Conformational dynamics and molecular interactions of natural products: unveiling functional structures in biological membranes

Abstract

Structural studies of natural products have been a driving force in the development of organic chemistry throughout its long history, especially in the early years. Recently, structure determination based on new concepts has also gained momentum. In this review we will mainly discuss the functional structures of natural products that account for the mechanisms of action largely from our studies. The topics include marine natural products, amphidinols, and ladder-shaped polyether compounds, which are known as potent antifungal agents and important marine biotoxins, respectively. Nuclear magnetic resonance studies for determining the stereochemistry of amphidinol 3 and its conformation in lipid-bilayer membranes will be presented in detail.

Introduction

Historically, the structural determination of natural products has advanced with progress in methodology for elucidating the bonding patterns of all constituent atoms.¹⁾ Structure elucidation is usually completed by assigning absolute stereochemistry, and the resulting molecular structure is used as searchable information by linking it to the compound name, and as necessary and sufficient information for total synthesis and other chemistry-related research. Additionally, analysis of chemical structure, especially stereochemistry, is crucial for elucidating the 3D shape of natural products with biological activities, which is referred to as ‘functional structure’ in this review; this structure is also called active conformation or competent conformation. It is generally known that when natural products or drugs bind to receptor proteins, the shape of the molecule has to change according to the shape of the binding site (induced fit). In this case, the conformational change is in a trade-off relationship between the additional energy required for the change (difference from the conformation in an unbound state) and the energy gained from binding. Therefore, it should be considered that stable conformers of natural products and drugs will change to some extent depending on the environment in which the compound is placed. In other words, the optimal molecular conformation differs depending on the interacting protein (and other biomolecules) responsible for each function in the processes of drug delivery, drug metabolism, and receptor binding.^2),3) For example, after oral administration, a drug passes through the digestive and circulatory systems to reach its target cell (e.g., cancer cell). The drug has the opportunity to bind to digestive enzymes, blood proteins, and drug-metabolizing enzymes before binding to the target receptor. Thus, the affinity of the drug for each of these proteins may have a significant impact on its life span and efficacy. In addition, the conformation of a drug has a significant effect on its solubility in plasma, biological membranes, and cytoplasm, as well as membrane permeability, all of which are important for delivery efficiency; here, in addition to the receptor-binding conformer, each of the favorable conformers is a functional structure. Therefore, because the binding affinity to proteins in these biological environments and the partition coefficient in each medium depend on the conformation of the molecule, the delivery efficiency can be improved by changing the conformation on a case-by-case basis; e.g., when fatty acids bind to binding proteins to transfer to mitochondria for β-oxidation, their alkyl chain takes on metastable dihedral angles, allowing the hair-pin conformation to fit into the pocket of the protein without difficulty.^4),5) In other words, compounds with flexible conformations are more likely to efficiently exhibit biological activity.⁶⁾ It can also be noted that such compounds are likely to have multiple conformations even under conditions where they do not interact with biomolecules (e.g., solution nuclear magnetic resonance [NMR] measurement conditions).⁷⁾ Thus, the relationship between conformational change and biological activity is a very intriguing research topic.

It is a well-known fact that the structural analysis method to be used depends on the rate of conformational change. In usual NMR analysis, natural products are measured in solution; the faster the molecules move, the higher the resolution of the signal obtained, and the easier the structural analysis becomes. On the other hand, it is not possible to observe each of the rapidly exchanging conformers, and the measured value (chemical shift or spin coupling constant) is the average of the conformers. In usual conditions for solution NMR, it is possible to consider that stable conformers are exchanging rapidly. Using techniques for determining conformation in solution NMR, such as nuclear Overhauser effect (NOE) and spin-spin coupling constants, it is possible to deduce a dominant conformer(s).^8),9) In this review, we briefly introduce a solution NMR study of ladder-shaped polyether compounds that cause food poisoning as an example of the conformational change. In addition, NMR can be used to analyze the conformation of compounds with high flexibility, which are incorporated in the hydrophobic parts of lipid membranes or micelles. As an example, we introduce a solution NMR study using small bicelles for the membrane-bound natural product, amphidinol 3 (AM3).

The fast molecular motion required for high-resolution solution NMR measurements cannot be achieved under usual membrane conditions. Therefore, different methods are needed to elucidate the conformational changes of natural products bound to lipid bilayer membranes (functional structure). For example, investigating the shape of molecules using solid-state NMR¹⁰⁾ and Förster resonance energy transfer (FRET) are promising approaches. As an example of a highly flexible molecule, we will explain in detail assignment of the stereochemistry and analysis of the functional structure of the antifungal natural product AM3 based on our research. Through studies of AM3, we will show that diverse methodologies can be used to elucidate functional structures. Even for structural analysis methods that are now considered routine, the aid of organic synthesis and other experimental techniques is still necessary in many cases, in order to accurately determine the structure. Particularly, in the case of natural products such as AM3, which have chiral centers scattered throughout the chain structure, it is not a straightforward process even with the newest technology, and organic synthesis plays a crucial role. Furthermore, we would like to deepen the discussion of what the problems are in elucidating the structures of conformers in biologically relevant conditions and how they may be solved. The aim of this review is also to address the challenges in efficiently elucidating the functional structures of natural products.

1. Evolution of the concept of structure determination in natural product chemistry

Before discussing functional structures, we would like to review what ‘structure of an organic compound’ means. The concept of the structure of compounds appeared in the 19th century. In terms of stereochemistry, van’t Hoff and Le Bel put forward the theory that carbon atoms have a tetrahedral structure, and the concepts of chiral carbon and optical isomerism introduced by Pasteur and Biot (and other chemists) were established around this time. In this process, natural products played an important role in forming the concept of the structure of organic compounds. Early chemists considered plant and animal extracts to be special substances that cannot be synthesized artificially, and they called them organic compounds. The general view at the time was that organic compounds can only be produced by living organisms. In the first half of the 19th century, F. Wöhler was one of the chemists who proved this idea wrong by demonstrating that the organic compound urea could be synthesized from inorganic substances. Since then, it has become widely accepted that organic compounds, regardless of whether they are naturally occurring or synthetic, obey the same natural laws as inorganic substances. As a matter of consequence, the structures of natural products were defined and determined in accordance with these rules and theories, which can be applied to all compounds. The key point is to determine the structure by clarifying the way all the interatomic bonds are formed, including the absolute stereochemistry derived from chiral centers, and this concept has continued basically to the present day.

This definition of ‘structure of an organic compound’ may be missing some important information. For example, ring conformation and slow single-bond rotation are often excluded from this definition. The need to determine these structural isomers depends on the lifetime of each isomer. That is, if the isomers can be separated at room temperature using conventional methods such as chromatography or crystallization, and if there is little or no isomerization in the time required to characterize the compound, such as by spectroscopy or biological assays, then the isomer is a target of structural determination. On the other hand, the energy barrier required for the internal rotation of usual C–C bonds is only a few kcal, and the rotation rate is fast enough at room temperature that it is not possible to separate the individual rotamers. In this way, the idea that the structure to be determined must be stable enough to be isolated at room temperature became common, and it was considered reasonable from both a fundamental scientific and technological standpoint; the activation free-energy required for these structure-isomerizing processes is generally assumed approximately 20 kcal/mol or more. Furthermore, structures defined in the conventional way provide sufficient information for chemical synthesis and, therefore, have been used in the determination of natural product structures. It is generally accepted in natural product chemistry that structural determination of a natural product is completed by the assignment of absolute stereochemistry. On the other hand, research to elucidate functional structures, which are a step beyond traditional structures, has also been actively conducted in recent years. In the following section, we would like to discuss the current situation of this research field.

2. Strategy for elucidating the functional structures of natural products

As mentioned in the introduction, bioactive natural products are subject to selection and elimination during the delivery process to the target molecule. The functional structures for these processes, such as metabolic stability and membrane permeability, may considerably differ, but bioactive natural products must be able to pass through all of these without incurring a significant energetic penalty. Therefore, it is highly likely that such natural products have a flexible three-dimensional structure. The functional structure of natural products is roughly synonymous to the three-dimensional shape or conformation while its flexibility upon interaction with biomolecules (proteins, nucleic acids, glycans, and lipids) is also an important factor. The structures of complexes of proteins (receptors) and binding molecules (ligands) have been determined mainly by means of crystal X-ray diffraction and cryo-electron microscopy, leading to major advances in life sciences and drug discovery. For natural products and drugs that target proteins, if the protein can be purified with its intrinsic three-dimensional structure, the conformation of the bound molecule at the binding site can be considered a functional structure; the external environment hardly affects the shape of the binding molecule. Therefore, structural research of functional structures is most advanced for ligands binding to proteins with relatively robust structures. A vast amount of knowledge has been accumulated about protein-ligand complexes; however, it is beyond the scope of this review and we will only cite a few review articles relevant to drug discovery research.^11),12)

Many natural products and drugs interact with biomolecules other than proteins, such as nucleic acids, glycans, and lipids, but their functional structures have hardly been elucidated. Because DNA-binding natural products are important research targets for anticancer drugs, considerable effort has been made to clarify their binding structures; in particular, the DNA complex structures of actinomycin¹³⁾ and calicheamicin¹⁴⁾ have been studied extensively. Nevertheless, there are still few examples of successful structural elucidation. When it comes to lipids and glycans, traditional approaches in structural biology are particularly less effective. In addition, when natural products bind to glycans, they rarely form clearly defined binding pockets, making it difficult to accurately determine the binding site. Solid-state NMR has been applied to elucidate the functional structures formed by glycans and natural products¹⁵⁾; prademicin A, an antibiotic derived from actinomycetes that selectively recognizes mannose, has attracted attention.¹⁶⁾ Prademicin A is thought to form dimers in the presence of Ca²⁺ and interact with the hydroxy groups of mannose at positions 2, 3, and 4.¹⁷⁾ As for glycans, however, there have been few successful cases, because research strategies are not systematic and researchers need to search for appropriate methodologies on a case-by-case basis.

As discussed in the following sections, natural products usually do not adopt a well-defined 3D structure during membrane penetration. In addition, some natural products are known to exhibit distinctive biological activities upon binding to biological membranes, with structures that differ from those in solution. To elucidate functional structures in membranes, it is important to consider the orientation of natural products relative to the membrane surface, i.e., the angle of their molecular axes with respect to the bilayer normal, which is the vector perpendicular to the membrane plane. For example, when using lipid bilayers as model biomembranes, it is often difficult to reproduce certain conditions that appear in biological membranes, because natural products tend to aggregate or disperse at the molecular level within model membranes. Furthermore, as discussed below, it is often difficult to find appropriate parameters describing the conformational fluctuation and molecular orientation that characterize the functional structure of flexible natural product molecules present in model lipid membranes. Which analytical methods should be used to solve these problems in model lipid membranes and determine their functional structure? There has been extensive research into the preparation of model membranes that mimic the lipid bilayer portion of biological membranes.¹⁸⁾ For compounds that act primarily on biological membranes, such as saponins and polyene antibiotics, model membranes can be customized to reproduce their biological activity (and probably the functional structure). Major biological lipids are commercially available, which makes it relatively easy to fine-tune the lipid composition of model membranes. Additionally, the parameters that characterize the orientation and mobility of lipids in membranes can be used for natural products to some extent. For example, the lateral diffusion coefficient is often used to describe the mobility of lipid molecules. When a fluorescence-labeled compound is available, this value can be determined for any membrane-bound molecules using a fluorescence quenching experiment or fluorescence correlation spectroscopy. If natural products can be labeled with deuterium, it is possible to estimate the orientation of the entire molecule relative to the membrane plane and the fluctuations of the labeled moiety using ²H NMR spectroscopy. For example, the interaction between natural products and sterols in membranes can be precisely examined using ²H NMR (see Section 4.3). We will discuss in detail the functional structure of AM3, which efficiently forms pores in lipid membranes with its highly flexible structure.

