2022 Volume 70 Issue 1 Pages 1-9
Biomembranes composed of various proteins and lipids play important roles in cellular functions, such as signal transduction and substance transport. In addition, some bioactive peptides and pathogenic proteins target membrane proteins and lipids to exert their effects. Therefore, an understanding of dynamic and complex intermolecular interactions among these membrane constituents is needed to elucidate their mechanisms. This review summarizes the major research carried out in the author’s laboratory on how lipids and their inhomogeneous distributions regulate the structures and functions of antimicrobial peptides and Alzheimer’s amyloid β-protein. Also, how to detect transmembrane helix–helix and membrane protein–protein interactions and how they are modulated by lipids are discussed.
Biomembranes play important roles in cellular functions, such as signal transduction and substance transport. The basic architecture of biomembranes is the lipid bilayer, in which various functional proteins are incorporated. Human erythrocyte membranes, one of the best-studied biomembranes and a model of mammalian plasma membranes, contain several hundred lipid species, the distribution of which is longitudinally asymmetric (Fig. 1). The outer leaflet contains mainly zwitterionic lipids, such as phosphatidylcholine (PC) and sphingomyelin (SM) with negatively charged sphingoglycolipids including gangliosides as minor components, whereas phosphatidylethanolamine (PE) and acidic phospholipids, such as phosphatidylserine (PS), phosphatidylinositol (PI), and phosphatidic acid (PA), are localized in the inner leaflet.1) Recently, the lateral distribution of lipids was also proposed to be non-uniform (Fig. 1). Sphingolipids (SM and sphingoglycolipids) and cholesterol form so-called ‘lipid raft’ microdomains, in which proteins involved in signal transduction are concentrated.2) Not only such diversity in lipid species and their inhomogeneous distributions but also physicochemical properties of lipid bilayers, e.g., membrane thickness and fluidity, also regulate the structure and function of membrane proteins.3)
The outer leaflet contains zwitterionic PC and SM with anionic gangliosides. Sphingolipids (SM and gangliosides) and cholesterol form ‘lipid raft’ microdomains with signaling proteins (not shown). In contrast, the inner leaflet is composed of zwitterionic PE and negatively charged phospholipids including PS, PI, and PA.
This review summarizes the major research carried out in the author’s laboratory on how the asymmetry and lateral heterogeneity of lipid distribution regulate the structures and functions of antimicrobial peptides and Alzheimer’s amyloid β-protein, respectively. Furthermore, methods to detect transmembrane helix–helix and membrane protein–protein interactions and how they are affected by membrane thickness and cholesterol are discussed.
Antimicrobial peptides (AMPs) are cationic and amphiphilic peptides typically composed of 15–50 amino acid residues produced by living organisms including humans as part of innate immunity and also promising candidates for antibacterial drugs.4–6) The cationicity is important for their selective interaction with bacterial cells because bacterial surfaces are more negatively charged than mammalian cell surfaces (Fig. 2). The bacterial cell membranes are rich in anionic phospholipids, mainly phosphatidylglycerol (PG) and cardiolipin (dimerized PG). Anionic peptidoglycan layers cover cell membranes. In addition, Gram-negative bacteria have outer membranes coated with negatively charged lipopolysaccharides (LPS). The majority of AMPs exert antimicrobial activity by permeabilizing membranes.7) This general mode of action endows them with the important properties: broad antimicrobial spectra and difficulty of emergence of resistant bacteria. For more details, refer to review articles by the author.8–12)
Anionic PG, cardiolipin, and zwitterionic PE mainly constitute cell membranes, which are covered by negatively charged peptidoglycan layers. In addition, Gram-negative bacteria have outer membranes coated with negatively charged LPS.
