investigations have revealed the broad biological activities of the extracts and isolates. The ethanol extract of C. orichioides and isolated phenolic compounds such as curculigoside A, curculigoside B, curculigine A, and curculigine D exhibited antiosteoporotic activity.¹³ In addition, cycloartane glycosides exhibited cytotoxic activity against HL-60 human promyelocytic leukemia cells.¹² As part of an ongoing research program aimed at discovering new bioactive constituents in medicinal plants, the MeOH extract of the rhizomes of C. orchioides and the subsequent CH₂Cl₂-soluble fraction were found to possess inhibitory effect against the lipopolysaccharide (LPS)-induced nitric oxide (NO) production in RAW264.7 cells (IC₅₀ = 27.7 and 20.0 μg/mL, respectively).

Liquid chromatography high-resolution mass spectrometry (LC-HRMS/MS)-based molecular networking analysis by the Global Natural Products Social (GNPS) molecular networking web platform efficiently organized and visualized the cluster of structurally related compounds. Its rapidly growing social library makes it possible to predict the structure of an extract before intensive chromatographic isolation and structure determination.¹⁴ In addition, a spectral library could be propagated using in silico analysis tools such as the Network Annotation Propagation (NAP).¹⁵ Recently, Nasir et al. applied a combination of molecular networking and LC-MS/MS profiling to annotate phenolic glycosides and norlignans in the rhizomes and leaf extract of C. latifolia.¹⁶ However, the molecular network of the CH₂Cl₂-solube fraction of C. orchioides showed a different pattern, with only a few compounds matching the GNPS library.

In the present study, MolNetEnhancer, which integrates outputs from molecular networking, MS2LDA, in silico tools (NAP or DEREPLICATOR), and automated chemical classification through ClassyFire, was applied to the generated molecular networking analysis.¹⁷ The clusters of triterpenoids were classified by MolNetEnhancer, and the detailed structure was predicted to be cycloartane by NAP. Through targeted isolation of these clusters, six cycloartane-type triterpenoids (1–6), including five new compounds, were isolated. Herein, we report the molecular networking-guided isolation and structure determination of compounds 1–6, as well as the structure–activity relationship (SAR) of their inhibitory effects against LPS-induced nitric oxide production in RAW264.7 cells.

Results and Discussion

The LC-HRMS/MS data of the CH₂Cl₂-soluble fraction of C. orchioides were analyzed by classical molecular networking via the GNPS web platform (https://gnps.ucsd.edu). At first, only one node in cluster 1 matched with 1,2-dihydroxyheptadec-16-en-4-yl acetate in the GNPS spectrum library, but the spectrum similarity was not good, with a mass difference of 94.09 and cosine score of 0.75 (Figure 1A). Therefore, an in silico fragment analysis tool (NAP) combined with the structure library of GNPS and SUPER NATURAL 2 (SUPNAT) (Figure 1B) was used to conduct analysis.¹⁸ The results of classical molecular networking and NAP were integrated by MolNetEnhancer of the GNPS web platform, which automatically classified the chemical class of each cluster (Figure 1C). Among them, the largest cluster (cluster 1) revealed as that of triterpenoids and the result of NAP also predicted the structure of each node as various triterpenoids containing cycloartane types. Because many cycloartane-type saponins have been previously reported in C. orchioides,⁹⁻¹² it was reasonable to select cycloartane-type triterpenoids among the predicted candidates. In addition, only a few aglycones were reported,¹⁹ and many of which are produced through hydrolysis from isolated saponins. Therefore, cluster 1 was expected to contain new cycloartane-type triterpenoids. Twelve subfractions from silica gel chromatography were tested for their inhibitory effects on NO production in LPS-induced RAW264.7 cells, and the defined subfraction COC9, containing triterpenoids, showed the most potent inhibitory effects (Figure S2). Therefore, this fraction was investigated.

Figure 1.

Molecular networking analysis of the CH₂Cl₂-soluble fraction of C. orchioides. (A) Spectrum match of the node of molecular networking with GNPS library. (B) Structures of top ranked NAP candidates using GNPS and SUPNAT library. (C) Automatic classification and visualization of each cluster by the MolNetEnhancer. The chemical class of the largest (the cluster filled with red color) clusters were revealed as triterpenoids. The singleton node was excluded in this figure.