Compared with progress in spectroscopic methods, the reliability and applicability of computational (or in-silico) methods have improved rapidly, and they are now frequently used to predict functional structures. In particular, molecular dynamics (MD) simulations have made it possible to predict the motion of individual molecules in systems of thousands of molecules over time periods of a few microseconds. When dealing with dynamic functional structures, computational methods prove to be the only practical way to visualize rapidly changing conformations. Another major advantage is that they can be applied to any kind of system, provided that a suitable initial state can be created. Therefore, if we can devote sufficient computational time to a molecular assembly such as a hydrated lipid bilayer, we may be able to obtain detailed information about the dynamics of molecular structures that are not obtainable by other methods. MD simulations have become an essential research methodology for the conformational analysis of natural products bound to biological membranes^19),20); later, we briefly describe MD simulations of the ion channel assembly of amphotericin B.

We then discuss the concept of “bimolecular interactions” that we employ to investigate the functional structures of membrane-bound molecules and membrane lipids. Biological membranes are extremely complex molecular assemblies that contain large amounts of proteins and various lipids, including glycolipids and sterols, and they are constantly changing their forms under non-equilibrium conditions. In such an environment, it is extremely difficult to accurately reproduce the interactions between natural products and lipids. On the other hand, because proteins and DNA maintain their original 3D structure even when removed from the living environment, it is possible to reproduce the intrinsic functional structure of molecular complexes of proteins or DNA with natural products in test tubes. These observations should also apply to some extent to lipid bilayers, provided certain prerequisites are met. As shown in Fig. 1, by observing only the labeled molecules (two blue lipid molecules are shown as an example in the figure), the molecular arrangement in a biological membrane (especially their interactions with neighboring molecules) can be reproduced with high accuracy in a model lipid bilayer. This idea is supported by the fact that model membranes can well mimic the physiological properties of biological membranes.^21),22) Thus, if a natural product is assumed to interact with a specific type of lipid in a biological membrane, the model membrane can be used to reproduce the natural product-lipid interaction, just as in the case of lipid-lipid interactions. We believe that this concept of bimolecular interactions also applies to natural products and drugs that interact with glycans and other endogenous molecules.

Fig. 1.

Limitations and advantages of model membranes. Under the following conditions, interactions between bound molecules in model lipid bilayers can recapitulate those in biological conditions. 1. Limitations of model membranes: Although model lipid bilayers cannot accurately reproduce the complexity and asymmetry of lipid distribution in biological membranes, they can mimic the interactions between individual lipid molecules with high accuracy. 2. Bimolecular interaction model: Rather than attempting to elucidate the aggregation state of many lipid molecules in the cell membrane, it is possible to clarify the structural requirements for lipid molecular interactions by focusing on the interaction between two adjacent molecules in model bilayers. 3. Common concepts of intermolecular interactions: The behavior of membrane-bound molecules can be predicted using common concepts regarding the interaction modes exhibited by structural moieties characteristic of lipids. For example, the head and tail of membrane phospholipids are governed by hydrophilic and hydrophobic interactions, respectively, and their interactions can be deduced by formulating the strength of each physicochemical interaction. Thus, it is possible to describe the interaction mode between two molecules as a whole. Extending these common concepts to natural products makes it possible to predict to some extent the interactions of membrane-bound natural products with surrounding lipids; e.g., using model bilayers, the interaction of AM3 with sterol can be accessed by ²H NMR (see Section 4.3).

3. Methods for elucidating the functional structures of natural products upon interacting with membranes

3.1. NMR observation of conformational exchange in natural products.

Small biomolecules often exert their biological activity by binding to membrane proteins. Hydrophilic compounds such as neurotransmitters bind directly to membrane proteins from the aqueous medium, whereas hydrophobic compounds such as lipid mediators bind to membrane proteins via partitioning to the lipid bilayer portion of the cell membrane. A number of natural products that bind to ion channels, typical transmembrane proteins, have long been studied, especially those that are important as marine biotoxins. These include hydrophilic compounds such as tetrodotoxin and saxitoxin and hydrophobic compounds such as ladder-shaped polyether compounds (Fig. 2). The latter are thought to first partition to lipid bilayers and then associate with ion channels, where the first step is also important for the expression of biological activity.^23)–26) During this process, natural products with strong biological activities are thought to take on multiple functional structures by changing their conformation to reach their action targets efficiently as discussed above. Based on this hypothesis, we will consider the functional structure of ladder-shaped polyethers that bind to voltage-gated ion channels. The effects of their structural changes on biological activity are still at the hypothesis stage, and further proof is awaited.

Fig. 2.

Structures of ladder-shaped polyether compounds with (A) 9-membered ether ring, brevetoxin A,³²⁾ ciguatoxin (CTX1B),³³⁾ brevisulcatic acid-4,³⁴⁾ and gambieric acid A,³⁶⁾ and (B) 7/8-ether rings, brevetoxin B,³⁷⁾ yessotoxin,³⁸⁾ gambierol,³⁹⁾ and gambierone.⁴⁰⁾

There are many examples of structural transformations that are accompanied by the broadening of signals in NMR, such as atropisomerism in diphenyl derivatives, C–N bond geometric isomerism in tri-substituted amides, and conformational isomerism in medium-sized rings. Among these, here we focus on the conformational exchanges that occur in the cyclic structure of ladder-shaped polyethers; conformational isomerism of medium-sized rings present in natural products often poses problems because it prevents the observation of NMR signals. Ladder-shaped polyether compounds are unique metabolites of dinoflagellates that have strong toxicity and specific biological activity.^27),28) The entire molecule is largely composed of rigid six-membered ether rings (and some seven-membered rings). Some of these compounds contain a nine-membered ether ring, oxonene, near the center of the molecule, which undergoes slow conformational change (Fig. 2A).²⁹⁾ Even when no nine-membered ring is present, there is always a seven- or eight-membered ring near the center of the molecule, but in most cases, this does not affect the NMR signal (Fig. 2B). This ether ring functions as a hinge, allowing the overall shape of the molecule to change dramatically through conformational alteration as discussed later.

As the causative toxins of massive fish kills and human poisoning, brevetoxins and ciguatoxins (CTXs) are important research topics. These toxins are known to activate voltage-gated ion channels,^30),31) so the structure-activity relationship has attracted much attention to gain a better understanding of the gating and voltage-sensing mechanisms of the channels. One of the problems with structure determination of these natural products based on NMR is the unobserved signals. A typical example that we met is the signal broadening of a 9-membered ether ring, such as the E-ring in brevetoxin-A,³²⁾ the F-ring in CTXs,³³⁾ and the E-ring in brevisulcatic acids^34),35) (Fig. 2). These rings are supposed to be important for the biological activities of the ladder-shaped polyethers as discussed later.

We discuss a case of brevisulcatic acids, which have the Na⁺ channel-activating activity similar to brevetoxins. Based on detailed analysis of 2D NMR spectra measured at room temperature, the partial structures of brevisulcatic acid-4 (BSX4) were elucidated such as ether rings A, B, C, H, I and J and the β-hydroxy-γ-methylene valeric acid moiety in the side chain.³⁴⁾ However, considering the molecular formula, 18 carbons and 26 protons (C-16 to C-31) were not observed due to extremely broad or missing NMR signals corresponding to H-16 to Me-48 on the rings E to G in the middle part of the molecule. This phenomenon was also observed in brevetoxin A³²⁾ and CTXs.³³⁾ In order to solve this missing signal problem, which occurred in compounds with the 9-membered ring, NMR spectra were measured for brevetoxin A at high temperatures and for CTXs at low temperatures (−20 °C). NMR spectra of BSX4 were remeasured at low temperature (−20 °C) to avoid degradation of the compound,³⁴⁾ resulting in appearance of new signals and sharpening of broad peaks in ¹³C NMR spectra (Fig. 3). The conformational alteration slowed down at −20 °C, which resulted in observations of two sets of ¹H and ¹³C NMR signals corresponding to the C10 to C33 portion of BSX4. These signals imply that BSX4 has multiple stable conformers at low temperatures analogous to brevetoxin A. On the other hand, the same NMR experiments suggest that the conformation of CTX converged into a single conformer at low temperature.

Fig. 3.

¹³C NMR spectra of BSX4 in pyridine-d₅ (a) at −20 °C and (b) at 27 °C.^{34) 13}C signals of the oxonine ring of BSX4 at 20 (C47, C47′), 33 (C24), and 90 (C23, C16) ppm broaden at room temperature (b), which sharpened at low temperature (a).

A model study was also conducted to examine in detail the slow structural changes of the 9-membered ether ring containing a double bond. Inoue et al. reported the conformational behavior of a synthetic 6/9/6 fused ring system (Fig. 4).⁴¹⁾ As observed in CTX1B, the NMR signals of the 9-membered ether ring broadened at room temperature even when attached to rigid 6-membered ether rings. As the temperature lowered, these signals sharpened and separated into two sets of peaks, indicating that two alternating conformers coexisted in the 6/9/6 fused ether ring system (Fig. 4).⁴¹⁾ Potent antifungal compounds, gambieric acids, also have the 9-membered ether ring in the middle of the molecules (Fig. 2). However, their NMR signals around ring F were observed clearly at room temperature.³⁶⁾ This difference may be due to the methyl group at C-30 interfering with the perturbation of the 9-membered ether ring. Therefore, substituents on the 9-membered ether ring greatly affect the conformation of ladder-shaped polyethers.

Fig. 4.

(Color online) The structure of the 6/9/6 model system and probable two conformers (left).⁴¹⁾ One of the possible transition-states of ciguatoxin-1B (A) that appeared during the conformational alteration in the middle of the molecule at 9-membered ring F and the grand-state conformer (B) (right). Images A and B were cited and modified from Ref. 29. (Reproduced with permission from Ref. 29. Copyright 1993 American Chemical Society.)

Ladder-shaped polyether compounds have a particular affinity for the α-helix domains of transmembrane proteins.^42),43) In the hydrophobic membrane interior, the relatively polar ether bonds and adjacent axial hydrogen are thought to be responsible for their interaction with the α-helix domains of proteins, in which the sequence of six- and seven-membered rings affect the interaction strength.^44),45) Therefore, for these compounds to reach their targets of action in vivo, they must pass through hydrophobic (and also hydrophilic) environments, where specific conformational changes of the medium ring are thought to contribute to their delivery during this process. In particular, synthetic ladder-like polyether compounds composed of rigid six-membered rings⁴⁶⁾ are known to be hard to dissolve in aqueous systems and organic solvents (personal communication from Prof. Tohru Oishi). Therefore, the medium rings near the middle of the molecule may increase solubility in the digestive and circulatory systems, as well as in the cytoplasm.

3.2. Methods for examining interaction between membrane lipids and natural products.

Many excellent studies have been reported on the functional structures of natural products and drugs that bind to membrane proteins, such as drug-receptor protein complexes. These studies have not only advanced drug development against cancers and infections but also deepened our understanding of cellular signal transduction. In this section, we will focus on natural products that bind to lipid membranes. This is because, whether the target is a protein or a lipid bilayer, this step is crucial for hydrophobic compounds to exert their activity. Therefore, we would like to mainly discuss the functional structures of natural products upon interacting with membrane lipids.

Because proteins often have a fixed shape (higher-order structure), this static structure is often a clue to inferring the function of the protein. However, in the case of natural products with flexible structures, it is not possible to directly elucidate their functional structure (active conformation) from their structure in the conventional meaning. For example, flexible molecules with long alkyl chains, such as 2-arachidonoylglycerol (2-AG), a binding molecule for opioid receptors (a type of G-protein coupled receptor), and diacylglycerol, which activates protein kinase C, are known as lipid mediators. Even in these molecules, one conformation should predominate when binding to the receptor. In other words, even if the structure is flexible, the possible conformers are often limited to a certain extent. The aforementioned study of fatty acids also revealed that alkyl chains take a highly restricted conformation in the pocket of the binding protein (fatty acid binding protein 3).^4),5) On the other hand, when lipid mediators are not bound to receptor proteins and are distributed in the lipid bilayer of biological membranes, they do not adopt a specific conformation and are thought to be highly flexible. Microdomains such as lipid rafts have been shown to play a central role in signal transduction in biological membranes, and the importance of the local distribution of lipid mediators and receptor proteins in signal transduction has been recognized. Therefore, the behavior of lipid mediators in lipid bilayers and their affinity with membrane lipids have become important research topics, but our understanding of the functional structures of lipidic molecules remains limited. The main problems that have hindered research into biomembranes can be summarized as two points: 1) the difficulty of reproducing the functional structures of natural products using model systems, particularly when the occurrence rate of the functional structure is scarce or when multiple structures occur in model membranes; and 2) the difficulties in elucidating the dynamic structural features of flexible molecules, such as the difference in their lateral and longitudinal mobilities, and how to measure and define these features. As examples of our research aimed at solving these problems, we would like to introduce our studies on amphoteric B and AM3 as membrane-binding natural products; we have included a separate section on AM3 to introduce our approach to these problems.