We have investigated interactions of several AMPs with lipid bilayers as a model of cell membranes as well as bacteria to elucidate their molecular mechanisms of action (Table 1). The major driving forces of membrane binding are electrostatic and hydrophobic interactions. AMPs, the hydrophobicity of which is generally low, exhibit relatively weak interaction with zwitterionic lipids and therefore normal mammalian cells. It should be noted that the hydrophobicity of amino acid residues depends on their positions along the peptide sequence with terminal positions contributing less to the overall hydrophobicity.13) AMPs selectively bind to negatively charged bilayers and bacterial cell membranes by electrostatic interaction14–18) (Fig. 3). Interestingly, however, fluorescent resonance energy transfer (FRET) experiments between the Trp residue incorporated in magainin 2 and pyrene-labeled lipids revealed that there is no specific interaction between the peptides and anionic lipids. Rather, AMPs are electrostatically concentrated above the membranes and then partitioned into the membranes.19) The membrane-bound peptides form amphiphilic secondary structures, typically α-helices, and are located at the water–membrane interface.15,20) In contrast, the α-helical magainin 2 peptide is specifically bound to the sugar group region of monosialoganglioside GM1, which is overexpressed on the surface of tumorigenic HeLa cells.19) Such specific interaction is also observed between tachyplesin I having a cyclic β-sheet structure and LPS.21)
| Peptide | Origin | Sequence |
|---|---|---|
| α-Helical | ||
| Magainin 1 | Xenopus laevis | GIGKF LHSAG KFGKA FVGEI MKS |
| Magainin 2 | GIGKF LHSAK KFGKA FVGEI MNS | |
| PGLa | GMASK AGAIA GKIAK VALKA L-amide | |
| Buforin 2 | Bufo bufo gargarizans | TRSSR AGLQF PVGRV HRLLR K |
| Cyclic-β-sheet | ||
| Tachyplesin I | Tachypleus tridentataus |  |
AMPs preferentially interact with bacterial and cancer cells over normal cells by virtue of electrostatic interaction. The mode of action is, however, different between bacterial and cancer cells. See the text for details.
Bound peptides, which are often partially oligomerized,17,22) expand the headgroup region of bilayers, inducing membrane thinning23) and positive curvature strain,24) leading to the formation of the unique toroidal pore, in which not only the hydrophilic faces of the amphiphilic secondary structures but also the headgroups of lipids constitute the pore lumen25–27) (Fig. 4). The structure markedly contrasts with the already known ‘barrel-stave’ ion channel formed by peptibols, such as alamethicin28) and hypelcin A29,30) in the presence of transmembrane voltage. Approximately 5 helices are involved in a pore formed by magainin 2 and its diameter is 2–3 nm,12,31,32) although the pore size is sensitive to subtle amino acid substitutions.33) This structure makes otherwise independent outer and inner leaflets a continuum, allowing a rapid scrambling of lipids.25,32) The pore formation rate increases with the decreasing polar angle of the amphiphilic helix.34) The lifetime of the pore is short (approx. ms) and decreases with an increasing number of positive charges.17,35) Upon its disintegration, some peptides are translocated into the inner leaflet.25,27,36) Buforin 2 (Table 1) is known to cross membranes without their permeabilization and interact with intracellular DNA/RNA.37) Its mechanism of translocation is the same as that of magainin 2: The extremely short lifetime of the pore prevents membrane permeabilization and enables effective translocation.38) The toroidal pore is also formed by the wasp venom mastoparan X39) and bee venom melittin.40,41) It should be noted that in lipid bilayers with a negative curvature strain, e.g., PS, the peptide-induced positive curvature strain is counteracted by surrounding lipids, resulting in membrane disruption due to the accumulation of large numbers of peptides.24)
AMPs bind to the membrane surface, forming amphiphilic structures at the membrane interface. Several molecules form toroidal pores with surrounding lipid molecules, inducing membrane permeabilization and lipid scrambling at the same time. Upon the disintegration of the pores, some peptides are translocated into the inner leaflet.