Compound 1 was isolated as a white amorphous powder. Its molecular formula was assigned as C₃₀H₄₈O₄ based on HRESIMS data (m/z 473.3624 [M + H]⁺; calcd 473.3625), indicating seven indices of hydrogen deficiency. The ¹H NMR data for 1 displayed two olefinic protons (δ_H 6.80, 5.96), three oxymethine protons (δ_H 4.57, 3.97, 3.39), four tertiary methyl protons (δ_H 1.12, 1.09, 1.03, 0.97), and three secondary methyl protons (δ_H 1.10, 0.93 × 2) (Table 1). The ¹³C NMR and HSQC spectra exhibited 30 carbon signals, including one carbonyl, two olefinic carbons, five quaternary carbons, seven methylenes, eight methines, and seven methyl groups (Table 2). The ¹H and ¹³C NMR data of 1 were similar to those of 12α,16β-dihydroxycycloartane-3,24-dione,¹⁹ except for the presence of a double bond and an additional oxymethine instead of a carbonyl group. The position of the α,β-unsaturated carbonyl group at the C-1/C-3 was determined by the HMBC correlations from H₂-19 (δ_H 1.25 and 0.95) to olefinic carbon (δ_C 153.9) and from CH₃-29 and CH₃-30 (δ_H 1.12 and 0.97) to a carbonyl carbon (δ_C 205.1). The proton signals of H₂-19 (δ_H 1.25 and 0.95) of 1 were shifted downfield compared with that of 3 (δ_H 0.65 and 0.59) because of the anisotropic effect caused by the alkene group between C-1 and C-2. Furthermore, through the HMBC correlations from CH₃-18 (δ_H 1.09) to C-12 (δ_C 73.1) and from H₂-15 (δ_H 2.06, 1.45) and H-20 (δ_H 1.89) to C-16 (δ_C 72.2), and from H-25 (δ_H 1.67), CH₃-26 (δ_H 0.93), and CH₃-27 (δ_H 0.93) to C-24 (δ_C 78.7), the planar structure of 1 was determined to be 12,16,24-trihydroxycycloartane-1-en-3-one (Figure 2). The relative configuration of 1 was deduced from the NOESY experiment. The NOE correlations of H₂-19/30-CH₃, H-6β, H-8, H-8/18-CH₃, and 18-CH₃/H-20 revealed the β-axial orientations of H-8, 18-CH₃, and H-20. In addition, the NOE correlations between H-16 and H-17 and 28-CH₃ revealed their α-orientations (Figure 2). Single-crystal X-ray diffraction analysis with Cu Kα radiation [Flack parameter = 0.00(5) and Hooft parameter = 0.02(4)] clearly corroborated the absolute configuration of 1 (Figure 3). Thus, the structure of 1 was determined to be (5R,8S,9R,10S,12S,13R,14S,16S,17R,20R,24S)-12,16,24-trihydroxycycloartane-1-en-3-one and named curculigone A.

Table 1. ¹H NMR (400^b, 500^c, and 800^d MHz) Data for Compounds 1–6^a.

no.	1b	2b	3c	4d	5b	6b
1	6.80 (d, 10.1)	6.80 (d, 10.1)	1.77 (m)	6.78 (d, 10.1)	2.36 (m)	1.58 (m)
			1.50 (m)		1.53 (m)	1.26 (m)
2	5.96 (d, 10.0)	5.95 (d, 10.0)	2.73 (m)	5.95 (d, 10.1)	2.76 (m)	1.78 (m)
			2.35 (m)		2.32 (m)	1.58 (m)
3						3.31 (m)
4
5	2.14 (m)	2.17 (m)	1.68 (dd, 4.3, 12.5)	2.07 (dd, 6.9, 10.2)	1.72 (m)	1.32 (m)
6	1.67 (m)	1.66 (m)	1.45 (m)	1.60 (m)	1.56 (m)	1.66 (m)
	1.00 (m)	1.02 (m)	0.81 (m)	1.08 (m)	0.98 (m)	0.80 (m)
7	1.47 (m)	1.47 (m)	1.38 (m)	1.56 (m)	1.40 (m)	1.34 (m)
	1.22 (m)	1.22 (m)	1.16 (m)	1.25 (m)	1.13 (m)	1.10 (m)
8	1.79 (m)	1.81 (dd, 5.1, 12.4)	1.65 (dd, 4.3, 13.0)	2.13 (dd, 3.9, 12.7)	1.69 (m)	1.49 (m)
9
10
11	2.12 (m)	2.11 (m)	2.40 (m)	1.96 (m)	2.09 (m)	1.96 (m)
	1.97 (m)	1.99 (m)	1.93 (m)	1.54 (m)	1.19 (m)	1.76 (m)
12	3.97 (dd, 5.6, 9.1)	3.94 (dd, 5.6, 9.1)	4.21 (m)	1.65 (m)	1.83 (m)	3.87 (dd, 6.0, 9.6)
					1.60 (m)
13
14
15	2.06 (m)	2.03 (m)	2.27 (dd, 8.1, 12.9)	2.01 (dd, 8.1, 13.1)	2.02 (m)	2.06 (dd, 8.1, 13.2)
	1.45 (m)	1.50 (m)	1.86 (m)	1.46 (dd, 4.7, 13.1)	1.47 (m)	1.41 (dd, 4.2, 13.2)
16	4.57 (ddd, 2.8, 8.0, 11.6)	4.64 (ddd, 2.8, 7.7, 12.0)	4.91 (ddd, 2.8, 7.8, 12.9)	4.47 (ddd, 2.5, 7.5, 12.5)	4.48 (ddd, 2.9, 7.8, 12.0)	4.56 (ddd, 2.8, 7.5, 12.6)
17	2.20 (dd, 8.0, 11.6)	2.13 (m)	2.91 (dd, 7.8, 11.0)	1.89 (dd, 7.5, 11.0)	1.89 (m)	2.18 (dd, 7.5, 11.2)
18	1.09 (s)	1.08 (s)	1.48 (s)	1.19 (s)	1.23 (s)	1.08 (s)
19	1.25 (m)	1.25 (d, 4.6)	0.65 (d, 4.0)	1.32 (d, 4.6)	0.83 (d, 4.2)	0.55 (d, 4.3)
	0.95 (m)	0.94 (d, 4.6)	0.59 (d, 4.3)	0.79 (d, 4.6)	0.60 (d, 4.3)	0.44 (d, 4.4)
20	1.89 (m)	1.24 (m)	2.55 (m)	2.33 (m)	2.29 (m)	1.87 (m)
21	1.10 (d, 6.5)	1.06 (d, 6.8)	1.52 (d, 6.7)	0.95 (d, 7.0)	0.95 (d, 7.1)	1.08 (d, 7.1)
22	1.75 (m)	2.07 (m)	2.42 (m)	4.06 (dt, 2.2, 10.1)	4.06 (dt, 2.2, 9.8)	2.36 (m)
	1.19 (m)	1.08 (m)	1.89 (m)			1.73 (m)
23	1.72 (m)	2.77 (m)	1.95 (m)	1.75 (m)	1.76 (m)	1.86 (m)
	1.30 (m)	2.56 (m)	1.81 (m)	1.68 (m)	1.67 (m)	1.72 (m)
24	3.39 (m)		3.68 (m)	3.63 (m)	3.64 (m)	3.39 (m)
25	1.67 (m)	2.64 (sept, 6.9)	1.58 (m)	1.84 (m)	1.81 (m)	1.65 (m)
26	0.93 (d, 6.8)	1.13 (d, 6.8)	1.12 (d, 6.8)	1.01 (d, 6.6)	1.01 (d, 6.6)	0.93 (d, 6.8)
27	0.93 (d, 6.8)	1.12 (d, 6.8)	1.10 (d, 6.8)	0.90 (d, 6.8)	0.91 (d, 6.8)	0.91 (d, 6.8)
28	1.03 (s)	1.04 (s)	1.34 (s)	0.92 (s)	0.92 (s)	0.99 (s)
29	1.12 (s)	1.12 (s)	1.17 (s)	1.11 (s)	1.11 (s)	0.98 (s)
30	0.97 (s)	0.97 (s)	1.07 (s)	0.97 (s)	1.05 (s)	0.81 (s)

