Forrestiacids A and B, Pentaterpene Inhibitors of ACL and Lipogenesis: Extending the Limits of Computational NMR Methods in the Structure Assignment of Complex Natural Products

Abstract

Forrestiacids A (1) and B (2) are a novel class of [4+2] type pentaterpenoids derived from a rearranged lanostane moiety (dienophile) and an abietane unit (diene). These unprecedented molecules were isolated using guidance by molecular ion networking (MoIN) from Pseudotsuga forrestii, an endangered member of the Asian Douglas Fir Family. The intermolecular hetero-Diels–Alder adducts feature an unusual bicyclo[2.2.2]octene ring system. Their structures were elucidated by spectroscopic analysis, GIAO NMR calculations and DP4 + probability analyses, electronic circular dichroism calculations, and X-ray diffraction analysis. This unique addition to the pentaterpene family represents the largest and the most complex molecule successfully assigned using computational approaches to predict accurately chemical shift values. Compounds 1 and 2 exhibited potent inhibitory activities (IC₅₀s < 5 μM) of ATP-citrate lyase (ACL), a new drug target for the treatment of glycolipid metabolic disorders including hyperlipidemia. Validating this activity 1 effectively attenuated the de novo lipogenesis in HepG2 cells. These findings provide a new chemical class for developing potential therapeutic agents for ACL-related diseases with strong links to traditional medicines.

The elevated level of low-density lipoprotein cholesterol (LDL-C) represents a key risk factor for atherosclerotic cardiovascular disease (CVD), which is associated with high morbidity and mortality as well as increasing health-care costs.^[1] Statins are effective at decreasing LDL-C and CVD, but are associated with some muscle- and skeletal-related side-effects.^[2] The adenosine triphosphate (ATP)-citrate lyase (ACL) is an important enzyme linking carbohydrate to lipid metabolism by converting citrate to acetyl-CoA, which serves as a common substrate for fatty acid and cholesterol biosynthesis.^[3] ACL has been recently identified as a novel drug target for hyperlipidemia, hypercholesterolemia and other metabolic disorders.^[4,5] ACL inhibitors may suppress the synthesis of fatty acids and cholesterol, and significantly lower the elevated level of LDL-C. An excellent example is the recently approved dicarboxylic acid derivative, bempedoic acid, which is now prescribed for use in statin resistant hypercholesterolemia.^[5]

There is still considerable effort required to identify improved ACL inhibitors as bempedoic acid has some treatment-limiting side-effects. The plant family Pinaceae is ranked among the top-20 privileged drug-prolific families that produce a high number of approved drugs.^[6] Natural products (NPs) from Pinaceae species have been used extensively in both traditional Asian and Native American medicines involving metabolic diseases making them ideal candidates for the discovery of ACL inhibitors.^[6,7] Continuing our research towards the identification of new bioactive NPs from endangered conifers,^[8] Pseudotsuga forrestii Craib, a relict and vulnerable Pinaceae species,^[7,9] was phytochemically investigated.^[10] In this study, by the implementation of HRMS/MS-based molecular ion networking (MoIN),^[11] a pair of unique [4+2]-adducts [i.e., forrestiacids A (1) and B (2), Figure 1] featuring an unprecedented pentaterpene carbon skeleton formed by intermolecular Diels–Alder cycloaddition between a rearranged 6/6/5/5-fused lanostane unit and an abietadiene unit, were obtained from the twigs/needles of P. forrestii. Herein, their isolation, structural characterization, biogenetic consideration, and inhibitory effects against ACL and lipogenesis are presented.

Figure 1.

A new class of pentaterpenes from Pseudotsuga forrestii (The bonds in red showing the connective sites between the tri- and di-terpenoid units).

Isolation of the novel chemical class of [4+2]-adducts was facilitated under the guidance of HPLC-HRMS/MS (positive mode)-based MoIN as enabled through the Cytoscape platform (https://cytoscape.org/).^[11] This analysis revealed two unique signals at m/z 769 with different retention times (Figure 2). Subsequently, target purification of these m/z 769 metabolites afforded the terpenoid [4+2] hetero-dimers 1 and 2.

Figure 2.

