Synthesis of a Library of Terpenoid Alkaloid–Like Compounds Containing Medium-Sized Rings via Reconstruction of the Humulene Skeleton

Graphical Abstract

We synthesized a library of terpenoid alkaloid-like compounds with diverse heteroatom-containing ring systems. Our strategy involves the introduction of a nitrogen atom into a humulene skeleton and subsequent ring reconstruction. A tartrate-resistant acid phosphatase (TRAP) assay using a receptor activator of nuclear factor-κB ligand (RANKL)-induced RAW264.7 cells allowed identifying a seed compound for bone-resorptive diseases.

Abstract

The construction of a chemical library based on natural products is a promising method for the synthesis of natural product-like compounds. In this study, we synthesized a terpenoid alkaloid-like compound library based on the humulene skeleton. Our strategy, which enables access to diverse ring systems such as 11-membered monocyclic, oxabicyclic, and medium-sized aza ring-containing scaffolds, involves the introduction of a nitrogen atom, an intermolecular C−O bond formation via Lewis acid-mediated epoxide-opening transannulation, and a ring-reconstruction strategy based on olefin metathesis. A cheminformatics analysis based on their structural and physicochemical properties revealed that the synthesized compounds have high three-dimensionality and high natural product likeness scores but with structural novelty. The usefulness of the terpenoid alkaloid-like compound library for drug discovery and the accessibility to structure-activity relationship studies were validated by performing an assay for osteoclast-specific tartrate-resistant acid phosphatase activity, resulting in the identification of a seed compound for bone-resorptive diseases such as osteoporosis.

Introduction

Natural products are essential in drug discovery owing to their structural diversity and biological activities. As a result of chemical research in this field, around 40 % of clinical drugs have been developed from natural products and their derivatives.¹ However, obtaining skeletally novel compounds from nature is challenging, which has caused a decline in pharmaceutical research utilizing natural compounds.^{2, 3} The construction of a chemical library based on complex natural products is a promising method for the synthesis of natural product-like compounds featuring a high percentage of sp³ carbons and high three-dimensionality. In fact, intense efforts have been devoted to the development of natural product skeletons to construct compound libraries that are structurally diverse and useful for drug discovery.^4-7

Terpenoid alkaloids have been used for a long time as pharmaceutical resources due to their structural diversity and promising biological activities. This class of natural compounds is characterized by having simultaneously nitrogen-containing alkaloid scaffolds and sp³-rich terpenoid scaffolds.^8-10 However, only limited terpene skeletons, such as secologanin⁹ and ent-kaurene,¹¹ are used as a source of terpenoid alkaloid moieties (Figure 1a). In addition, alkaloids containing medium-sized aza rings belong to a useful and rare class of natural products consisting of manzamine A¹² and nakadomarin A.¹³ Therefore, the construction of terpenoid alkaloid-like compounds containing medium-sized rings using terpene skeletons that are not employed in natural biosynthesis will be useful for drug discovery.

Figure 1

(a) Structures of reserpine and aconitine. Red bonds show terpenoid skeletons. (b) Aniline-introduced terpenoid alkaloid-like compounds based on the humulene skeleton.

Previously, we constructed terpenoid alkaloid-like compound libraries based on the humulene skeleton,^{14, 15} a 11-membered-ring-containing sesquiterpenoid that is widely found in plants such as hops.¹⁶ This library included compounds with diverse ring systems, and cheminformatics and biological investigations demonstrated their utility in drug discovery.¹⁴ However, in this strategy, anilines were used as the nitrogen source. Aniline is included in “structural alerts,” which are functional groups found in drugs and referred to as toxicophores.¹⁷ For instance, acetaminophen is oxidized by P450 to an electrophilic species and forms a covalent bond with glutathione. This results in a decrease in the hepatic GSH concentration and the covalent modification of hepatic proteins, causing the inhibition of cellular protein function and cell death.^18-21

In this study, we constructed a humulene-based terpenoid alkaloid-like compound library, which contains sp³-rich structures and medium-sized aza-ring systems, by introducing aliphatic amines to avoid structure alerts and molecular flatness due to the aniline moiety. Our strategy involved the introduction of a nitrogen atom, an intermolecular C−O bond formation via Lewis acid-mediated epoxide-opening transannulation, and a ring-reconstruction based on olefin metathesis, enabling access to diverse ring systems. Herein, we describe the synthetic strategy, the synthesis of a chemical library, and the evaluation of its usefulness for drug discovery by means of cheminformatics and biological analyses.