^aAssignments were based on COSY and HSQC experiments. Compound 3 was dissolved with pyridine-d₅, and others used CDCl₃.

Table 2. ¹³C NMR (100^b, 125^c, and 200^d MHz) Data for Compounds 1–6^a.

no.	1b	2b	3c	4d	5b	6b
1	153.9, CH	154.0, CH	33.6, CH2	153.9, CH	33.4, CH2	32.2, CH2
2	126.9, CH	126.9, CH	37.6, CH2	126.7, CH	37.5, CH2	30.4, CH2
3	205.1, C	205.3, C	215.0, C	205.3, C	216.6, C	78.7, CH
4	46.3, C	46.3, C	50.3, C	46.0, C	50.3, C	40.5, C
5	45.7, CH	45.7, CH	48.7, CH	44.3, CH	48.5, CH	47.3, CH
6	20.5, CH2	20.5, CH2	21.8, CH2	19.7, CH2	21.5, CH2	21.2, CH2
7	25.0, CH2	25.0, CH2	26.6, CH2	23.8, CH2	26.1, CH2	26.3, CH2
8	47.3, CH	47.3, CH	49.1, CH	44.7, CH	48.0, CH	48.6, CH
9	23.1, C	23.2, C	21.2, C	24.4, C	21.0, C	19.6, C
10	29.2, C	29.0, C	26.0, C	29.9, C	26.0, C	26.0, C
11	39.2, CH2	39.6, CH2	40.3, CH2	27.6, CH2	26.5, CH2	38.3, CH2
12	73.1, CH	73.2, CH	72.4, CH	32.6, CH2	33.0, CH2	73.6, CH
13	49.6, C	49.6, C	50.1, C	45.8, C	45.8, C	49.5, C
14	46.8, C	46.7, C	47.1, C	47.3, C	46.9, C	46.8, C
15	48.4, CH2	45.7, CH2	50.8, CH2	46.0, CH2	47.1, CH2	49.0, CH2
16	72.2, CH	71.7, CH	71.7, CH	71.7, CH	72.1, CH	72.6, CH
17	49.2, CH	49.6, CH	49.3, CH	52.0, CH	52.3, CH	49.1, CH
18	17.9, CH3	17.7, CH3	18.6, CH3	17.8, CH3	18.9, CH3	17.7, CH3
19	32.2, CH2	32.2, CH2	29.6, CH2	30.1, CH2	29.8, CH2	30.0, CH2
20	31.2, CH	29.4, CH	31.6, CH	34.9, CH	35.0, CH	31.3, CH
21	17.7, CH3	17.0, CH3	17.8, CH3	16.1, CH3	15.9, CH3	17.8, CH3
22	33.5, CH2	29.3, CH2	34.1, CH2	74.2, CH	74.2, CH	33.5, CH2
23	30.9, CH2	36.7, CH2	32.3, CH2	34.3, CH	34.5, CH2	30.9, CH2
24	78.7, CH	217.1, C	77.4, CH	76.2, CH	76.1, CH	78.7, CH
25	34.0, CH	41.0, CH	34.2, CH	32.5, CH	32.6, CH	33.9, CH
26	18.7, CH3	18.4, CH3	19.8, CH3	18.8, CH3	18.7, CH3	18.8, CH3
27	17.4, CH3	18.3, CH3	17.7, CH3	18.7, CH3	19.0, CH3	17.4, CH3
28	21.1, CH3	21.3, CH3	22.1, CH3	19.6, CH3	20.3, CH3	21.4, CH3
29	21.6, CH3	21.6, CH3	22.7, CH3	21.4, CH3	22.2, CH3	25.4, CH3
30	19.1, CH3	19.1, CH3	20.8, CH3	19.1, CH3	20.8, CH3	14.0, CH3