Amphotericin B (AmB, Fig. 5) is an important antibiotic for the treatment of severe deep-seated fungal infections,^47),48) and its mechanism of action is thought to be primarily via the formation of ion-permeable channels in fungal cell membranes.⁴⁹⁾ AmB recognizes ergosterol (Erg) present in the fungal cell membrane and forms pores through which ions can pass. The mixed complex of AmB-Erg formed in model membranes shows ion-channel activity. Yet, it was unclear whether this complex represents an ion channel formed in the fungal cell membrane responsible for antifungal activity; the above-mentioned problem 1 was a major obstacle. To address this issue, solid-state NMR, circular dichroism (CD), and MD simulations were used. First, it was necessary to investigate how the difference in AmB concentration between the cell membranes and model membranes affects the channel structure. The concentration of AmB that exerts antifungal activity on the cell membrane is thought to be less than 1:1000 (molar ratio to lipid) while it is necessary to incorporate AmB in model membranes at more than 1:20 to elucidate the channel structure with solid-state NMR. CD spectroscopy is capable of evaluating the intermolecular interaction of AmB at low concentrations; the heptaene moiety of AmB exhibits characteristic ultraviolet-visible spectrum absorption bands at 280–330 nm, and their intensities and wavelength in the CD spectrum sensitively reflect the association state of AmB.^50),51) Thus, the complex formation of AmB in the Erg membrane was investigated by changing AmB content. As a result, the CD spectrum was essentially unchanged between AmB-lipid ratios of 1:1000 and 1:10, implying that the high AmB concentration in NMR conditions reflects the association state of AmB in vivo.⁵¹⁾ Solid-state ²H NMR experiments showed that when AmB and Erg were mixed with membrane lipid (palmitoyloleoylphosphatidylcholine [POPC]) both at 5 mol% of total lipids, the rapid movement of both molecules slowed down, suggesting the formation relatively large complexes in model membranes.⁵²⁾ Assuming that the model membrane can reproduce the biological membrane complex, the next step is to elucidate the channel structure formed in the model membrane. Because ²H NMR measurements further revealed that the movement of the AmB-Erg mixed complex in the membrane was almost fixed, it was considered possible to measure the distance between the isotope-labeled positions of AmB/AmB and AmB/Erg for structure elucidation of the complex by solid-state NMR. As a result, the structure of the channel assembly was successfully constructed (Fig. 5) based on the interatomic distances obtained from the rotational echo double resonance (REDOR) spectra in solid-state NMR.⁵¹⁾ The next question was whether the AmB assembly deduced from NMR actually functions as an ion channel. To this end, MD simulations were performed to reproduce the ion channel assemblies in a perturbed state (mimicking those in the cell membranes) using the NMR-derived structure as the initial configuration. The resultant perturbing channel assemblies were subjected to estimation of an ion permeation rate, indicating that an ion-channel assembly formed by seven molecules of AmB-Erg reproduced the values recorded in electrophysiological experiments. The result supported the idea that the ion channel structure constructed in the model membrane reproduces (or closely mimics) the functional structure of AmB, which may account for its antibiotic activity⁵¹⁾ (Fig. 5B); details of these studies, including the synthesis of isotope-labeled AmB and Erg, are described in a recent review.⁵³⁾

Fig. 5.

Functional structure of AmB as an ion channel complex with Erg. Seven molecules of AmB and Erg form an ion channel (B and C). The AmB and cholesterol assembly does not form a stable ion-conducting pore (D). Color codes for panels B, C, and D: cyan, yellow and pink in AmB denote the ionic headgroup, the hydrophobic heptaene face, and the hydrophilic polyol side, respectively. Blue and green denote Erg and POPC, respectively. (Reproduced with permission from Ref. 51. Copyright 2022 American Association for the Advancement of Science.)

Similarly, there are also studies that have reported the functional structures of salinomycin and erythromycin, which exert their characteristic biological activities by binding to the lipid bilayer portion of cell membranes.^54),55) In these studies, subtle conformational changes of salinomycin as an ionophore within bicelles were detected by conventional NMR methods, and the orientation and position of erythromycin within bicelles and micelles were primarily inferred from paramagnetic relaxation enhancement experiments.

4. Structural study of amphidinol 3 in solution and lipid bilayers for elucidating its functional structure

4.1. Stereochemistry of AM3.

The next story we would like to introduce is our 30 years struggle with AM3. Dinoflagellates belonging to the genus Amphidinium produce a unique class of the polyhydroxy-polyene metabolites known as amphidinols. The first member of the amphidinols, amphidinol 1 (Fig. 6), was discovered in 1991 from Amphidinium klebsii collected at Ishigaki Island, Japan.⁵⁶⁾ Amphidinol 1 exhibited significant hemolytic and antifungal activities, surpassing the potency of standard saponin and the well-known antifungal agent AmB, respectively.⁵⁶⁾ Since that discovery, over 20 amphidinols have been identified.^57)–63) These compounds share a conserved region flanked by two tetrahydropyran rings, with structural variation primarily observed in their polyhydroxy and polyene chains. AM3 (Fig. 6), which was isolated from A. klebsii in 1996,⁵⁸⁾ exhibits the strongest antifungal and hemolytic actions among amphidinols reported to date.⁶⁴⁾ Our extensive investigation into AM3, which includes past misinterpretations and subsequent corrections, provides valuable insights into the determination of the complete stereochemistry of complex acyclic natural products. Our challenge also represents a pioneering study on the active conformations, modes of molecular interaction, and mechanisms of action of membrane-associated natural products.

Fig. 6.

Planar structure of amphidinol 1 and amphidinol 3 (AM3).

The stereochemistry of AM3 was proposed in 1999⁶⁵⁾ through combined application of a J-based configuration analysis (JBCA) method,⁹⁾ chemical degradations, and a modified Mosher’s method.⁶⁶⁾ For details of the JBCA method, please refer to the original paper and the review,^9),67) but a brief overview is presented here. As depicted in Fig. 7, the dihedral angel dependence of vicinal proton-proton coupling constant, ³J_HH, is well known as the Karplus relationship. Electronegative substituents, such as hydroxy groups, generally decrease ³J_HH values by 1–2 Hz. A similar relationship is seen for long-range carbon-proton coupling constants, ²J_CH and ³J_CH (Fig. 7). Vicinal carbon-proton coupling constant ³J_CH follows dihedral angle dependence similar to ³J_HH. In the case of geminal coupling ²J_CH, when a ¹³C is bonded to an electronegative substituent such as an oxygen or a halogen (¹H-C-¹³C-X system), a well-defined relationship exists between ²J_CH values and the dihedral angels of ¹H/X. In 1,2-dioxygenated systems, such as 1,2-diols, ²J_CH values exhibit a positive shift by 1–2 Hz. The JBCA method leverages the dihedral angle dependence of these coupling constants to determine the relative configuration between adjacent chiral carbons or between chiral carbons separated by a methylene. This method was initially applied to the configuration determination of maitotoxin^68),69) and later to that of AM3.⁶⁵⁾

Fig. 7.

Dihedral angle dependence of coupling constants,^{67) 3}J_HH, ²J_CH, and ³J_CH. (a) ³J_HH values at the ¹H/¹H dihedral angles of 180° and 60°. (b) ²J_CH values at the dihedral angles between ¹³C-attached oxygen and ¹H of 180° and 60°. (c) ³J_CH values at the ¹³C/¹H dihedral angles of 180° and 60°. The figures in parentheses represent the values of 1,2-dioxygenated systems.

An example of the application of the JBCA method to AM3 is shown in Fig. 8: determination of the C20–C21 relative configuration. In the original study,⁶⁵⁾ the stereochemistry was determined without accounting for the flexibility of this bond. However, in this analysis, we reconsider the conformational exchange between two rotamers, as illustrated in Fig. 8. The spin coupling constants are expected to reflect the averaged contributions of the two rotamers, yielding values that closely match the observed coupling constants. This demonstrates the effectiveness of the JBCA method, even in systems undergoing conformational exchange.

Fig. 8.

Relative stereochemical determination of the C20–C21 of AM3 by considering alternating rotamers around the bond.

Using a combination of the JBCA method and nuclear Overhauser effect (NOE) data, the relative configuration of the regions highlighted in yellow in Fig. 9 was determined.⁶⁵⁾ To fully elucidate the stereochemistry of AM3, degradation reactions and subsequent derivatizations were carried out, as depicted in Fig. 9.⁶⁵⁾ The absolute stereochemistry at C2 was determined by chiral chromatography, where retention times of the derivatized degradation product were compared with those of authentic samples. For the stereocenters at C6, C10, C14, and C39, the periodate degradation products of AM3 were derivatized to α-methoxy-α-(trifluoromethyl)phenylacetyl (MTPA) esters and their absolute configurations were determined using the modified Mosher's method. The absolute stereochemistry at C23 was elucidated using NMR comparison of the MTPA derivative of the periodate degradation product with authentic samples. Based on these analyses, the complete stereochemistry of AM3 was proposed in 1999, as shown in Fig. 10.⁶⁵⁾

Fig. 9.

(Color online) Stereochemistry elucidation of AM3 reported in 1999.⁶⁵⁾ The relative stereochemistry of the portions highlighted in yellow was estimated using the JBCA method and NOE interpretation of intact AM3. Absolute configurations at C2, C6, C10, C14, C23, and C39 were determined by degradation of AM3 and subsequent derivatization.

Fig. 10.

(Color online) Proposed and revised stereostructures of AM3.

The stereochemistry of AM3 proposed in 1999 was subsequently examined by Oishi et al. through organic synthesis of partial structures.^70)–74) Details of the synthetic works are omitted, but the stereochemistry was revised three times,^70),73),75) as summarized in Fig. 10. Ultimately in 2019, 20 years after the proposed stereostructure was published,⁶⁵⁾ the stereochemistry of AM3 was unambiguously and completely determined through total synthesis.⁷⁶⁾ Surprisingly, the two tetrahydrofuran rings were found to have antipodal stereochemistry, indicating that these ring moieties are likely formed via distinct biosynthetic pathways.

Now, we re-examine the stereochemical assignment from 1999 in light of the revised correct stereochemistry. Misassignments occurred in three regions: the absolute configuration at C2,⁶⁵⁾ and the relative configurations of C38–C39⁷⁵⁾ and C50–C51⁷³⁾ (Fig. 10).

While the exact reason for the misassignment of the absolute stereochemistry at C2 remains unclear, it is possible that a certain degradation product of AM3 coincidentally exhibited the same retention time as the authentic samples in chiral chromatography, leading to its mistaken identification as the target degradation product containing C2. At that time, a UV detector was used instead of mass spectrometry, which may have contributed to the misinterpretation of the HPLC peaks. On the other hand, the misassignments of the relative stereochemistry of the C38–C39 and C50–C51 bonds are attributed to misinterpretation of the JBCA method. We will review this below.