MoIN of the mentioned MeOH extract of P. forrestii. left: Organized landscape of complete network; right: the MoIN cluster with signals at m/z 769.

Forrestiacid A (1), colorless crystals from MeOH, possesses the molecular formula C₅₀H₇₂O₆ as assigned by the HRESIMS ion at m/z 769.5392 [M + H]⁺ (calcd 769.5402) and ¹³C NMR data (Table S1). In the upfield region of its ¹H NMR spectrum, seven singlet methyls resonating at δ_H 0.62 (s, H₃-20′), 0.78 (s, H₃-18), 1.09 (s, H₃-28/H₃-29), 1.11 (s, H₃-19), 1.16 (s, H₃-19′), 1.18 (s, H₃-26), and three doublet methyls at 0.74 (d, J = 6.5 Hz, H₃-21), and 1.04 (d, J = 6.5 Hz, H₃-15′ and H₃-16′) were observed (Table S2). In the down-field region, three olefinic protons were recognized, with two [δ_H 4.72, 4.46 (each br s)] ascribed to an exocyclic methylene group and the rest (δ_H 5.36, br s) for a trisubstituted double bond. The ¹³C NMR data of 1 (Table S1), with the aid of DEPT and HSQC NMR experiments, exhibited 50 carbon resonances comprising two ketocarbonyls (δ_C 216.7 and 213.8), two carboxyls (δ_C 186.6 and 184.2), six olefinic carbons, eight sp³ quaternary carbons, seven sp³ methines, 15 sp³ methylenes, and 10 methyls. In combination with the number of the methyls and methyl-derived groups (i.e., exocyclic methylenes and carboxyls), the observation of 50 carbon signals and 15 indices of hydrogen deficiency indicated that compound 1 is a hetero-dimer of a C₃₀-triterpenoid moiety fused with a C₂₀-diterpenoid unit.

For the triterpenoid part, a close comparison of the NMR data of 1 with the co-occurring neoabiestrine F,^[12] a rearranged lanostane with an intriguing spiro[4,4]nonane ring, permitted the presence of a similar spiro-lanostane segment in the structure of 1. More details came from the interpretation of its 2D NMR data. The presence of two discontinuous spin systems of CH₂(15)-CH₂(16) and CH₃(21)-CH(20)-CH₂-(22) as evidenced from the ¹H-¹H COSY correlations, and the HMBC correlations from H₃-18 to the spiro-carbon (C-13) and C-16/C-17/C-20, from H₂-30 to C-13/C-14/C-15, and from H₃-19 to C-9, confirmed the assignment of an 8(14→13)-abeo-17,13-friedo-lanost-8,14(30)-diene^[13] fragment for 1 (Figure 3 a). Consistent with this, the observed quaternary carbon signal at δ_C 67.9 was typical for a double allyl nodal carbon atom (C-13) in this class of rearranged 6/6/5/5-fused lanostanes.^[13] The carbonyl at δ_C 216.7 was assigned to C-3 based on its HMBC correlations with the geminal methyls H₃-28 and H₃-29 as a keto group, whereas the correlations from H₂-22 and H-24 to a carbonyl at δ_C 213.8 allowed it to be positioned at C-23 as an additional keto group. The HMBC spectrum of 1 exhibited correlations from H₃-26 to C-24, C-25, and a carboxyl at δ_C 184.2, indicating C-27 to be carboxylated.

Figure 3.

2D NMR correlations of forrestiacid A (1).

The remaining 20 carbon resonances comprising a carboxyl, four methyls, six methylenes, five methines, and four quaternary carbons, constructed a diterpenoid unit. The presence of an isopropyl group [δ_H 1.04 (6 H, d, J = 6.5 Hz), H₃-16′/H₃-17′; 2.34 (m), H-15′], which in conjunction with the HMBC correlations from H-14′ to C-7′/C-12′, from H₃-16′(17′) to C-13′/C-15′, from H₃-19′ to C-3′/C-4′/C-5′/C-18′, and from H₃-20′ to C-1′/C-5′/C-9′/C-10′ (Figure 3 a), further depicted this diterpenoid unit featuring an abieta-13-ene skeleton. The carboxyl group at δ_C 186.6 was attached to C-4′ based on its HMBC correlation with H₃-19′.