^aAssignments were based on COSY and HSQC experiments. Compound 3 was dissolved with pyridine-d₅, and others used CDCl₃.

Figure 2.

Key HMBC, COSY, and NOESY correlations of compound 1.

Figure 3.

X-ray ORTEP plot for the molecular structure of compound 1.

Compound 2 was obtained as a white amorphous powder. The molecular formula of 2 was deduced from the HRESIMS data as C₃₀H₄₆O₄ (m/z 471.3467 [M + H]⁺; calcd 471.3469), indicating eight indices of hydrogen deficiency. A comparison of the ¹H and ¹³C NMR data of 1 and 2 indicated that 2 was similar to 1 except the hydroxyl group was replaced by a carbonyl group (δ_C 217.1) in 2 (Tables 1 and 2). The positions of the two carbonyl groups at C-3 and C-24 were determined by the HMBC correlations of H-1 (δ_H 6.80), CH₃-29 (δ_H 1.12), and CH₃-30 (δ_H 0.97)/C-3 (δ_C 205.3) and H-25 (δ_H 2.64), CH₃-26 (δ_H 1.13), and CH₃-27 (δ_H 1.12)/C-24 (δ_C 217.1) (Figure 4). The relative configuration of 2 was determined to be the same as that of 1 based on their similar NOE correlations (Figure 4). Furthermore, single-crystal X-ray diffraction analysis using Cu Kα radiation was performed on 2, and the Flack parameter −0.14(8) and Hooft parameter −0.16(7) permitted the absolute configuration (Figure 5). Thus, the structure of 2 was determined to be (5R,8S,9R,10S,12S,13R,14S,16S,17R,20R)-12,16-dihydroxycycloartane-1-en-3,24-one and named curculigone B.

Figure 4.

Key HMBC, COSY, and NOESY correlations of compound 2.

Figure 5.

X-ray ORTEP plot for the molecular structure of compound 2.

Compound 3 was obtained as a white amorphous powder. Its molecular formula was determined by the HRESIMS data as C₃₀H₅₀O₄ (m/z 497.3598 [M + Na]⁺; calcd 497.3601), indicating six indices of hydrogen deficiency, one less than that of 1. In the ¹H NMR and ¹³C NMR data, the signals corresponding to the double bond disappeared, indicating the reduction of the double bond between C-1 and C-2 in 1 (Tables 1 and 2). The positions of the carbonyl group and three hydroxy groups were determined by the HMBC correlations of CH₃-29, CH₃-30/C-3, H₂-11, CH₃-18/C-12, H₂-15, H-20/C-16, and H-25, CH₃-26, CH₃-27/C-24 (Figure 6). As with compounds 1 and 2, the relative and absolute configurations of 3 were determined by NOESY experiment and single-crystal X-ray diffraction analysis [Flack parameter = 0.08(5) and Hooft parameter = 0.07(4)] (Figure 7). Thus, the structure of 3 was determined as (5R,8S,9R,10R,12S,13R,14S,16S,17R,20R,24S)-12,16,24-trihydroxycycloartane-3-one and named curculigone C.

Figure 6.

Key HMBC, COSY, and NOESY correlations of compound 3.

Figure 7.

X-ray ORTEP plot for the molecular structure of compound 3.