Results and Discussion

Our strategy for constructing a terpenoid alkaloid-like library is shown in Figure 2. Humulene diepoxide was selected as a starting material. Aliphatic amine-introduced medium-sized oxabicyclic compounds were obtained via sequential epoxide-opening reactions and subsequent transannulation. Then, a ring reconstruction strategy involving a cascade ring-opening/ring-closing metathesis reaction between an internal double bond in the humulene skeleton and an allyl or butenyl group furnished diverse medium-sized aza ring-containing compounds (Figure 2). Humulene diepoxides 1 and 1’ were prepared as a 7 : 3 diastereomeric mixture according to a previous work.¹⁴ A screening of nucleophiles and Lewis acids revealed that the combination of benzylamine and lanthanum(III) triflate afforded nitrogen-introduced products in good yields. We decided to use 3,4-dimethoxybenzylamine as a nucleophile because it could be easily removed. Subsequently, humulene diepoxides 1 and 1’ were reacted with 3,4-dimethoxybenzylamine in the presence of lanthanum(III) triflate, followed by N-alkylation with allyl bromide or but-3-en-1-yl trifluoromethanesulfonate. After separation, compounds 2 and 3 were obtained as major products, corresponding to the alkyl groups introduced, along with compounds 4–6 as minor products (Scheme 1). Except for 4, the synthesized compounds contained medium-sized rings, i. e., a 10-membered oxabicyclic ring in 2, an 11-membered ring in 3, and 9-membered oxabicyclic rings in 5 and 6. A plausible mechanism for the synthesis of these compounds can be briefly described as follows. Bicyclic compound 2 would be produced by the nucleophilic attack of 3,4-dimethoxybenzylamine on the C-2 position of 1 and sequential epoxide opening reaction. ¹H and ¹³C NMR spectra of 2 a and 2 b were obtained as a mixture of two compounds but showed the same spectra after separation by HPLC. This supports that 2 a and 2 b exist as a mixture of conformational isomers, as previously reported.¹⁴ Meanwhile, 11-membered ring compound 3 would be formed from a bicyclic tetrahydrofuran intermediate derived from 1’, although the mechanism for the cleavage of the C−O bond is unclear. Tetrahydropyran 4, which was also isolated in a previous work, would be most likely formed via the cleavage of the C-8−C-9 bond of the bicyclic tetrahydropyran intermediate in a retro-ene reaction. Finally, the nucleophilic attack of 3,4-dimethoxybenzylamine on the C-2 position of 1 or 1’ and sequential nucleophilic addition to an olefinic C-10 carbon, followed by epoxide-ketone rearrangement, respectively, would afford bicyclic compounds 5 and 6.

Figure 2

Synthetic plan for terpenoid alkaloid-like compounds based on the humulene skeleton by introducing aliphatic amines.

Scheme 1

Synthesis of terpenoid alkaloid-like compounds.

Next, we applied the ring reconstruction strategy to 2 a, 2 b, 4 a, and 4 b. The ring-closing metathesis of compounds 4 a and 4 b using a second-generation Grubbs’ catalyst afforded medium-sized aza ring-containing bicyclic compounds 7 a and 7 b, respectively. The ring-rearrangement metathesis reaction of 2 a and 2 b in the presence of Stewart-Grubbs’ catalyst and ethylene gas afforded 7-membered and medium-sized aza ring-containing compounds 9 a and 9 b as well as ring-opening product 8 b. Compounds 9 a and 9 b were not obtained in the absence of ethylene gas and were probably generated through 8 a and 8 b. (Scheme 2).

Scheme 2

Ring-rearrangement of terpenoid alkaloid-like compounds.

Finally, the 3,4-dimethoxybenzyl groups of the synthesized compounds were removed using 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ), affording terpenoid alkaloid-like compounds 10–16 in moderate yields (see Supporting Information). Meanwhile, 17 b was obtained from 9 b via deprotection of 3,4-dimethoxy benzyl groups (Scheme 3). The formation of 17 b can be considered to proceed via a N-radical cation formed by DDQ, which mediated C−N bond formation and cleavage to construct the pyrrolidine moiety. Our humulene-based terpenoid alkaloid-like compound library contains 31 compounds with various ring systems, such as 11-membered monocyclic, oxabicyclic, and medium-sized aza ring-containing scaffolds (Figure S1).

Scheme 3

Deprotection of 9b.