Overall, the triterpene unit was fused with the diterpene via two newly generated C—C bonds of C-25-C-8′ and C-24-C-12′ to form an additional cyclohexene ring (Figure 3 a). This was evidenced by the HMBC correlations of H₃-26/C-8′, H-24/C-13′, H-12′/C-24, and H-14′/C-25. The planar structure of 1 was thus delineated. Its relative configuration was established by analyses of proton-proton coupling constants (Table S2) and the ROESY spectrum. The ROE correlations (Figure 3 b) of H₃-19/H-2β, H-2β/H₃-29, H₃-28/H-5, H-5/H-7α, H-7β/H₃-18, and H-30b/H-7α ascertained the relative configuration of the triterpenoid moiety to be consistent with those of neoabiestrine F and analogs.^[12] Meanwhile, the ROE correlation between H₃-19′ and H₃-20′, and those between the axial protons of H-5′ (δ_H 1.69, br d, J = 12.0 Hz) with H-1′α and of H-1′α with H-9′ established the relative configuration of the A/B rings in the diterpenoid moiety as those of usual abietanes^[14] (Figure 3 b). The ring C in the diterpenoid portion was fused with the newly formed cyclohexene to form a rigid bicyclo-[2.2.2]octene core and its stereochemistry was determined as shown in Figure 3 b. Analysis of the proton-proton coupling constants revealed that H-24 (br s) and H-12′ (br s) were both equatorially oriented due to their small J values. The strong ROE correlation of H-14′ with H₃-20′ positioned the bridge fragment (C-13′ to C-17′) on the β-side of this molecule. H-24 was concluded to adopt the β-orientation from its ROE correlations with H-14′ and H₃-16′(17′), whereas Me-26 α-configured owing to the ROE correlation of H₃-26/H-9′. The gross structure of 1 was then elucidated as depicted in Figure 1.

Forrestiacid B (2) has the same molecular formula (C₅₀H₇₂O₆) as 1 according to the protonated ion at m/z 769.5406 in its HRESIMS. Its ¹H and ¹³C NMR data (Tables S1 and S2) closely resembled those of 1, with noticeable differences occurring at CH-24 (δ 3.35/59.0 in 2 vs. δ 2.79/62.2 in 1), C-25 (δ_C 55.8 in 2 vs. δ_C 49.8 in 1), and their neighboring groups. Further interpretation of the HMBC spectrum of 2 revealed its planar structure to be the same as 1, implying that the stereogenic centers C-24 and C-25 in the [4+2]-cyclohexene ring of 2 would adopt different configurations. This inference was confirmed by the ROESY experiment, in which the cross-peaks of H₃-20′/H-14′, H-14′/H₃-26′, and H-24/H-9′ demonstrated the β-orientation for Me-26 but an α-orientation for H-24 (Figure S2).

To further consolidate the relative configuration assignments for C-24 and C-25 of the above diastereomers 1 and 2, GIAO (the Gauge Independent Atomic Orbital method) ¹H and ¹³C NMR chemical shift (CS) calculations using the corrected mean absolute error (CMAE) and DP4 + probability analyses were performed (for details, see Table S6–S14 in the Supporting Information).^[15] The results obtained through DP4 + probability analysis predicted (24S,25R)-1 to match well with the experimental data of 1 with 100 % probability [DP4 + (all (H and C) data) and scaled DP4 + (all (H and C) data)] (Table 1 and Table S13). Similarly, (24R,25S)-2 matches closely with the experimental data of 2 with 100 % probability [DP4 + (H and C data)] (Table 1 and Table S14). In addition, for scaled CS values (Table S9–S12), the carbon- and proton-CMAE analysis of (24S,25R)-1 and (24R,25S)-2 (Table 1) showed better agreement with the DP4 + study. Overall, the CMAE and DP4 + probability analyses verified the relative configurations of compounds 1 and 2 to be (5R*,10S*,13R*,17S*,20R*,24S*,25R*,4′R*,−5′R*,8′S*,9′R*,10′R*,12′R*) and (5R*,10S*,13R*,17S*,20R*,−24R*,25S*,4′R*,5′R*,8′S*,9′R*,10′R*,12′R*), respectively.