Compound 4 was obtained as a white amorphous powder and exhibited an m/z value of 473.3623, corresponding to the molecular formula of C₃₀H₄₈O₄ ([M + H]⁺, calcd 473.3625), with seven indices of hydrogen deficiency. Its molecular weight was the same as that of 1, but the chemical shifts of the oxymethines showed a slight difference in ¹H and ¹³C NMR data (Tables 1 and 2). The positions of the three hydroxy groups were determined to be C-16, C-22, and C-24 by the HMBC correlations of H₂-15 (δ_H 2.01 and 1.46), H-20 (δ_H 2.33)/C-16 (δ_C 71.7), H-20 (δ_H 2.33), CH₃-21 (δ_H 0.95)/C-22 (δ_C 74.2), and H-25 (δ_H 1.84), CH₃-26 (δ_H 1.01), CH₃-27 (δ_H 0.90)/C-24 (δ_C 76.2). In addition, the ¹H–¹H COSY correlations of H-22/H₂-23/H-24 verified the positions of two hydroxy groups at C-22 and C-24. The relative configuration of 4 was elucidated by the ROESY experiment with the aid of computational 3D modeling after MM2 energy minimization (Figure 8). As with compounds 1–3, the relative configurations of H-8β, CH₃-18β, CH₃-30β, H-5α, and CH₃-28α were deduced from their ROESY correlations. In addition, the large coupling constant between H-17 and H-20 (J = 11.0 Hz), together with the ROESY correlation between CH₃-18β and H-20, clearly determined the β-configuration of H-20. The coupling constant between H-20 and H-22 exhibited a large value (J = 10.1 Hz), and two possible configurations were simulated using computational 3D modeling for the preferred match. In the case of the 3D model of 20S*,22R*, the dihedral angle between H-20 and H-22 was −163.3°, whereas that of the model of 20S*,22S* was 69.4°. Thus, the configurations of H-20 and H-22 were determined to be 20S*,22R* according to the Karplus equation (Figure 8A).²⁰ Furthermore, in the computational model for the 22R*,24R* configuration, the distance between H-22 and H-24 was 3.83 Å, which is too far to show the NOE effect. For 22R*,24S*, the distance of H-22 and H-24 (2.66 Å) corresponded with the observed ROESY correlation between H-22 (δ_H 4.06) and H-24 (δ_H 3.63) (Figure 8B).²¹ The optical rotation of compound 4 was measured as +3.5 (c 0.5, methanol), which has the same polarity as compounds 1–3. Therefore, the structure of 4 was determined as (5R,8S,9S,10S,13R,14S,16S,17R,20S,22R,24S)-16,22,24-trihydroxycycloartane-1-en-3-one and named curculigone D.

Figure 8.

3D model simulation after MM2 minimization (minimum RMS gradient = 0.01) for the comparison of relative configuration of compound 4. (A) Comparison of 3D computational models of (20S*,22R*)-4 and (20S*,22S*)-4 and ³J values upon dihedral angle. (B) Comparison of (20S*,22R*,24S*)-4 and (20S*,22R*,24R*)-4 and calculated interproton distances.

Compound 5 was obtained as a white amorphous powder and assigned the molecular formula C₃₀H₅₀O₄ based on its HRESIMS data (m/z 475.3777 [M + H]⁺, calcd 475.3782) and ¹³C NMR data. With one less hydrogen deficiency than 4, along with the disappearance of the olefinic moiety in ¹H and ¹³C NMR data revealed the reduction of the double bond between C-1 and C-2 of 4 (Tables 1 and 2). The planar structure of 5 was determined to be 16,22,24-trihydroxycycloartane-3-one by HMBC experiment. The large coupling constant between H-20 and H-22 (J = 9.8 Hz) matched well with the 3D modeling for 20S*,22R*, which showed a dihedral angle of −163.9°. Moreover, the observed NOESY correlation of H-22 and H-24 permitted the configuration of 22R*,24S* (2.66 Å) rather than 22R*,24R* (3.83 Å) (Figure 9). The positive optical rotation value of +5.9 (c 0.5, methanol) confirmed the absolute configuration of 5. Therefore, the structure of 5 was determined as (5R,8S,9S,10R,13R,14S,16S,17R,20S,22R,24S)-16,22,24-trihydroxycycloartane-3-one and named curculigone D.

Figure 9.

3D computational model for the compound 5 and key NOESY correlation.

Compound 6 was obtained as a white amorphous powder. Its molecular formula was determined by the HRESIMS data as C₃₀H₅₂O₄ ([m/z 499.3757 [M + Na]⁺; calcd 499.3758). The ¹H and ¹³C NMR data of 6 resembled those of 3, except for the absence of the carbonyl group and the presence of an additional hydroxyl group (Tables 1 and 2). These data were consistent with those of (24S)-9,19-cyclolanostane-3β,12α,16β,24-tetrol, which was produced by the enzymatic hydrolysis of (24S)-12α,16β,24-trihydroxy-9,19-cyclolanostan-3β-hydroxyl-β-d-glucopyranoside.¹²

After structure determination, the clusters classified as triterpenoids by MolNetEnhancer were annotated according to their structures (Figure 10). The structures predicted by the NAP module with GNPS and the SUPNAT library were close with the actual structure, except for the position of double bond, and/or the number of hydroxyl group. The isolated compounds were mostly present as adducts of [M + H – H₂O]⁺ or [M + H – 2H₂O]⁺, whereas most of the predicted candidates were present as protonated or a sodium adduct in the spectrum library. This gap in the adducts may result in a difference between predicted and actual structures.

Figure 10.

Annotation of compounds 1–6 on the triterpenoid clusters.

All compounds were tested for their inhibitory effects on LPS-induced NO production in RAW264.7 macrophages using aminoguanidine as a positive control (Table 3). The cytotoxicity of the compounds was examined using the MTT reagent, and none of the test compounds showed significant cytotoxicity at their effective concentrations for the inhibition of NO production (Figure S41, Supporting Information). A comparison of the IC₅₀ values of compounds 1–6 suggested the α,β-unsaturated carbonyl moiety on the A-ring was critical for NO inhibition. The lack of an inhibitory effect on NO production in compound 2, despite having an α,β-unsaturated carbonyl moiety in the A-ring, could be due to the introduction of an additional carbonyl group at C-24.