The relative configuration between C38 and C39 stereocenters was determined by assuming a rotational exchange of the bond, as depicted in Fig. 11a.⁶⁵⁾ Judging from the observed ³J_CH values, this interpretation seemed reasonable. However, it is possible that the observed coupling constants were incorrect and/or insufficient for accurate stereochemical determination. We employed two methods to measure ^2,3J_CH data. The first was a 2D hetero half-filtered total correlated spectroscopy (TOCSY; heteronuclear long-range couplings [HETLOC]) experiment,⁷⁷⁾ which gives cross-peak separated by ¹J_CH in the F1 direction and displaced by ^2,3J_CH in the F2 direction within TOCSY-like ¹H-¹H 2D spectra, allowing the ^2,3J_CH values to be read from the displacement of the cross peaks along F2. However, ^2,3J_CH values important for stereochemical determination between C38 and C39 could not be reliably extracted from the HETLOC spectra due to weak signal intensity and severe overlap with other signals. As a result, we calculated the corresponding ^2,3J_CH values from the signal intensities of a heteronuclear multiple bond correlation (HMBC)-type experiment, which is the second method to measure ^2,3J_CH values.⁶⁵⁾ Upon checking the HMBC spectra again for this review, we found that the cross-peak intensities corresponding to ³J (C37, H39) and ³J (C40, H38) were very weak, with low signal-to-noise ratios. These weak signals likely resulted in small apparent ^2,3J_CH values, even if the true coupling constants were larger, leading to a systematic misreading of the corresponding spin coupling constants.

Fig. 11.

(a) Proposed relative stereochemistry of the C38–C39 of AM3 determined by considering alternating rotamers around the bond.⁶⁵⁾ These ³J_CH values were likely misread from the cross peaks with low signal-to-noise ratio in the HMBC-type experiment. (b) A possible rotamer equilibrium based on the revised stereochemistry.

Next, we review the application of the JBCA method to the C50–C51 bond. This region was a point of contention during the preparation of the original article in 1999. Figure 12a shows the stereochemical determination reported in the paper, which appeared reasonable based on the measured coupling constants. However, we noticed that the ²J (C51, H50) value fluctuated slightly between −2.5 and −3.5 Hz, probably depending on the mixing time used in the HETLOC experiments. Empirically, a mixing time of 30 ms seems optimal for measuring ²J_CH, because longer mixing times tend to reduce the absolute values of ²J_CH. In the original article,⁶⁵⁾ we adopted −2.5 Hz as ²J (C51, H50) (Fig. 12a), but had we employed −3.5 Hz, the interpretation would be like that shown in Fig. 12b, which aligns with the correct stereochemistry. Additionally, the revised rotamer of the C50–C51 bond can account for the characteristic long-range NOEs observed in AM3 (Fig. 12c). Although the JBCA method is effective even in systems undergoing conformational exchange, the potential for misassignment of stereochemistry remains high in such cases. Careful interpretation of spin coupling constants is essential, and as many coupling constants as possible should be obtained, particularly in systems with conformational exchange.

Fig. 12.

(Color online) (a) Relative stereochemistry of C50–C51 of AM3 proposed in the original article.⁶⁵⁾ (b) A reconsidered configuration assignment for the C50–C51 bond by using −3.5 Hz as ²J (C51, H50). (c) Characteristic long-range NOEs support the rotamer in (b).

4.2. Membrane permeabilizing activity of AM3.

Along with the isolation and planar structure determination of AM3, its sterol-dependent membrane activity was also identified,⁵⁸⁾ which was considered to be related with its hemolytic and antifungal activities. AM3 induces the leakage of fluorescent dye calcein encapsulated within liposomes only when liposomes contain cholesterol or Erg.^58),78),79) The effect of sterols is striking, where within sterol-free membranes, AM3 exhibited minimal membrane permeabilizing activity. This enhanced permeabilization was attributed to pore formation in the membrane, rather than membrane disruption through surfactant action.⁸⁰⁾ The size of pore formed by AM3 was estimated to be between 2.0 and 2.9 nm using osmotic protection experiments with erythrocytes.⁸⁰⁾

It is now well established that most antimicrobial peptides, including alamethicin and melittin, form pores according to two primary models: the barrel-stave and toroidal pore models (Fig. 13). In the barrel-stave model, peptide helices insert into the membrane and aggregate to form a cylindrical superstructure with a hydrophilic lumen.^81),82) This model applies not only to antimicrobial peptides but also natural products such as AmB,⁵¹⁾ as shown in Fig. 5. In contrast, the toroidal model involves peptides that are associated with the lipid headgroups, causing the lipid monolayer to bend continuously from the outer to inner leaflets, forming a toroidal hole. Thus, the toroidal pore lumens are composed of both the peptides and the lipid headgroups.

Fig. 13.

(Color online) Schematic presentation of barrel-stave (left) and toroidal (right) pores.

We have been investigating AM3 pores with these two models in mind but have encountered conflicting results. Initially, we hypothesized that membrane permeabilization by AM3 likely follows a toroidal pore model, based on the finding that its activity is less affected by variations in membrane thickness.⁸¹⁾ This is because, in principle, toroidal pores are less sensitive to membrane thickness, whereas the activity of barrel-stave pore formers typically decreases in thicker membranes due to their inability to penetrate these membranes. Later, however, solid-state ³¹P NMR measurements of phospholipid membranes containing AM3 suggested that the barrel-stave model is more plausible than the toroidal model.⁸³⁾ This conclusion stemmed from the observation that ³¹P NMR spectra were not significantly changed even at high AM3 molar ratios (Fig. 14). This is inconsistent with the expectation of the toroidal pore formation, because toroidal pores should produce isotropic signals in solid-state ³¹P NMR spectra due to the rapid reorientation of the phospholipid headgroups at the toroidal curves.⁸⁴⁾

Fig. 14.

³¹P NMR spectra of POPC-cholesterol bilayers. AM3:cholesterol:POPC molar ratios were 0:1:18 (a), 1:1:18 (b), and 3:1:16 (c). The curved headgroups in toroidal pores should give an isotropic ³¹P NMR signal at around 0 ppm. (Reproduced with permission from Ref. 83. Copyright 2014 American Chemical Society.)

Recently we demonstrated through channel recording and atomic force microscopy that AM3 can form different types of pores in a concentration-dependent manner.⁸⁵⁾ At a lower concentration (20 nM), AM3 exhibited single-channel activity characterized by discrete open-closed transitions. The channel conductance was approximately 1.7 nS, allowing us to estimate the channel diameter to be around 0.8 nm without ion selectivity. Pores with defined conductance levels are often attributed to barrel-stave pore formation,^86),87) whereas rapid and undefined current fluctuations are typically attributed to toroidal pores.^81),86) Accordingly, our observation suggested that AM3 forms barrel-stave pores at this concentration. Conversely, at a higher AM3 concentration (2.0 µM or more), we detected giant-conductance single channels with conductance of 20–50 nS, which allowed us to estimate the pore diameter ranging from 2.6 to 4.0 nm. This is roughly consistent with the aforementioned pore size of AM3 on the erythrocyte membrane. Additionally, atomic force microscopy observations at high AM3 concentrations revealed that although the specific type of pore remains chaotic, AM3 forms domain-like aggregates where large pores may be generated. In any case, we discovered a concentration-dependent morphological change in AM3 pore structure, transitioning from typical barrel-stave pores to large channels within domain-like aggregates. This concentration-dependent variation in pore size can account for the discrepancy between hemolytic activity and calcein leakage⁷⁸⁾: AM3 exhibits hemolytic activity at nanomolar concentration, where it forms non-selective ion channels with diameters less than 1 nm, whereas calcein leakage occurs at micromolar concentrations, where large channels with diameters of a few nm are formed.

4.3. Sterol interaction of AM3.

As described above, AM3 exhibits sterol-dependent pore-forming activity on membranes. However, this may not necessarily indicate a direct molecular interaction between AM3 and sterol in membranes. It is possible that sterols alter the physicochemical properties of membranes, such as thickness, order, phase state and so on, which may indirectly promote AM3 binding and pore-formation. In fact, a similar discussion had arisen regarding the Erg-dependent channel formation of AmB.^52),88) Thus, we aimed to prove direct interaction of AM3 with sterol in membranes using solid-state ²H NMR and short-range energy transfer experiments.

Solid-state ²H NMR signals of deuterated sterol reflect their motional property in the membrane. When deuterated sterols undergo rapid axial rotation in the membrane, solid-state ²H NMR spectra exhibit narrow quadrupole splitting. If the rotation of the sterol becomes slowed due to interaction with a sterol-binder, its rotational time scale should be comparable to the ²H NMR time scale (microsecond), resulting in signal broadening. Such spectral change in ²H NMR was used to detect the interaction between deuterated sterols and AM3 in the membrane.⁸³⁾ This technique has successfully revealed direct interactions for AmB/Erg⁵²⁾ and theonellamide A/sterols in membranes.⁸⁹⁾

The quadrupole splitting patterns indicated that both 3d-cholesterol and 3d-ergosterol in lipid bilayers undergo fast rotational motion in the absence of AM3, whereas the doublet intensities were significantly reduced in the presence of AM3 (Fig. 15), showing that rotational motions of sterol become sluggish. In contrast, 3d-epicholesterol, which is an epimer of cholesterol at the 3-OH group, showed no significant change in doublet intensity in the presence or absence of AM3 (Fig. 15). These spectral features clearly indicate that the fast rotation of cholesterol and Erg is significantly attenuated due to their direct molecular interaction with AM3, whereas epicholesterol has a weaker interaction with AM3.⁸³⁾ This observation aligns with the fact that AM3 permeabilizes membranes containing cholesterol or ergosterol, but not those containing epicholesterol.⁸³⁾ These ²H NMR results explicitly demonstrate the direct molecular interaction between AM3 and 3β-hydroxysterols in membranes (Fig. 15) and suggest that AM3 recognizes the sterol 3-OH group, including its stereochemistry.

Fig. 15.

(Color online) ²H NMR spectra of 3d-sterols incorporated into lipid bilayers. (a) The membranes contained 3d-cholesterol but AM3, (b) both 3d-cholesterol and AM3, (c) 3d-ergosterol but AM3, (d) both 3d-ergosterol and AM3, (e) 3d-epicholesterol but AM3, and (f) both 3d-epicholesterol and AM3. Isotropic signals at 0 ppm are mostly due to residual deuterium water. (g) Schematic representation of the interaction between AM3 and sterols in bilayers. (Reproduced with permission from Ref. 83. Copyright 2014 American Chemical Society.)

Next, we aimed to obtain distance information between AM3 and sterol in the membrane. To measure intermolecular distances between AmB and Erg in the membrane, we utilized solid-state NMR, as described above.^51),90) However, the larger molecular size of AM3 compared with AmB presents significant challenges in synthesizing site-selectively ¹⁹F-labeled AM3. Consequently, we opted to leverage FRET. Although FRET is typically used to measure distances between fluorophore pairs on the nanometer scale, the standard FRET distance range is too long to elucidate detailed molecular interactions between small molecules. In addition, fluorescent labeling of AM3 with relatively large fluorophores is impractical. Thus, we conceived the idea of short-range FRET (Fig. 16).⁹¹⁾

Fig. 16.

(Color online) Chemical structures of cholesterol and cholestatrienol (CTL) and schematic presentation of short-range FRET between AM3 and CTL (top). (a) FRET between AM3 and CTL in isopropanol and (b) in DOPC bilayers, and (c) that between AM3 and acetyl-CTL in DOPC. AM3 was excited at 270 nm. The content of CTL or acetyl-CTL in DOPC was 10 mol%. Red and black curves denote the presence and absence of AM3, respectively. AM3, CTL, and acetyl-CTL concentrations were 9.57 µM. Acetyl-CTL, which lost its interaction with AM3, was used as a negative control. (Reproduced with permission from Ref. 91. Copyright 2022 The Chemical Society of Japan.)