Table 1:

Mean absolute error (MAE), corrected MAE, and DP4 + probability analyses (sarottinmr.weebly.com) for (24S,25R)-1 and (24R,25S)-2. The PCM/mPW1PW91/6–311 + G(d,p) level of theory was used for the analysis.

	1		2		Units
	calc. (24S,25R)-1	calc. (24R,25S)-2	calc. (24S,25R)-1	calc. (24R,25S)-2
All DP4 + data	100.00	0.00	0.00	100.00	%
MAE (carbon)	4.93	5.32	4.90	5.02	ppm
MAE (proton)	0.20	0.28	0.22	0.23	ppm
CMAE (carbon)	1.86	2.56	2.17	1.96	ppm
CMAE (proton)	0.17	0.22	0.18	0.17	ppm

The reliability of the absolute configuration assignments was then corroborated by the electronic circular dichroism (ECD) calculation using time-dependent density functional theory (TDDFT).^[16] In the ECD curve of 1, a positive Cotton effect (CE) at 222 nm was observed (Δε + 13.35), in contrast to a negative one for 2 (Δε−27.16, 227 nm) (Figure S3). Such a CE around 220 nm is aroused by the electronic transitions involving the electrons of the carbonyl at C-23 and the Δ^13′(14′)-olefinic bond.^[17] As shown in Figure S3, the respective similarities between the experimental and computed ECD spectra of 1 and 2 allowed the determination of (5R,10S,13R,17S,20R,24S,25R,4′R,5′R,8′S,9′R,10′R,12′R)−1 and (5R,10S,13R,17S,20R,24R,25S,4′R,5′R,8′S,9′R,10′R,−12′R)-2. Finally, the above assignments (including both relative and absolute configurations) were unambiguously confirmed by a single-crystal X-ray crystallography study of 1 with Ga Kα radiation [absolute structure parameter 0.17-(11)] (Figure 4).^[18]

Figure 4.

ORTEP representation of forrestiacid A (1).

Considering that neoabiestrine F (Scheme 1),^[12] a rearranged lanostane with an intriguing spiro[4,4]nonane moiety to be the precursor for the isolates, and several abietenes (e.g., dehydroabietic acid^[14]) were found co-occurred in the title plant, it is reasonable to propose that the hetero-dimers would be formed by an intermolecular [4+2] Diels–Alder (DA) cycloaddition between the triterpene and diterpene units. The diterpenoid levopimaric acid^[19] would act as the ‘diene’ in the DA reaction. Meanwhile, neoabiestrine F, a rearranged lanostane bearing an 23-oxo-24(E)-enoic acid fragment could serve as the electron-deficient ‘dienophile’. The triterpenoid and diterpenoid units were connected via the newly formed C-25-C-8′ and C-24-C-12′ bonds. Interestingly, in this study, 1 and 2 were isolated as a pair of natural isomers derived from exo-/endo-selectivities in [4+2] dimerization, but with the endo-cycloadducts (1) as the favored one (isolated amount 1.2 g for 1 and only 0.1 g for 2). Besides, the [4+2] adducts 1 and 2 derived from the E-configured dienophile adopt the cis ring fusion mode (i.e., 27-COOH and H-24 being cofacial), as illustrated.

Scheme 1.

Proposed biosynthetic pathway for forrestiacids A (1) and B (2).

The DA reaction has been an extremely valuable synthetic method for the construction of unsaturated six-membered rings.^[20] A number of naturally occurring terpenoid [4+2]-adducts have been reported, and have attracted attention due to their complex architectures, bioactivities, and synthetic challenges.^[21] Among them, most are terpenoid homo-dimers, whereas the hetero-dimers are quite rare.^[21–23] As for the triterpene-diterpene [4+2] adducts, only three examples have been encountered so far: two cycloartane-labdane adducts from Pseudolarix amabilis (Pinaceae)^[24] and a lupane-abietane adduct from two Hippocratea species (Celastraceae).^[25] Forrestiacids represent a novel chemical class of [4+2] hetero-dimers formed between a lanostane-derived triterpene and an abietane-type diterpene.