Table 3. Inhibitory Effects of Compounds 1–6 against LPS-Induced NO Production in RAW264.7 Cells and Their Cytotoxicity^a.

compound	IC50 (μM)b	CC50 (μM)c
1	12.4 ± 0.9	76.7 ± 1.0
2	>100	>100
3	>100	>100
4	11.8 ± 1.2	66.9 ± 3.1
5	29.4 ± 0.2	89.1 ± 2.6
6	41.7 ± 0.7	54.8 ± 4.6

^aAminoguanidine was used as the positive control (IC₅₀ = 15.9 ± 1.7 μM). Results are expressed as the mean IC₅₀ values in μM from triplicate experiments.

^bIC₅₀: 50% inhibitory concentration.

^cCC₅₀: 50% cytotoxic concentration.

In conclusion, the MolNetEnhacer module, coupled with the in silico fragment analysis tool NAP, was applied to the molecular networking analysis of the rhizomes of C. orchioides. The workflow of MolNetEnhancer efficiently classified the chemical class of each cluster, and the largest cluster in the molecular network was revealed to be triterpenoid. Targeted isolation of this cluster afforded six cycloartane-type triterpenoids (1–6). Among them, the cylcloartane-1-en-3-one-type compound (1 and 4) exhibited potent inhibitory effects against LPS-induced NO production in RAW264.7 cells. Hence, cycloartane-type triterpenoids from the rhizomes of C. orchioides may have potential for the treatment of inflammation-associated diseases and are worthy of further study for their mechanism of action.

Experimental Section

General Experimental Procedures

The optical rotations were measured using a JASCO DIP-1000 polarimeter. UV and IR spectra were obtained using a JASCO UV-550 spectrophotometer and JASCO FT-IR 4100 spectrometer, respectively. NMR spectra were recorded on Bruker AVANCE 400, 500, and 800 MHz spectrometers using CDCl₃ and pyridine-d₅ as solvents. HR-ESI-MS and UPLC-MS/MS analyses were performed using an Orbitrap Exploris 120 mass spectrometer coupled with a Vanquish UHPLC system (ThermoFisher Scientific). Chromatographic elution was conducted on a YMC Triart C₁₈ (100 × 2.1 mm, 1.9 μm, 0.3 mL/min) column at a temperature of 30 °C, with a mobile phase of water +0.1% formic acid (A) and CH₃CN + 0.1% formic acid (B), and the gradient consisted of a linear gradient of 10–100% B (0–10 min). Mass detection was performed in the m/z range of 200–2000, and the resolution of the Orbitrap mass analyzer was fixed at 60,000 for the full MS scan and 15,000 for the data-dependent MSⁿ scan. The following parameters were used during MS measurements: a spray voltage of 3.5/2.5 kV for the positive/negative modes, ion transfer tube temperature of 320 °C, HESI probe vaporizer temperature of 275 °C, RF lens 70 (%). Ultrapure nitrogen (>99.999%) was used as both sheath and auxiliary gas of the HESI probe and set to 50 and 15 arb, respectively. A normalized higher-energy collision dissociation energy of 30% was used for the collision of ions in the Orbitrap detector. MS/MS fragmentation was obtained by the data-dependent MSⁿ mode to obtain an MS² spectrum of the four most intense ions, with a dynamic exclusion filter to exclude the repeated fragmentation of ions within 2.5 s after acquiring the MS² spectrum. Column chromatography was performed using a silica gel column (Merck, 70-230 mesh). MPLC was performed using a Biotage Isolera Prime chromatography system with Lichroprep RP-18 (Merck, 40-63 μm). Preparative HPLC was performed using a Waters HPLC system equipped with two Waters 515 pumps, a 2996 photodiode array detector, and a YMC J’sphere ODS H-80 column (4 μm, 150 × 20 mm, i.d., flow rate 6.0 mL/min). Thin layer chromatography was performed using precoated silica gel 60 F₂₅₄ (0.25 mm, Merck) plates, and spots were detected using a 10% vanillin-H₂SO₄ aqueous spray reagent.

Plant Material

Dried rhizomes of C. orchioides were purchased from the Kyungdong herbal market (Seoul, South Korea) in May 2021 and were identified by one of the authors (B.Y.H.). A voucher specimen (CBNU-2021-05-CO) was deposited at the Herbarium of the College of Pharmacy, Chungbuk National University, South Korea.