To implement this concept, we first confirmed that the triene part of AM3 exhibits weak fluorescence, allowing us to use AM3 itself as a fluorophore without fluorescent labeling. We then examined whether AM3 and cholestatrienol (CTL), a fluorescent cholesterol with a conjugated triene moiety, can function as a FRET donor and acceptor, respectively, due to the overlap of the emission spectrum of AM3 with the excitation spectrum of CTL. Given the weak fluorescence quantum yield of AM3, FRET between AM3 triene and CTL was calculated to extend to a maximum distance of 1 nm, validating the concept of short-range FRET (Fig. 16). The addition of AM3 to CTL-containing liposomes increased the emission of CTL upon excitation of AM3, which demonstrated the occurrence of short-range FRET between the AM3 polyene and the CTL rings, confirming their close vicinity in the membrane. Notably, short-range FRET was not observed in the organic solvent isopropanol (Fig. 16), suggesting that the membrane environment is necessary for the molecular interaction. In addition, titration experiments using short-range FRET revealed that AM3 and CTL form a 1:1 complex in the membrane.

In summary, ²H NMR measurements indicated that AM3 recognizes the sterol OH group and its stereochemistry, and short-range FRET experiments demonstrated that the triene moiety of AM3 is in close proximity to the ring backbone of sterols. In the next section, we take account of these findings and further examine the interaction of AM3 with sterols based on its conformation in the membrane.

4.4. Conformation of sterol and membrane-bound AM3.

To gain insight into the molecular recognition of AM3 with sterol, it is essential to understand the conformation of AM3 in membranous environments. Prior to this, we consider the conformation of AM3 in the organic solvent. The JBCA method, applied to AM3 dissolved in methanol-pyridine, determined the rotational conformation of each bond for configuration assignments, thereby allowing for the estimation of conformation of AM3.⁶⁵⁾ As a result, the C20–C54 portion of AM3 adopted a hairpin-like turn structure at the two tetrahydropyran rings.^65),92) Although the JBCA method misassigned two relative configurations of AM3, the hairpin-like turn structure remains valid even when the revised stereochemistry is applied in the analysis.

We now return to the conformational analysis of AM3 as a functional structure in membranous environments. We leveraged isotropic bicelles as a membrane model.⁹³⁾ Bicelles are generally comprised of a long-chain phospholipid and a short-chain one, such as dimyristoylphosphatidylcholine (DMPC) and dihexanoylphosphatidylcholine (DHPC), respectively (Fig. 17).⁹⁴⁾ Bicelles possess properties of both bilayer vesicles and micelles, while retaining a bilayered lamellar structure, which provides an advantage over micelles. At high DMPC/DHPC ratios (>2.5), bicelles adopt magnetically aligned disk-shaped features. Conversely, at ratios of 0.5 or lower, bicelles become isotropic (also referred to as fast-tumbling) and do not align with the magnetic field. The diameter of bicelles at a DMPC/DHPC ratio of 0.5 is estimated to be 8–10 nm, which is slightly larger than typical micelles with a diameter of 5 nm, thus allowing high-resolution NMR observations of membrane-associated molecules. Therefore, we employed isotopic bicelles to elucidate the 3D structure of AM3 in a membrane environment.^93),94) Although the detected ³J_HH values of AM3 were limited in isotropic bicelles compared with in methanol-pyridine due to signal overlapping and lower signal intensity, the values were relatively consistent across both media (Fig. 17). In addition, characteristic long-range NOEs, including H32/H38 and H43/H51 (Fig. 12c), indicative of a hairpin turn at the two tetrahydrofuran rings, were also observed in both media. It is important to note that the original conformational study of AM3 in bicelles was based on the incorrect stereochemistry,⁹³⁾ because it had been conducted prior to the revision of its configuration. However, reevaluation of the conformation in bicelles using the revised stereochemistry did not change the conclusion that AM3 adopts a turn structure at the two tetrahydropyran rings.

Fig. 17.

(Color online) Schematic model of oriented and isotropic bicelles (left) and comparison of ³J_HH values of AM3 between in methanol-pyridine⁶⁵⁾ and in isotropic bicelles⁹³⁾ (right).

Notable as an example of a functional structure, the rotatable bonds C38–C39 and C50–C51, which caused the misassignment in our stereochemical analysis, have both gauche and anti conformations. These bonds are more likely to be in the gauche form in order for AM3 to take on the hairpin-turn conformation shown in Fig. 18. In other words, this coincidence supports the hypothesis that the functional structure is flexible. This is because C–C bonds that take multiple conformers under NMR conditions function as hinges that enable the conformational changes necessary for the formation of channel aggregates in the membrane.

Fig. 18.

(Top) Functional structure of AM3 interacting with sterol. The C21–C30 portion is suggested to participate in sterol recognition. In this model, the AM3-sterol interaction is stabilized by van der Waals and/or π-π interactions between the AM3 polyene (highlighted in yellow) and the steroid skeleton, as well as by multiple hydrogen bonds between AM3 and the sterol OH group. The red boxes denote that the gauche conformers about the C38–C39 and C50–C51 bonds promote the hairpin-like conformation. (Bottom) A plausible barrel-stave pore structure constructed based on the AM3-sterol complex model in a lipid bilayer. Pale blue molecules indicate sterols.

Before constructing the AM3-sterol complex model, we summarize our findings: AM3 strictly recognizes the OH group moiety of sterol; the triene portion of AM3 and the alicyclic structure of sterols are in close proximity, within 1 nm, in the membrane; and AM3 takes a hairpin-like turn structure even in a membrane environment. In addition, we recently reported the structure-activity relation using truncated derivatives of AM3, suggesting that the C21–C30 polyol moiety is involved in sterol recognition.⁹⁵⁾ To integrate these findings, we constructed a model of the AM3-sterol complex (Fig. 18), which is stabilized by van der Waals and/or π-π interactions between the AM3 polyene chain and the sterol skeleton, as well as by multiple hydrogen bonds between the sterol OH group and AM3. The formation of these multiple hydrogen bonds is probably attributed to the strict recognition of the sterol OH group by AM3. The C21–C30 chain probably acts as a lid covering the sterol molecule, as supported by the finding that the AM3 derivative lacking the C1–C30 region completely lost its ability to bind sterol, whereas the derivative retaining C10–C30 exhibited sterol-binding capability.⁹⁵⁾

It is challenging to specify the conformation of the C1–C20 moiety of AM3 due to its inherent flexibility. As noted above, because AM3 forms barrel-stave pores, particularly at low concentrations, the C1–C20 moiety of AM3 in the functional structure likely adopts an extended conformation, constituting the hydrophilic inner lining of the pore. Based on this assumption, we constructed a plausible barrel-stave pore structure formed by AM3-sterol complexes, in which the two tetrahydrofuran rings reside at the water-lipid interface, and the polyhydroxy chain of AM3 constitutes the pore lumen (Fig. 18). Given that the size of the AM3 barrel-stave pore is approximately 0.8 nm, as described above, MD simulations reproducing this size may offer a more detailed view of the pore structure.

Conclusions and future perspectives

Since the 1980s, research into elucidating the mechanisms of biological activities of natural products by target identification, particularly proteins, has become common. This research trend also influenced the natural product chemistry community, leading to the launch of ‘dynamic natural product chemistry’ in Japan.⁹⁶⁾ In recent years, coupled with advances in structural biology and molecular simulation, it has become possible to elucidate the three-dimensional structures of natural products and drugs when they exert their effects. In this review, we have discussed the importance of the flexibility in molecular shape, using the term “functional structure” to refer to the conformation and its mobility required for expression of in vivo functions, including the delivery process. Particularly, we focused on natural products that work in lipid bilayers, such as AM3, to explain how we tried to elucidate the functional structure. In this study the assignment of the entire configuration of AM3 was achieved through total synthesis, and this complete structural determination made it possible to deduce how AM3 permeabilizes lipid membranes. Based on the NMR spectra of AM3 in lipid bilayers, its functional structure was proposed when forming an ion-channel assembly with sterols in biological membranes.

As in the case of AM3, accurate analysis of configuration and conformation is required to determine the functional structures of natural products and drugs in lipid bilayers. We have devoted many pages to NMR experiments because we mainly focus on this technique in our research, but other experimental techniques such as fluorescence spectroscopy, X-ray crystal diffraction, and cryo-electron microscopy are also useful. These experimental techniques were established in the 1980s and 1990s and have been frequently improved up to the present day. For example, in the field of crystal X-ray diffraction, breakthroughs have been made, as seen in time-resolved XFEL crystallography and the crystal sponge method,⁹⁷⁾ and it is expected that breakthroughs will continue in other fields as well. Furthermore, recent advances in computer and information sciences are expected to lead to dramatic advances in spectroscopic technology for structural analysis. For example, theoretical predictions of structure and electronic state play an important role in the calculation of NMR chemical shifts and the simulation of CD spectra, techniques that are now widely used in specific fields. Recently, advancements in computational resources and software have made these techniques accessible to a broader range of researchers.

On the other hand, NMR measurements of natural products often require several milligrams of highly pure sample, and even larger amounts are sometimes required to evaluate biological activities. For this reason, when obtaining natural products from biological materials, it often takes a great deal of effort and time to collect the materials and isolate the target compound, ensuring minimal environmental impact. In addition, when it is necessary to prepare labeled compounds, chemical synthesis and derivatization of natural products are often obstacles to the progress of research. It is expected that the recent rapid progress in in-silico technologies will solve various problems in natural product and bioorganic chemistry. In this sense, there may still be room for development in organic chemistry, biophysics, and structural biology to facilitate research into functional structures.

The concept of compound structure, established in the last century, provides sufficient information for chemical synthesis and still serves as a foundational framework for chemistry today. However, in recent years, the importance of the 3D shape of molecules (or conformation), which cannot be sufficiently defined by this conventional concept, has come to be recognized. This idea has spurred active research, particularly in the structural analysis of short-lived molecules, including reaction intermediates and transition states, where the focus extends to their electronic states and orbital interactions. Biophysical and biochemical studies encompass the structural and functional dynamics of reaction intermediates and functional structures in enzyme reactions.⁹⁸⁾ It is also noteworthy that advances in techniques such as picosecond spectroscopy⁹⁹⁾ and theoretical/computational methods have led to significant advances in the structural analysis of short-lived species.¹⁰⁰⁾ Another important aspect of ‘functional structure’ is the conformation of molecules influenced by their environment, such as the conformation and orientation of a bound compound in lipid membranes. In dynamic systems such as cell membranes, where the microenvironment changes constantly, analyzing functional structures is highly challenging. Therefore, the analysis of environment-dependent structural changes is still lagging behind, mainly due to the lack of efficient experimental techniques.

It is known that bioactive natural products such as polyketides with an acyclic or macrocyclic structure have relatively flexible conformation, inferring that this contributes to the expression of biological activity.⁶⁾ Because binding affinity is affected not only by enthalpy gain estimated from the static structure of the bound form but also by entropy loss when flexibility reduced by binding and hydration are considered, it is important to maintain the flexibility the binding molecules.¹⁰¹⁾ It has also been reported that even for transmembrane proteins, which are common drug binding targets, flexibility of the protein structure increases the binding affinity.¹⁰²⁾ Furthermore, focusing on the structural motifs of natural products involved in conformational transformation, Hoffman described natural products that bind to the target as “flexible molecules with a predetermined conformation”⁷⁾ and provided guidelines for compound design, by citing studies by Still and others.¹⁰³⁾ Regarding drug discovery strategies, drugs with conformation adaptability are more likely to have favorable interactions with transport systems, including brain-blood barrier passage, for improving the delivery of drugs.¹⁰⁴⁾ Thus, innovations in analytical methods will enable high-resolution elucidation of the functional structure of natural products and drugs in vivo, providing profound insights for drug discovery and drug design.

The awarding of the 2024 Nobel Prize in Chemistry to the developers of AlphaFold2 highlights the growing impact of computational science including artificial intelligence in accurately predicting the three-dimensional structures of proteins and other molecules. Looking ahead, it is anticipated that such tools will expand to larger molecular assemblies and systems. Although simulating an entire organism remains a distant goal, breakthroughs in computational speed, such as quantum computing, could make all-atom MD simulations millions of times faster. This means that simulations currently requiring a month might be completed in seconds, enabling long-term predictions for complex systems. However, recreating a complete organism at the atomic level will require an astronomical amount of time and resources, making it likely infeasible to fully capture the essence of life. Currently, limitations in computational resources and scalability make it unrealistic to fully model highly complex, non-equilibrium systems such as living organisms and cells using microscopic approaches such as all-atom MD or quantum mechanical calculations. To address this challenge, it is essential to bridge the gap with empirical rules and parameterizations that connect physical laws to the principles underlying biological systems. For instance, understanding biological complexity across multiple levels requires new parameters and rules derived from a range of in silico approaches, including multiscale modeling and artificial intelligence.