Biosynthesis of natural [4+2]-adducts occurs via either non-catalytic spontaneous dimerization or enzyme-catalytic reaction. In the present study, spontaneous reactions between neoabiestrine F and levopimaric acid either at room temperature for a period of 6-month or under refluxing in toluene overnight were performed, but no expected adduct could be detected by TLC and HPLC analysis (details, please refer to the experimental section in the Supporting Information). Natural Diels–Alderases^[26] might be responsible for their biosynthesis, which warrants future exploration.

The two terpenoid [4+2]-adducts were evaluated in ACL assays to assess their activity.^[14] The known ACL inhibitor BMS 303141^[27] was used as the positive control. As illustrated in Table 2, forrestiacids A (1) and B (2) displayed significant inhibition against ACL, with IC₅₀ values of 4.12 and 3.57 μM, respectively. In contrast, the two possible precursors, neoabiestrine F and levopimaric acid, were inactive (IC_50s > 20 μM). Hence, the existence of the bicyclo[2.2.2]octene core is crucial for the ACL inhibitory activity of the hetero-dimers. As evidenced by the data, the stereoisomers (endo- or exo-type) showed comparable effects. ACL inhibitors may suppress the synthesis of fatty acids and cholesterol, and significantly lower the elevated level of LDL-C.^[5] Interestingly, forrestiacids and bempedoic acid^[5] all are dicarboxylic acid derivatives.

Table 2:

ACL inhibitory effects of compounds 1 and 2.

Compounds	IC50 (μM)[a]
1	4.12±0.62
2	3.57±0.39
BMS 303141[27]	0.46±0.13

^[a]These data are expressed as the mean ± SEM of triplicate experiments.

The major component, forrestiacid A (1, yield 0.008 %), was then subjected to a de novo lipogenesis assay (Figure 5). It elicited dual inhibition on [¹⁴C]-labeled acetate incorporation into fatty acid and cholesterol in HepG2 cells, in a concentration-dependent manner. Remarkably, the treatment of compound 1 resulted in dramatic reductions in fatty acid (≈75 %, P < 0.05), and cholesterol (≈93 %, P < 0.05) syntheses at all test concentrations (10, 20, and 40 μM). The positive control, 1 mM metformin caused a dual decrease in fatty acid and cholesterol, only by 33.59 and 30.04 %, respectively. Due to the modest activity against ACL shown in Table 2 relative to the highly significant activity in lipogenesis shown in Figure 5, a docking study with ACL and four other protein targets laying key roles in lipogenesis (i.e., ACC, FAS, SCD1, and PPAR-α, for details see SI) revealed a potential synergistic interaction with two or more of these targets leading to the potent inhibition of lipogenesis. This further elevates the potential value of these molecules as therapeutic agents in the control of lipogenesis.

Figure 5.

Effects of compound 1 on de novo lipogenesis in HepG2 cells. Detection of syntheses of fatty acid (A) and cholesterol (B) in HepG2 after treated with 1 for 20 h and incubated with ¹⁴C labeled acetate for another 4 h. n = 3 for all groups. * P < 0.05 compared with DMSO group. All error bars, s.e.m.

In summary, two new pentaterpenoids [4+2] hetero-dimers were isolated and characterized from the vulnerable conifer P. forrestii, under the guidance of MS/MS-based MoIN. The intriguing pentaterpene-like skeleton comprises a rearranged 6/6/5/5-fused lanostane-type triterpene unit conjugated with an abietane-type diterpene unit by an intermolecular Diels–Alder cycloaddition. The applications of quantum NMR calculations with the aid of DP4 + probability analysis provide ideal complementary tools to facilitate the elucidation process in some challenging cases.^[28] Forrestiacids represent one of the largest terpenoid molecules published to date,^[28] and expand the utility of computational chemical shift NMR approaches as effective and powerful strategies for assign and verify the correct NMR structure. Both exhibited potent inhibitory activities against ACL (IC₅₀s < 5 mM). Forrestiacid A (1) is shown to significantly inhibit the de novo fatty acid and cholesterol synthesis in HepG2 cells. These unique naturally occurring lipogenesis inhibitors reveal the importance of protecting endangered plants and interrogating traditional Asian and Native American medicines as potential sources of new and complex NP therapeutics.