Next, a cheminformatics analysis was performed to gain insight into the structural and physiochemical properties of the humulene-based terpenoid alkaloid-like compound library using a data set of randomly selected natural products, FDA-approved drugs, and synthetics obtained from the publicly available databases COCONUT online²² and ZINC-20.²³ Moreover, the web applications NaPLeS²⁴ and NP-Scout²⁵ were used to identify natural product-likeness. Specifically, NaPLeS can be used to calculate natural product-likeness scores based on the environment of the atoms in a molecule, that is, the number of fragment occurrences found in natural products or synthetics.²⁴ Meanwhile, NP-Scout can quantify the natural product-likeness and identify the atoms that contribute to the classification of small molecules as a natural product or a synthetic molecule.²⁵ Quantifying the natural product-likeness using computational methods helps guide the synthesis of natural product-like compounds and optimize the natural product-likeness of lead compounds. As shown in Figure 3a and b, the compounds included in our library showed high scores (NP-likeness score>0, NP class probability>0.5), indicating that they contain substructures commonly found in natural products. According to the similarity mapping generated by NP-Scout, the 3,4-dimethoxybenzyl moiety led to decreased similarities to natural products, whereas core skeletons derived from humulene, including scaffolds rearranged via olefin metathesis, showed high similarities (Figure 3c). The principal component analysis shown in Figure 3d reflects well these results, that is, the distribution of our compounds is in good agreement with that of the natural products. Conversely, Tanimoto similarity scores, which represent molecular similarity based on molecular fingerprints between molecule pairs,²⁶ revealed a low structural similarity to reported natural products with average Tanimoto scores between a terpenoid alkaloid-like compound and natural products of 0.12. A similarity score of less than 0.4 was set by Pye et al. as a criteria of structural novelty,²⁷ indicating that the synthesized compounds contain unique scaffolds that are not biosynthesized in nature.

Figure 3

Cheminformatics analysis of terpenoid alkaloid-like compounds based on the humulene skeleton. (a) Ratio of each NP-likeness score of terpenoid alkaloid-like compounds (31 compounds, red line), natural products selected randomly from COCONUT (n=880, green line), drugs selected randomly from ZINC-20 (n=500, blue line), and synthetics selected randomly from ZINC-20 (n=500, black line). (b) Number of compounds in each similarity score for terpenoid alkaloid-like compounds. (c) Examples of similarity maps generated by NP-Scout. (d) Principal component analysis (PCA) plot for terpenoid alkaloid-like compounds (31 compounds, red dots), drugs selected randomly from ZINC-20 (n=500, not shown in Figure 3d), and natural products selected randomly from COCONUT (n=500, green dots). (e) Principal moment of inertia (PMI) plot for terpenoid alkaloid-like compounds (31 compounds, red dots), natural products selected randomly from COCONUT (150 compounds, green dots), and drugs selected randomly from ZINC-20 (150 compounds, blue dots). The dashed red line shows the 3D score of 1.07.

Finally, to evaluate the molecular shape of our compounds, a principal moment of inertia analysis was conducted,^28-30 finding that our library overlapped with drugs and natural products and the density of terpenoid alkaloid-like compounds was shifted away from the disk-like axis. The average of the 3D score (I₁/I₃+I₂/I₃) among terpenoid alkaloid-like compounds was 1.19. The 3D score represents the flatness of a molecule. Specifically, a 3D score greater than 1.07^{31, 32} indicates that the molecule is highly steric, which is one of the factors that make natural products promising in drug discovery.^32-34 The results of the cheminformatics analysis illustrate that the humulene-based terpenoid alkaloid-like compound library can approach unexploited chemical space with high three-dimensionality.

To identify potential pharmacologically active compounds among the synthesized terpenoid alkaloid-like compounds, their biological activities were extensively screened, and some compounds showed osteoclastogenesis-suppressive activity. Bone remodeling consists of two processes, i. e., bone formation and bone resorption. Osteoclasts and osteoblasts play important roles in bone resorption and cooperate to maintain the mechanical competence of bones.^{35, 36} Aging or diseases such as osteoporosis or rheumatoid arthritis causes an imbalance of these two types of cells in the regulation of the bone remodeling processes. The receptor activator of the nuclear factor-κB ligand (RANKL) is supplied by osteoblasts, inducing a signal transduction and expressing downstream genes, which result in the differentiation of the osteoclast precursor into multinucleated bone-resorptive cells. Therefore, the RANKL/RANK signaling pathway is a promising target for the treatment of bone metastasis. Moreover, a humanized anti-RANKL antibody, denosumab, is used to treat osteoporosis and other bone diseases.³⁷ However, no small-molecule RANKL-induced osteoclastogenesis inhibitors are used in clinical trials.