In all-atom MD simulations, parameters are derived from first-principles calculations of electronic states, linking physicochemical research directly to parameter acquisition. Conversely, coarse-grain simulations are based on grouping multiple atoms into larger units called “beads”. Designing these beads, which do not directly correspond to physical objects, involves integrating results from all-atom calculations. This approach enables the modeling of large-scale systems such as cell membranes, where each bead’s movements are parameterized to reflect the underlying atomic dynamics. Extending this methodology to whole-cell modeling will require the integration of physical laws and biological phenomena. This involves defining key parameters and systematically calculating their values across multiple hierarchical levels of varying complexity and scales. Statistical thermodynamics and condensed matter physics play crucial roles in these studies, allowing researchers to model large molecular assemblies in biological systems without delving into the fine details of chemical structures. This approach has significantly advanced our understanding of biological phenomena. However, in order to define the parameters in these physical fields, it is still very important to accurately elucidate the functional structure of biomolecules. Functional structures of natural products, in particular, are likely to shed light on biological functions through their specific activities. In cell membranes, the functional structure of the lipids that surround membrane proteins and their complexes can be elucidated using the same methods as for natural products. Substantial structural data on lipid conformation have been obtained through X-ray crystallography and cryo-electron microscopy.^105),106) Based on this information, in-silico approaches are now being applied to predict lipid functional structures in interaction with membrane proteins.¹⁰⁷⁾ These computational methods are increasingly capable of accurately simulating biological events in cell membranes at scales achievable with current techniques, such as coarse-grained MD simulations. The systematic accumulation and organization of such fundamental data will drive advancements in science and technology, enabling broad applications in fields such as drug delivery assessment, rational drug design, and the development of artificial cell membranes.

Acknowledgements

We are grateful to Prof. Takeshi Yasumoto for providing us with an opportunity to contribute a review article to this Journal. It should be noted that thanks to the efforts of Prof. Yasumoto, many natural products have been isolated and their structures were identified, including amphidinol, CTX, yessotoxin, gambieric acid, gambierol, and maitotoxin. The research topics in this article were largely conducted in Tohoku University, the University of Tokyo, Osaka University, and Kyushu University. We would like to thank all of the members equally in these institutions. Among those, we are particularly grateful to the previous members of these research groups including Drs. Keiichi Konoki, Masanao Kinoshita, Hiroshi Tsuchikawa, Yuichi Umegawa, and Tohru Oishi; and to the late Prof. Kazuo Tachibana. For research on phospholipid bilayers, we are also grateful to Prof. J. Peter Slotte, Åbo Akademi University and Prof. Wataru Shinoda, Okayama University for productive collaborations.

Notes

Edited by Keisuke SUZUKI, M.J.A.

Correspondence should be addressed to: M. Murata, Institute for Protein Research, Osaka University, 3-2 Yamadaoka, Suita, Osaka 565-0871, Japan (e-mail: [e-mail]); N. Matsumori, Department of Chemistry, Graduate School of Science, Kyushu University, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan (e-mail: [e-mail]).