Molecular Networking-Based Analysis

LC-MS/MS data were uploaded to the GNPS (http://gnps.ucsd.edu) for classical molecular networking. The product ion and fragment ion mass tolerances were set to 0.02 Da. The edge between the nodes was created when they displayed cosine scores above 0.7 and more than six matched peaks. Furthermore, these edges were kept in the network if and only if each of the nodes appeared in each of the top 10 most similar nodes. The network was downloaded and visualized using Cytoscape 3.8.2,²² and the NAP tool in the GNPS web platform was used for in silico analysis of the molecular network. The following parameters were used for this purpose: 10 first candidates, mass tolerance of 5 ppm, database of GNPS and SUPNAT2 (Supernatural II). After analysis, the resulting file was exported to Cytoscape, and the structures were visualized using the ChemViz2 plug-in. The workflow and results of classical molecular networking and NAP could be browsed on the following GNPS Web site links:

https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=e68dc185eac44e4889272286f5e971b8

https://proteomics2.ucsd.edu/ProteoSAFe/status.jsp?task=d5b5890d96df4b4b85df42566c7e5916

The MolNetEnhancer workflow was conducted via “Advanced Views-Experimental Views” tab of the result of classical networking with the job ID of NAP above. The result (https://proteomics2.ucsd.edu/ProteoSAFe/status.jsp?task=d5b5890d96df4b4b85df42566c7e5916) was downloaded and imported to Cytoscape, and the color of the clusters was automatically filled upon the chemical subclass.

Extraction and Isolation

The dried rhizomes of C. orchioides (3.0 kg) were extracted with MeOH (3 × 18 L) by maceration for 3 days at room temperature (25 °C) and then filtered and evaporated under reduced pressure to obtain a MeOH extract. A suspension of the extract (150 g) in distilled water was partitioned sequentially with n-hexane (2 × 2 L), CH₂Cl₂ (2 × 2 L), EtOAc (2 × 2 L), and n-BuOH (2 × 2L). The CH₂Cl₂-soluble fraction (8.3 g) was subjected to silica gel column chromatography and eluted using a CH₂Cl₂–MeOH gradient system (30:1 to 0:1) to yield 12 fractions (COC1–COC12). COC9 (2.5 g) was further separated by RP-MPLC and eluted with a MeOH–H₂O gradient system (20:80 to 100:0) to obtain 18 subfractions (COC9-1–COC9-18). COC9-12 (134.2 mg) was further purified by preparative HPLC (MeCN–H₂O, 57:43, isocratic) to yield (24S)-9,19-cyclolanostane-3β,12α,16β,24-tetrol (6) (t_R = 22.5 min, 2.1 mg), compound 1 (t_R = 27.1 min, 4.6 mg), and compound 2 (t_R = 49.0 min, 12.2 mg). COC9-13 (116.8 mg) was isolated by preparative HPLC (MeCN–H₂O, 55:45, isocratic) to obtain compounds 4 (t_R = 41.5 min, 1.6 mg) and 3 (t_R = 43.0 min, 7.2 mg). COC9-14 (80.4 mg) was purified using preparative HPLC (MeCN–H₂O, 65:35, isocratic) to yield compound 5 (t_R = 32.8 min, 2.5 mg).

Curculigone A (1):

White amorphous powder; [α]_D²⁵ +5.5 (c 1.0, MeOH); UV (MeOH) λ_max 270 nm; IR ν_max (film) 3390, 2956, 2925, 2854, 1705, 1384, and 1102 cm^–1; ¹H NMR (400 MHz, CDCl₃) and ¹³C NMR (100 MHz, CDCl₃), see Tables 1 and 2; HRESIMS m/z 473.3624 [M + H]⁺ (calcd for C₃₀H₄₉O₄, 473.3625).

Curculigone B (2):

White amorphous powder; [α]_D²⁵ +3.2 (c 0.5, MeOH); UV (MeOH) λ_max 269.2 nm; IR ν_max (film) 3398, 2950, 2915, 2856, 1711, 1705, 1384, and 1102 cm^–1; ¹H NMR (400 MHz, CDCl₃) and ¹³C NMR (100 MHz, CDCl₃), see Tables 1 and 2; HRESIMS m/z 471.3467 [M + H]⁺ (calcd for C₃₀H₄₇O₄, 471.3469).

Curculigone C (3):

White amorphous powder; [α]_D²⁵ +14.3 (c 1.0, MeOH); UV (MeOH) λ_max 210 nm; IR ν_max (film) 3405, 2951, 2925, 2848, 1710, and 1036 cm^–1; ¹H NMR (500 MHz, pyridine-d₅) and ¹³C NMR (125 MHz, pyridine-d₅), see Tables 1 and 2; HRESIMS m/z 497.3598 [M + Na]⁺ (calcd for C₃₀H₅₀NaO₄, 497.3601).

Curculigone D (4):

White amorphous powder; [α]_D²⁵ +3.5 (c 0.5, MeOH); UV (MeOH) λ_max 270 nm; IR ν_max (film) 3393, 2947, 2922, 2858, 1704, and 1106 cm^–1; ¹H NMR (800 MHz, CDCl₃) and ¹³C NMR (200 MHz, CDCl₃), see Tables 1 and 2; HRESIMS m/z 473.3623 [M + H]⁺ (calcd for C₃₀H₄₉O₄, 473.3625).