Tartrate-resistant acid phosphatase (TRAP) is a glycosylated monomeric metalloprotein enzyme expressed in mammals³⁸ and a differentiation marker of osteoclasts. TRAP activity reflects the balance of bone remodeling and is used to screen inhibitors of osteoclast differentiation. To verify the potential of the terpenoid alkaloid-like compound library to treat osteoporosis and other bone diseases, we investigated the inhibitory effect on RANKL-induced TRAP activity in RAW264.7 cells (Figure S3). The results showed that compounds 2–9 with 3,4-dimethoxybenzyl groups inhibited TRAP activity. Conversely, deprotected compounds 10–17 did not show inhibition, suggesting that the 3,4-dimethoxybenzyl group is essential for the inhibitory activity. Compounds 5 and 6 containing a 12-oxabicyclo[6.3.1]dodecane moiety exhibited the strongest inhibitory activity, with the allyl group being a suitable substituent (5 a vs. 5 b). Bicyclic compound 5 a showed potent TRAP inhibitory activity with an IC₅₀ value of 4.8 μM and without remarkable cytotoxicity (Figure 4). A known small-molecule RANKL inhibitor, AS2676293, was developed for reducing bone metastasis of breast cancer cells; however, it is not used in clinical and its inhibitory activity with an IC₅₀ of 0.15 μM, as confirmed in a TRAP staining assay, is lower than that of 5 a.³⁹ Therefore, since 5 a is expected to be a seed compound for new RANKL inhibitors, we performed preliminary structure-activity relationship studies based on the structure of 5 a. Compounds 18–20 were synthesized according to the methodology used for the construction of the terpenoid alkaloid-like compound library. The results showed that 5 a still exhibited the strongest inhibitory activity, revealing that an electron-rich benzylamine is a suitable nucleophile. In particular, the 3,4-dimethoxybenzyl group was found to play an important role in TRAP inhibitory activity, even though contributed to the decrease in similarities to natural products. The use of unnatural nucleophiles can be expected to be useful in the development of natural product-like compound libraries that are complementary to chemical space of existing natural products. Overall, these results demonstrate that our methodology not only generates privileged resources for drug discovery but also enables access to structure-activity relationship studies.

Figure 4

Inhibitory effect of TRAP activity in RANKL-induced RAW264.7 cells. (a) Inhibitory effects of TRAP activity and cell viability of 5 a and its derivatives (n=3 samples per group). The IC₅₀ values for inhibition of TRAP activity are 4.8 μM for 5 a, 7.1 μM for 18, 34 μM for 19, and 9.5 μM for 20, respectively. (b) TRAP staining images of RANKL-induced osteoclast and effect of 5 a on osteoclast differentiation.

Conclusions

In conclusion, we have described the synthesis of a humulene-based terpenoid alkaloid-like compound library containing 31 compounds with various ring systems. The compounds were synthesized by introducing an aliphatic amine via Lewis acid-catalyzed epoxide-opening reaction and ring-reconstruction via olefin metathesis. Especially, the ring-reconstructed molecules are characterized by medium-sized aza ring moieties, which are rare in nature but found in bioactive natural products. Our strategy allows optimizing the ring size by changing the electrophile used for N-alkylation. A cheminformatics analysis revealed that the synthesized compounds have high three-dimensionality and high natural product-likeness scores but with structural novelty. The similarity maps calculated by NP-Scout indicate that 3,4-dimethoxybenzyl moieties reduced the natural product-likeness scores; however, they can be easily removed and converted to other functionalities. The biological screening demonstrates the usefulness of this library as a promising source for drug discovery and to perform structure-activity relationship studies. Compound 5 a was identified as a promising seed compound for the treatment of osteoporosis or other bone diseases based on a TRAP activity assay. This strategy, which is characterized by the introduction of a nitrogen atom and ring reconstruction, can be applied to compounds containing medium-sized rings or macrocycles with olefins and will help develop valuable compounds for drug discovery.

Supporting Information Summary

The authors have cited additional references within the Supporting Information.^40-45

Acknowledgments

This work was supported by Grants-in-Aid for Scientific Research (no. 19H02837) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), the Uehara Memorial Foundation, and Tokyo Biochemical Research Foundation.

Conflict of Interests

The authors declare no conflict of interest.

Open Research

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.