References

• 1)  Nakanishi, K. (1999) A brief history of natural products chemistry. In Comprehensive Natural Products Chemistry (eds. Barton, D., Nakanishi, K. and Meth-Cohn, O.). Pergamon Press, Oxford, pp. 1–31.
• 2)   Daina,  A.,  Michielin,  O. and  Zoete,  V. (2017) SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci. Rep. 77, 42717.
• 3)   Bocci,  G.,  Carosati,  E.,  Vayer,  P.,  Arrault,  A.,  Lozano,  S. and  Cruciani,  G. (2017) ADME-Space: a new tool for medicinal chemists to explore ADME properties. Sci. Rep. 7, 6359.
• 4)   Matsuoka,  S.,  Sugiyama,  S.,  Matsuoka,  D.,  Hirose,  M.,  Lethu,  S.,  Ano,  H. et al. (2015) Water-mediated recognition of simple alkyl chains by heart-type fatty-acid-binding protein. Angew. Chem. Int. Ed. 54, 1508–1511.
• 5)   Sugiyama,  S.,  Matsuoka,  D.,  Hara,  T.,  Sonoyama,  M.,  Matsuoka,  S. and  Murata,  M. (2022) Experimental and theoretical investigations into the mechanism of interactions between membrane-bound fatty acids and their binding protein: A model system to investigate the behavior of lipid acyl chains in contact with proteins. Chem. Phys. Lipids 247, 105227.
• 6)   Larsen,  E.M.,  Wilson,  M.R. and  Taylor,  R.E. (2015) Conformation-activity relationships of polyketide natural products. Nat. Prod. Rep. 32, 1183–1206.
• 7)   Hoffmann,  R.W. (2000) Conformation design of open-chain compounds. Angew. Chem. Int. Ed. 39, 2054–2070.
• 8)   Foster,  M.P.,  McElroy,  C.A. and  Amero,  C.D. (2007) Solution NMR of large molecules and assemblies. Biochemistry 46, 331–340.
• 9)   Matsumori,  N.,  Kaneno,  D.,  Murata,  M.,  Nakamura,  H. and  Tachibana,  K. (1999) Stereochemical determination of acyclic structures based on carbon-proton spin-coupling constants. A method of configuration analysis for natural products. J. Org. Chem. 64, 866–876.
• 10)   Medeiros-Silva,  J.,  Jekhmane,  S.,  Paioni,  A.L.,  Gawarecka,  K.,  Baldus,  M.,  Swiezewska,  E. et al. (2018) High-resolution NMR studies of antibiotics in cellular membranes. Nat. Commun. 9, 3963.
• 11)   Latorraca,  N.R.,  Venkatakrishnan,  A.J. and  Dror,  R.O. (2017) GPCR dynamics: Structures in motion. Chem. Rev. 117, 139–155.
• 12)   Rosenbaum,  M.I.,  Clemmensen,  L.S.,  Bredt,  D.S.,  Bettler,  B. and  Strømgaard,  K. (2020) Targeting receptor complexes: a new dimension in drug discovery. Nat. Rev. Drug Discov. 19, 884–901.
• 13)   Kamitori,  S. and  Takusagawa,  F. (1994) Multiple binding modes of anticancer drug actinomycin D: X-ray, molecular modeling, and spectroscopic studies of d(GAAGCTTC) 2-actinomycin D complexes and its host DNA. J. Am. Chem. Soc. 116, 4154–4165.
• 14)   Kraka,  E.,  Tuttle,  T. and  Cremer,  D. (2007) The reactivity of calicheamicin gamma(1)(I) in the minor groove of DNA: the decisive role of the environment. Chemistry 13, 9256–9269.
• 15)   Nakagawa,  Y.,  Masuda,  Y.,  Yamada,  K.,  Doi,  T.,  Takegoshi,  K.,  Igarashi,  Y. et al. (2011) Solid-state NMR spectroscopic analysis of the Ca²⁺-dependent mannose binding of pradimicin A. Angew. Chem. Int. Ed. 50, 6084–6088.
• 16)   Nakagawa,  Y. and  Ito,  Y. (2022) Mannose-binding analysis and biological application of pradimicins. Proc. Jpn. Acad. Ser. B 98, 15–29.
• 17)   Nakagawa,  Y.,  Doi,  T.,  Masuda,  Y.,  Takegoshi,  K.,  Igarashi,  Y. and  Ito,  Y. (2011) Mapping of the primary mannose binding site of pradimicin A. J. Am. Chem. Soc. 133, 17485–17493.
• 18)   Pinheiro,  M.,  Magalhães,  J. and  Reis,  S. (2019) Antibiotic interactions using liposomes as model lipid membranes. Chem. Phys. Lipids 222, 36–46.
• 19)   Miyazaki,  Y. and  Shinoda,  W. (2022) Cooperative antimicrobial action of melittin on lipid membranes: A coarse-grained molecular dynamics study. Biochim. Biophys. Acta Biomembr. 1864, 183955.
• 20)   Garza-Miyazato,  D.,  Hanashima,  S.,  Umegawa,  Y.,  Murata,  M.,  Kinoshita,  M.,  Matsumori,  N. et al. (2024) Mode of molecular interaction of triterpenoid saponin ginsenoside Rh2 with membrane lipids in liquid-disordered phases. Biochim. Biophys. Acta Biomembr. 1866, 184366.
• 21)   Bloom,  M.,  Evans,  E. and  Mouritsen,  O.G. (1991) Physical properties of the fluid lipid-bilayer component of cell membranes: a perspective. Q. Rev. Biophys. 24, 293–397.
• 22)   Mouritsen,  O.G. and  Bloom,  M. (1993) Models of lipid-protein interactions in membranes. Annu. Rev. Biophys. Biomol. Struct. 22, 145–171.
• 23)   Mori,  M.,  Oishi,  T.,  Matsuoka,  S.,  Ujihara,  S.,  Matsumori,  N.,  Murata,  M. et al. (2005) Ladder-shaped polyether compound, desulfated yessotoxin, interacts with membrane-integral α-helix peptides. Bioorg. Med. Chem. 13, 5099–5103.
• 24)   Yamada,  K.,  Kuriyama,  H.,  Hara,  T.,  Murata,  M.,  Irie,  R.,  Harntaweesup,  Y. et al. (2014) Interaction analysis of a ladder-shaped polycyclic ether and model transmembrane peptides in lipid bilayers by using Förster resonance energy transfer and polarized attenuated total reflection infrared spectroscopy. Bioorg. Med. Chem. 22, 3773–3780.
• 25)   Konoki,  K.,  Hashimoto,  M.,  Nonomura,  T.,  Sasaki,  M.,  Murata,  M. and  Tachibana,  K. (1998) Inhibition of maitotoxin-induced Ca²⁺ influx in rat glioma C6 cells by brevetoxins and synthetic fragments of maitotoxin. J. Neurochem. 70, 409–416.
• 26)   Matile,  S.,  Berova,  N. and  Nakanishi,  K. (1996) Exciton coupled circular dichroic studies of self-assembled brevetoxin-porphyrin conjugates in lipid bilayers and polar solvents. Chem. Biol. 3, 379–392.
• 27)   Yasumoto,  T. (2005) Chemistry, etiology, and food chain dynamics of marine toxins. Proc. Jpn. Acad. Ser. B 81, 43–51.
• 28)   Yasumoto,  T. (2001) The chemistry and biological function of natural marine toxins. Chem. Rec. 1, 228–242.
• 29)   Yasumoto,  T. and  Murata,  M. (1993) Marine toxins. Chem. Rev. 93, 1897–1909.
• 30)   Catterall,  W.A. and  Risk,  M. (1981) Toxin T4₄₆ from Ptychodiscus brevis (formerly Gymnodinium breve) enhances activation of voltage-sensitive sodium channels by veratridine. Mol. Pharmacol. 19, 345–348.
• 31)   Bidard,  J.N.,  Vijverberg,  H.P.,  Frelin,  C.,  Chungue,  E.,  Legrand,  A.M.,  Bagnis,  R. et al. (1984) Ciguatoxin is a novel type of Na⁺ channel toxin. J. Biol. Chem. 259, 8353–8357.
• 32)   Shimizu,  Y.,  Chou,  H.N.,  Bando,  H.,  Van Duyne,  G. and  Clardy,  J. (1986) Structure of brevetoxin A (GB-1 toxin), the most potent toxin in the Florida red tide organism Gymnodinium breve (Ptychodiscus brevis). J. Am. Chem. Soc. 108, 514–515.
• 33)   Murata,  M.,  Legrand,  A.M.,  Ishibashi,  Y.,  Fukui,  M. and  Yasumoto,  T. (1990) Structures and configurations of ciguatoxin from the moray eel Gymnothorax javanicus and its likely precursor from the dinoflagellate Gambierdiscus toxicus. J. Am. Chem. Soc. 112, 4380–4386.
• 34)   Suzuki,  R.,  Irie,  R.,  Harntaweesup,  Y.,  Tachibana,  K.,  Holland,  P.T.,  Harwood,  D.T. et al. (2014) Brevisulcatic acids, marine ladder-frame polyethers from the red tide dinoflagellate Karenia brevisulcata in New Zealand. Org. Lett. 16, 5850–5853.
• 35)   Irie,  R.,  Suzuki,  R.,  Tachibana,  K.,  Holland,  P.T.,  Harwood,  D.T.,  Shi,  F. et al. (2016) Brevisulcatic acids from a marine microalgal species implicated in a toxic event in New Zealand. Heterocycles 92, 45–54.
• 36)   Nagai,  H.,  Murata,  M.,  Torigoe,  K.,  Satake,  M. and  Yasumoto,  T. (1992) Gambieric acids, new potent antifungal substances with unprecedented polyether structures from a marine dinoflagellate Gambierdiscus toxicus. J. Org. Chem. 57, 5448–5453.
• 37)   Lin,  Y.-Y.,  Risk,  M.,  Ray,  S.M.,  Van Engen,  D.,  Clardy,  J.,  Golik,  J. et al. (1981) Isolation and structure of brevetoxin B from the “red tide” dinoflagellate Ptychodiscus brevis (Gymnodinium breve). J. Am. Chem. Soc. 103, 6773–6775.
• 38)   Satake,  M.,  Terasawa,  K.,  Kadowaki,  Y. and  Yasumoto,  T. (1996) Relative configuration of yessotoxin and isolation of two new analogs from toxic scallops. Tetrahedron Lett. 37, 5955–5958.
• 39)   Satake,  M.,  Murata,  M. and  Yasumoto,  T. (1993) Gambierol: A new toxic polyether compound isolated from the Marine Dinoflagellate Gambierdiscus toxicus. J. Am. Chem. Soc. 115, 361–362.
• 40)   Rodríguez,  I.,  Genta-Jouve,  G.,  Alfonso,  C.,  Calabro,  K.,  Alonso,  E.,  Sánchez,  J.A. et al. (2015) Gambierone, a ladder-shaped polyether from the dinoflagellate Gambierdiscus belizeanus. Org. Lett. 17, 2392–2395.
• 41)   Inoue,  M.,  Sasaki,  M. and  Tachibana,  K. (1997) Construction of fused oxonene ring and reproduction of conformational behavior shown by ring F of ciguatoxin. Tetrahedron Lett. 38, 1611–1614.
• 42)   Torikai,  K.,  Oishi,  T.,  Ujihara,  S.,  Matsumori,  N.,  Konoki,  K.,  Murata,  M. et al. (2008) Design and synthesis of ladder-shaped tetracyclic, heptacyclic, and decacyclic ethers and evaluation of the interaction with transmembrane proteins. J. Am. Chem. Soc. 130, 10217–10226.
• 43)   Alonso,  E.,  Fuwa,  H.,  Vale,  C.,  Suga,  Y.,  Goto,  T.,  Konno,  Y. et al. (2012) Design and synthesis of skeletal analogues of gambierol: Attenuation of amyloid-β and tau pathology with voltage-gated potassium channel and N-methyl-D-aspartate receptor implications. J. Am. Chem. Soc. 134, 7467–7479.
• 44)   Sasaki,  M. and  Tachibana,  K. (2007) Design and synthesis of simplified polycyclic ethers and evaluation of their interaction with an α-helical peptide as a model of target proteins. Tetrahedron Lett. 48, 3181–3186.
• 45)   Torikai,  K.,  Yari,  H.,  Mori,  M.,  Ujihara,  S.,  Matsumori,  N.,  Murata,  M. et al. (2006) Design and synthesis of an artificial ladder-shaped polyether that interacts with glycophorin A. Bioorg. Med. Chem. Lett. 16, 6355–6359.
• 46)   Kawamura,  S.,  Toriyama,  K.,  Umeno,  K.,  Yasuno,  Y.,  Satake,  M. and  Oishi,  T. (2024) Unified synthesis of NOPQ and STUV rings of brevisulcenal-F. Chem. Lett. 53, upae179.
• 47)   Donovick,  S.,  Gold,  W.,  Pagano,  J.F. and  Stout,  H.A. (1956) Amphotericins A and B, antifungal antibiotics produced by a streptomycete. I. In vitro studies. Antibiot. Annu. 3, 579–586.
• 48)   Hartsel,  S. and  Bolard,  J. (1996) Amphotericin B: new life for an old drug. Trends Pharmacol. Sci. 17, 445–449.
• 49)   Bolard,  J. (1986) How do the polyene macrolide antibiotics affect the cellular membrane properties? Biochim. Biophys. Acta 864, 257–304.
• 50)   Vertut-Croquin,  A.,  Bolard,  J.,  Chabbert,  M. and  Gary-Bobo,  C. (1983) Differences in the interaction of the polyene antibiotic amphotericin B with cholesterol- or ergosterol-containing phospholipid vesicles. A circular dichroism and permeability study. Biochemistry 22, 2939–2944.
• 51)   Umegawa,  Y.,  Yamamoto,  T.,  Dixit,  M.,  Funahashi,  K.,  Seo,  S.,  Nakagawa,  Y. et al. (2022) Amphotericin B assembles into seven-molecule ion channels: An NMR and molecular dynamics study. Sci. Adv. 8, eabo2658.
• 52)   Matsumori,  N.,  Tahara,  K.,  Yamamoto,  H.,  Morooka,  A.,  Doi,  M.,  Oishi,  T. et al. (2009) Direct interaction between amphotericin B and ergosterol in lipid bilayers as revealed by 2H NMR spectroscopy. J. Am. Chem. Soc. 131, 11855–11860.
• 53)   Umegawa,  Y.,  Tsuchikawa,  H.,  Shinoda,  W. and  Murata,  M. (2024) NMR and molecular simulation studies on the structure elucidation of the amphotericin B ion channel using ¹³C and ¹⁹F-labelling. Org. Biomol. Chem. 23, 1233–1252.
• 54)   Matsumori,  N.,  Morooka,  A. and  Murata,  M. (2007) Conformation and location of membrane-bound salinomycin-sodium complex deduced from NMR in isotropic bicelles. J. Am. Chem. Soc. 129, 14989–14995.
• 55)   Matsumori,  N.,  Morooka,  A. and  Murata,  M. (2006) Detailed description of the conformation and location of membrane-bound erythromycin A using isotropic bicelles. J. Med. Chem. 49, 3501–3508.
• 56)   Satake,  M.,  Murata,  M.,  Yasumoto,  T.,  Fujita,  T. and  Naoki,  H. (1991) Amphidinol, a polyhydroxypolyene antifungal agent with an unprecedented structure, from a marine dinoflagellate, Amphidinium klebsii. J. Am. Chem. Soc. 113, 9859–9861.
• 57)   Paul,  G.K.,  Matsumori,  N.,  Murata,  M. and  Tachibana,  K. (1995) Isolation and chemical structure of amphidinol 2, a potent hemolytic compound from marine dinoflagellate Amphidinium klebsii. Tetrahedron Lett. 36, 6279–6282.
• 58)  Paul, G.K., Matsumori, N., Konoki, K., Sasaki, M., Murata, M. and Tachibana, K. (1996) Structure of amphidinol 3 and its cholesterol-dependent membrane perturbation: a strong antifungal metabolite produced by dinoflagellate, Amphidinium klebsii. In Harmful and Toxic Algal Blooms (eds. Yasumoto, T., Oshima, Y. and Fukuyo, Y.). Intergovernmental Oceanographic Commission of UNESCO, Paris, pp. 503–506.
• 59)   Morsy,  N.,  Matsuoka,  S.,  Houdai,  T.,  Matsumori,  N.,  Adachi,  S.,  Murata,  M. et al. (2005) Isolation and structure elucidation of a new amphidinol with a truncated polyhydroxyl chain from Amphidinium klebsii. Tetrahedron 61, 8606–8610.
• 60)   Echigoya,  R.,  Rhodes,  L.,  Oshima,  Y. and  Satake,  M. (2005) The structures of five new antifungal and hemolytic amphidinol analogs from Amphidinium carterae collected in New Zealand. Harmful Algae 4, 383–389.
• 61)   Meng,  Y.,  Van Wagoner,  R.M.,  Misner,  I.,  Tomas,  C. and  Wright,  J.L.C. (2010) Structure and biosynthesis of amphidinol 17, a hemolytic compound from Amphidinium carterae. J. Nat. Prod. 73, 409–415.
• 62)   Nuzzo,  G.,  Cutignano,  A.,  Sardo,  A. and  Fontana,  A. (2014) Antifungal amphidinol 18 and its 7-sulfate derivative from the marine dinoflagellate Amphidinium carterae. J. Nat. Prod. 77, 1524–1527.
• 63)   Satake,  M.,  Cornelio,  K.,  Hanashima,  S.,  Malabed,  R.,  Murata,  M.,  Matsumori,  N. et al. (2017) Structures of the largest amphidinol homologues from the Dinoflagellate Amphidinium carterae and structure–activity relationships. J. Nat. Prod. 80, 2883–2888.
• 64)   Morsy,  N.,  Houdai,  T.,  Matsuoka,  S.,  Matsumori,  N.,  Adachi,  S.,  Oishi,  T. et al. (2006) Structures of new amphidinols with truncated polyhydroxyl chain and their membrane-permeabilizing activities. Bioorg. Med. Chem. 14, 6548–6554.
• 65)   Murata,  M.,  Matsuoka,  S.,  Matsumori,  N.,  Paul,  G.K. and  Tachibana,  K. (1999) Absolute configuration of amphidinol 3, the first complete structure determination from amphidinol homologues: Application of a new configuration analysis based on carbon-hydrogen spin-coupling constants. J. Am. Chem. Soc. 121, 870–871.
• 66)   Ohtani,  I.,  Kusumi,  T.,  Kashman,  Y. and  Kakisawa,  H. (1999) High-field FT NMR application of Mosher’s method. The absolute configurations of marine terpenoids. J. Am. Chem. Soc. 113, 4092–4096.
• 67)  Matsumori, N. and Murata, M. (2018) NMR studies on natural product–Stereochemical determination and conformational analysis in solution and in membrane. In Experimental Approaches of NMR Spectroscopy (eds. The Nuclear Magnetic Resonance Society of Japan). Springer, Singapore, pp. 383–414.
• 68)   Sasaki,  M.,  Matsumori,  N.,  Maruyama,  T.,  Nonomura,  T.,  Murata,  M.,  Tachibana,  K. et al. (1996) The complete structure of maitotoxin, Part I: Configuration of the C1-C14 side chain. Angew. Chem. Int. Ed. Engl. 35, 1672–1675.
• 69)   Nonomura,  T.,  Sasaki,  M.,  Matsumori,  N.,  Murata,  M.,  Tachibana,  K. and  Yasumoto,  T. (1996) The complete structure of maitotoxin, Part II: Configuration of the C135-C142 side chain and absolute configuration of the entire molecule. Angew. Chem. Int. Ed. Engl. 35, 1675–1678.
• 70)   Oishi,  T.,  Kanemoto,  M.,  Swasono,  R.,  Matsumori,  N. and  Murata,  M. (2008) Combinatorial synthesis of the 1,5-polyol system based on cross metathesis: Structure revision of amphidinol 3. Org. Lett. 10, 5203–5206.
• 71)   Kanemoto,  M.,  Murata,  M. and  Oishi,  T. (2009) Stereoselective synthesis of the C31-C40/C43-C52 unit of amphidinol 3. J. Org. Chem. 74, 8810–8813.
• 72)   Manabe,  Y.,  Ebine,  M.,  Matsumori,  N.,  Murata,  M. and  Oishi,  T. (2012) Confirmation of the absolute configuration at C45 of amphidinol 3. J. Nat. Prod. 75, 2003–2006.
• 73)   Ebine,  M.,  Kanemoto,  M.,  Manabe,  Y.,  Konno,  Y.,  Sakai,  K.,  Matsumori,  N. et al. (2013) Synthesis and structure revision of the C43-C67 part of amphidinol 3. Org. Lett. 15, 2846–2849.
• 74)   Tsuruda,  T.,  Ebine,  M.,  Umeda,  A. and  Oishi,  T. (2015) Stereoselective synthesis of the C1-C29 part of amphidinol 3. J. Org. Chem. 80, 859–871.
• 75)   Wakamiya,  Y.,  Ebine,  M.,  Murayama,  M.,  Omizu,  H.,  Matsumori,  N.,  Murata,  M. et al. (2018) Synthesis and stereochemical revision of the C31-C67 fragment of amphidinol 3. Angew. Chem. Int. Ed. 57, 6060–6064.
• 76)   Wakamiya,  Y.,  Ebine,  M.,  Matsumori,  N. and  Oishi,  T. (2020) Total synthesis of amphidinol 3: A general strategy for synthesizing amphidinol analogues and structure-activity relationship study. J. Am. Chem. Soc. 142, 3472–3478.
• 77)   Kurz,  M.,  Schmieder,  P. and  Kessler,  H. (1991) HETLOC, an efficient method for determining heteronuclear long-range couplings with heteronuclei in natural abundance. Angew. Chem. Int. Ed. Engl. 30, 1329–1331.
• 78)   Paul,  G.K.,  Matsumori,  N.,  Konoki,  K.,  Murata,  M. and  Tachibana,  K. (1997) Chemical structures of amphidinols 5 and 6 isolated from marine dinoflagellate Amphidinium klebsii and their cholesterol-dependent membrane disruption. J. Mar. Biotechnol. 5, 124–128.
• 79)   Morsy,  N.,  Houdai,  T.,  Konoki,  K.,  Matsumori,  N.,  Oishi,  T. and  Murata,  M. (2008) Effects of lipid constituents on membrane-permeabilizing activity of amphidinols. Bioorg. Med. Chem. 16, 3084–3090.
• 80)   Houdai,  T.,  Matsuoka,  S.,  Matsumori,  N. and  Murata,  M. (2004) Membrane-permeabilizing activities of amphidinol 3, polyene-polyhydroxy antifungal from a marine dinoflagellate. Biochim. Biophys. Acta Biomembr. 1667, 91–100.
• 81)   Yang,  L.,  Harroun,  T.A.,  Weiss,  T.M.,  Ding,  L. and  Huang,  H.W. (2001) Barrel-stave model or toroidal model? A case study on melittin pores. Biophys. J. 81, 1475–1485.
• 82)   Bobone,  S.,  Roversi,  D.,  Giordano,  L.,  De Zotti,  M.,  Formaggio,  F.,  Toniolo,  C. et al. (2012) The lipid dependence of antimicrobial peptide activity is an unreliable experimental test for different pore models. Biochemistry 51, 10124–10126.
• 83)   Espiritu,  R.,  Matsumori,  N.,  Tsuda,  M. and  Murata,  M. (2014) Direct and stereospecific interaction of amphidinol 3 with sterol in lipid bilayers. Biochemistry 53, 3287–3293.
• 84)   Sani,  M.A.,  Rajput,  S.,  Keizer,  D.W. and  Separovic,  F. (2024) NMR techniques for investigating antimicrobial peptides in model membranes and bacterial cells. Methods 224, 10–20.
• 85)   Iwamoto,  M.,  Sumino,  A.,  Shimada,  E.,  Kinoshita,  M.,  Matsumori,  N. and  Oiki,  S. (2017) Channel formation and membrane deformation via sterol-aided polymorphism of amphidinol 3. Sci. Rep. 7, 10782.
• 86)   Fennouri,  A.,  Mayer,  S.F.,  Schroeder,  T.B.H. and  Mayer,  M. (2017) Single channel planar lipid bilayer recordings of the melittin variant MelP5. Biochim. Biophys. Acta Biomembr. 1859, 2051–2057.
• 87)   Wiedman,  G.,  Fuselier,  T.,  He,  J.,  Searson,  P.C.,  Hristova,  K. and  Wimley,  W.C. (2014) Highly efficient macromolecule-sized poration of lipid bilayers by a synthetically evolved peptide. J. Am. Chem. Soc. 136, 4724–4731.
• 88)   Venegas,  B.,  González-Damián,  J.,  Celis,  H. and  Ortega-Blake,  I. (2003) Amphotericin B channels in the bacterial membrane: role of sterol and temperature. Biophys. J. 85, 2323–2332.
• 89)   Espiritu,  R.A.,  Matsumori,  N.,  Murata,  M.,  Nishimura,  S.,  Kakeya,  H.,  Matsunaga,  S. et al. (2013) Interaction between the marine sponge cyclic peptide theonellamide a and sterols in lipid bilayers as viewed by surface plasmon resonance and solid-state 2H nuclear magnetic resonance. Biochemistry 52, 2410–2418.
• 90)   Umegawa,  Y.,  Nakagawa,  Y.,  Tahara,  K.,  Tsuchikawa,  H.,  Matsumori,  N.,  Oishi,  T. et al. (2012) Head-to-tail interaction between amphotericin B and ergosterol occurs in hydrated phospholipid membrane. Biochemistry 51, 83–89.
• 91)   Hieda,  M.,  Tsujimura,  K.,  Kinoshita,  M. and  Matsumori,  N. (2022) Formation of a tight complex between amphidinol 3 and sterols in lipid bilayers revealed by short-range energy transfer. Bull. Chem. Soc. Jpn. 95, 1753–1759.
• 92)   Houdai,  T.,  Matsuoka,  S.,  Morsy,  N.,  Matsumori,  N.,  Satake,  M. and  Murata,  M. (2005) Hairpin conformation of amphidinols possibly accounting for potent membrane permeabilizing activities. Tetrahedron 61, 2795–2802.
• 93)   Houdai,  T.,  Matsumori,  N. and  Murata,  M. (2008) Structure of membrane-bound amphidinol 3 in isotropic small bicelles. Org. Lett. 10, 4191–4194.
• 94)   Matsumori,  N. and  Murata,  M. (2010) 3D structures of membrane-associated small molecules as determined in isotropic bicelles. Nat. Prod. Rep. 27, 1480–1492.
• 95)   Matsumori,  N.,  Hieda,  M.,  Morito,  M.,  Wakamiya,  Y. and  Oishi,  T. (2024) Truncated derivatives of amphidinol 3 reveal the functional role of polyol chain in sterol-recognition and pore formation. Bioorg. Med. Chem. Lett. 98, 129594.
• 96)  Ogura, T. (1997) Dynamic Aspects of Natural Products Chemistry. CRC Press, Tokyo.
• 97)   Inokuma,  Y.,  Yoshioka,  S.,  Ariyoshi,  J.,  Arai,  T.,  Hitora,  Y.,  Takada,  K. et al. (2013) X-ray analysis on the nanogram to microgram scale using porous complexes. Nature 495, 461–466.
• 98)   Schramm,  V.L. (2011) Enzymatic transition states, transition-state analogs, dynamics, thermodynamics, and lifetimes. Annu. Rev. Biochem. 80, 703–732.
• 99)   Maiuri,  M.,  Garavelli,  M. and  Cerullo,  G. (2019) Ultrafast spectroscopy: State of the art and open challenges. J. Am. Chem. Soc. 142, 3–15.
• 100)   Maeda,  S.,  Harabuchi,  Y.,  Ono,  Y.,  Taketsugu,  T. and  Morokuma,  K. (2015) Intrinsic reaction coordinate: Calculation, bifurcation, and automated search. Int. J. Quantum Chem. 115, 258–269.
• 101)   Chodera,  J.D. and  Mobley,  D.L. (2013) Entropy-enthalpy compensation: role and ramifications in biomolecular ligand recognition and design. Annu. Rev. Biophys. 42, 121–142.
• 102)   Amaral,  M.,  Kokh,  D.B.,  Bomke,  J.,  Wegener,  A.,  Buchstaller,  H.P.,  Eggenweiler,  H.M. et al. (2017) Protein conformational flexibility modulates kinetics and thermodynamics of drug binding. Nat. Commun. 8, 2276.
• 103)   Smith,  P.W. and  Still,  W.C. (1988) The effect of substitution and stereochemistry on ion binding in the polyether ionophore monensin. J. Am. Chem. Soc. 110, 7917–7919.
• 104)   Wu,  D.,  Chen,  Q.,  Chen,  X.,  Han,  F.,  Chen,  Z. and  Wang,  Y. (2023) The blood-brain barrier: Structure, regulation and drug delivery. Signal Transduct. Target. Ther. 8, 217.
• 105)   Biou,  V. (2023) Lipid-membrane protein interaction visualised by cryo-EM: a review. Biochim. Biophys. Acta Biomembr. 1865, 184068.
• 106)   Sych,  T.,  Levental,  K.R. and  Sezgin,  E. (2023) Lipid-protein interactions in plasma membrane organization and function. Annu. Rev. Biophys. 51, 135–156.
• 107)   Renard,  K. and  Byrne,  B. (2021) Insights into the role of membrane lipids in the structure, function and regulation of integral membrane proteins. Int. J. Mol. Sci. 22, 9026.