Curculigone E (5):

White amorphous powder; [α]_D²⁵ +5.9 (c 0.5, MeOH); UV (MeOH) λ_max 210 nm; IR ν_max (film) 3383, 2957, 2930, 2850, 1709, and 1098 cm^–1; ¹H NMR (400 MHz, CDCl₃) and ¹³C NMR (100 MHz, CDCl₃), see Tables 1 and 2; HRESIMS m/z 475.3777 [M + H]⁺ (calcd for C₃₀H₅₁O₄, 475.3782).

(24S)-9,19-Cyclolanostane-3β,12α,16β,24-tetrol (6):

White amorphous powder; [α]_D²⁵ +26.0 (c 1.0, MeOH); UV (MeOH) λ_max 224 nm; IR ν_max (film) 3365, 2957, 2925, 2848, and 1036 cm^–1; ¹H NMR (400 MHz, CDCl₃) and ¹³C NMR (100 MHz, CDCl₃), see Tables 1 and 2; HRESIMS m/z 499.3757 [M + Na]⁺ (calcd for C₃₀H₅₂NaO₄, 499.3758).

X-ray Crystallographic Analysis of Compounds 1–3

Single crystals of compounds 1–3 were recrystallized by an evaporation method with MeOH and analyzed using a Bruker D8 Venture diffractometer equipped with a monochromatic fine-focus Cu Kα (λ = 1.54178 Å) radiation source. Data collection was carried out using a PHOTON 100 CMOS detector at 223(2) K using the APEX2 software (Bruker AXS Inc.). The crystal structure was refined by full-matrix least-squares refinement using the SHELXL-2014 computer program.²³ Further analysis of the properties of the single crystals was performed using PLATON.²⁴ Molecular graphics were computed using the Mercury 4.2 software. Crystallographic data for 1–3 were deposited in the Cambridge Crystallographic Data Centre (deposition numbers CCDC 2173493, CCDC 2173494, and CCDC 2173491). These data can be obtained free of charge at www.ccdc.cam.ac.uk.

Crystal Data of Curculigone A (1):

Monoclinic crystal, C₃₀H₅₀O₅, M_r = 490.70, size 0.155 × 0.070 × 0.055 mm³, a = 13.4533(4) Å, b = 7.5536(2) Å, c = 13.6495(3) Å, α = 90.00°, β = 92.5160(10)°, γ = 90.00°, V = 1385.74(7) ų, T = 223(2) K, space group P2₁, Z = 2, μ = 0.613 mm^–1; 29287 collected reflections, 5803 independent reflections (R_int = 0.0275), R₁ (all data) = 0.0445, wR₂ (all data) = 0.1252. The absolute structure was determined by the Flack parameter x = 0.00(5) and Bijvoet pair analysis by Hooft method (Hooft y = 0.02(4), P2 (true) = 1.000, P3 (true) = 1.000, P3 (racemic twin) = 0.000).

Crystal Data of Curculigone B (2):

Orthorhombic crystal, C₃₀H₄₆O₄, M_r = 470.67, size 0.564 × 0.053 × 0.030 mm³, a = 7.9126(2) Å, b = 31.3030(7) Å, c = 31.6212(8) Å, α = 90.00°, β = 90.00°, γ = 90.00°, V = 7832.3(3) ų, T = 223(2) K, space group P2₁2₁2_1,Z = 12, μ = 0.604 mm^–1; 91532 collected reflections, 16472 independent reflections (R_int = 0.0714), R₁ (all data) = 0.0738, wR₂ (all data) = 0.1569. The absolute structure was determined by the Flack parameter x = −0.14 (8) and Bijvoet pair analysis by Hooft method (Hooft y = −0.16(7), P2 (true) = 1.000, P3 (true) = 1.000, P3 (racemic twin) = 0.000).

Crystal Data of Curculigone C (3):

Monoclinic crystal, C₃₁H₅₄O₅, M_r = 506.74, size 0.600 × 0.097 × 0.088 mm³, a = 13.4720(15) Å, b = 7.5719(6) Å, c = 13.7835(9) Å, α = 90.00°, β = 91.428 (6)°, γ = 90.00°, V = 1405.6(2) ų, T = 223(2) K, space group P2_1,Z = 2, μ = 0.618 mm^–1; 25265 collected reflections, 5828 independent reflections (R_int = 0.0281), R₁ (all data) = 0.0432, wR₂ (all data) = 0.1242. The absolute structure was determined by the Flack parameter x = 0.08 (5) and Bijvoet pair analysis by Hooft method (Hooft y = 0.07(4), P2 (true) = 1.000, P3 (true) = 1.000, P3 (racemic twin) = 0.000).

Measurement of LPS-Induced Nitric Oxide Production and Cell Viability

As an indicator of NO synthesis, the nitrite concentration in the supernatant of RAW264.7 cells was measured according to the Griess reaction following a previously reported protocol. Cell viability of the remaining cells was determined by a MTT (Sigma Chemical Co., St. Louis, MO)-based colorimetric assay.²⁵

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.2c03243.

• 1D NMR, 2D NMR, and HRESIMS of compounds 1–6 (PDF)

• X-ray crystallography data of 1 (CIF)

• X-ray crystallography data of 2 (CIF)

• X-ray crystallography data of 3 (CIF)

The authors declare no competing financial interest.