Non-standard abbreviation list

AM3

amphidinol 3

AmB

amphotericin B

BSX4

brevisulcatic acid-4

CD

circular dichroism

CTL

cholestatrienol

CTX

ciguatoxin

Erg

ergosterol

DHPC

dihexanoylphosphatidylcholine

DMPC

dimyristoylphosphatidylcholine

DOPC

dioleoylphosphatidylcholine

FRET

Förster resonance energy transfer

HETLOC

heteronuclear long-range couplings

HMBC

heteronuclear multiple bond correlation

JBCA

J-based configuration analysis

MD

molecular dynamics

MTPA

α-methoxyl-α-trifluoromethyl-phenylacetic acid

NMR

nuclear magnetic resonance

NOE

nuclear Overhauser effect

POPC

palmitoyloleoylphosphatidylcholine

REDOR

rotational echo double resonance

TOCSY

total correlated spectroscopy

Profile

Michio Murata received his Ph.D. from Tohoku University in 1986 (Prof. Takeshi Yasumoto). He worked for the Suntory Institute for Bioorganic Research (1983–1985), then held positions as Assistant Professor, Faculty of Agriculture, Tohoku University (1985–1993); Associate Professor, School of Science, the University of Tokyo (1993–1999); Professor, Graduate School of Science, Osaka University (1999–2024); and Specially Appointed Professor, Protein Research Institute, Osaka University (2024–present). He received the Award for the Encouragement of Young Scientists, the Japan Bioscience, Biotechnology and Agrochemistry Society (1991); The Chemical Society of Japan Award for Creative Work (2006); Fellow of Royal Society of Chemistry (2014); Honorary Doctorate from Åbo Akademi University (2022), and the Chemical Society of Japan Award (2022).

Masayuki Satake received his Ph.D. from Tohoku University in 1994 (Prof. Takeshi Yasumoto). He has held positions as Assistant Professor and Associate Professor, Faculty of Agriculture, Tohoku University (1994–2004); visiting research Professor, University of North Carolina Wilmington (2005–2006, Prof. J.L.C. Wright); and Associate Professor, School of Science, the University of Tokyo (2006–present).

Nobuaki Matsumori received his Ph.D. from the University of Tokyo in 1997, under the supervision of Profs. Kazuo Tachibana and Michio Murata. His dissertation focused on the configuration analysis of complicated natural products. Following this, he pursued postdoctoral research in chemical biology under Profs. Sueharu Horinouchi and Minoru Yoshida, also at the University of Tokyo (1997–1999). In 1999, he joined Osaka University as an Assistant Professor and later as an Associate Professor in 2010, working with Prof. M. Murata. In 2000, he spent a year as a visiting scientist at MIT. Since 2014, he has been a Professor at Kyushu University, where his current research focuses on the functional analysis of lipids.