A combination of two AMPs occasionally exhibits synergistic activity. Magainin 2 and PGLa, both from the same origin (Table 1), form a 1 : 1 complex, in which the two helices align in parallel, and exhibit much higher antimicrobial activity but also greater cytotoxicity than the individual components alone.17,42–44) Another example is a mixture of magainin 2 and tachyplesin I, which exhibits bacteria-selective synergism, although the underlying mechanism is remains unclear.45)
For the development of anti-infective drugs, improvement of the therapeutic index is crucial. The attachment of polyethylene glycol (MW 5000) at the N-terminus of tachyplesin I can almost nullify its cytotoxicity at the expense of significant reduction in antimicrobial activity.27) The drawback can be partially overcome by the augmentation of positive charges, as exemplified in the case of magainin 2.46)
Anticancer agents are another application of AMPs because the surfaces of cancer cells are more negatively charged than normal cells47,48) (Fig. 3). AMPs that are activated at a weakly acidic pH would show augmented selectivity for cancer cells because the pH in tumor tissue (5.6–6.849) or 6.2–6.950)) is lower than that in normal tissue (approx. 7.4). For this purpose, the charge-reversal peptide HE (GIHHW LHSAH EFGEH FVHHI MNS-amide) was designed using magainin 2 as a template.51) Its charge reverses from −1.5 at pH 7.4 to +6 at pH 5.5. The peptide exerted moderate toxicity against human renal adenocarcinoma ACHN cells at pH 6.0 (EC50 approx. 100 µM), but not at least up to 100 µM at pH 7.4, and was nontoxic against human normal glomerular mesangial cells even at this low pH. To enhance tumoricidal activity, we introduced 2,3-diaminopropionic acid (Dap) residues with a pKa value of 6.3 instead of His.52) The best peptide with 8 Dap residues (GIXXW LHSAX XFGXX FVXXI ZNS-amide, X = Dap, Y = norleucine) exhibited EC50 values of approx. 5 and 60 µM against the multidrug-resistant human pancreas carcinoma cell line PANC-1 at pH 6.0 and 7.4, respectively. Furthermore, the introduction of Dap also reduced cytotoxicity against normal HEK293 cells and glomerular mesangial cells at pH 7.4 (EC50 > 100 µM).
Alzheimer’s disease (AD) is a major form of dementia, and amyloid β-protein (Aβ), most commonly composed of 40 or 42 amino acid residues, plays a central role in its pathogenesis53–55) (Fig. 5(a)). Self-assembled forms of Aβ rich in β-sheet structures (oligomers and fibrils) impair neuronal cells.56) Aβmainly originates from the neuronal cell surface amyloid precursor protein, which is sequentially cleaved by β- and γ-secretases in recycling endosomes,55) and is most likely released into the synaptic terminal.57) Thus, Aβ has a better chance to interact with neuronal membranes. Accumulating evidence has suggested that the binding of Aβ to membranes plays an pivotal role in the aggregation of Aβ.58–62)
The β-sheet is stabilized by salt bridges (open squares) and π–π interactions (gray squares).
Yanagisawa et al. identified a specific form of Aβ bound to the monosialoganglioside GM1, a sphingoglycolipid abundant in neuronal membranes (Fig. 5(b)), in brains exhibiting the early pathological changes associated with AD, and also suggested that the GM1-bound form of Aβ may serve as a seed for the formation of Aβ aggregates.63) Inspired by this discovery, we have systematically investigated Aβ–membrane interaction using lipid bilayers as well as neuronal cells. The membrane-mediated amyloidogenesis mechanism also explains the reason why aged rodents rarely develop the characteristic lesions of AD.64) Rodent Aβ forms much less toxic fibrils on neuronal cells than its human counterpart. Our major findings are illustrated in Fig. 6, and also refer to reviews by the author for more details.58–61,65,66) Aβ does not interact with membranes without GM1 or with uniformly distributed GM1, corresponding to the healthy condition. An increase in the cholesterol level triggers the formation of GM1 clusters, which are recognized by Aβ.67–72) The protein initially forms an α-helical form, which is converted to a β-sheet rich oligomer composed of 10–20 Aβ molecules with an increase in the protein density in the membrane.73,74) Interestingly, this oligomer is nontoxic, in marked contrast to toxic oligomers formed in aqueous solution.74) A further increase in the protein concentration leads to the formation of toxic amyloid fibrils,75–77) which bind to a complex of toll-like receptors 4, 6, and CD36, inducing the nuclear translocation of nuclear factor-kappaB (NFκB), sequentially activating the NLRP3 inflammasome and caspases-8, -9, and -2, finally resulting in apoptosis.78)
Aβ recognizes a cluster of GM1 and is sequentially converted to α-helical forms, nontoxic β-sheet oligomers, and finally toxic, tape-shaped amyloid fibrils upon increases in the protein concentration in the membrane. The ‘amyloid tape’ recognizes a TLR4-TLR6-CD36 complex, inducing apoptosis. See the text for details.
The harmful fibril has a unique structure, which is a flat, tape-like structure composed of a single β-sheet layer formed by mixed in-register parallel and 2-residue-shifted antiparallel structures.79) The latter configuration is stabilized by multiple salt bridges and hydrophobic/π–π interactions80) (Fig. 5(c)). In low polarity environments in membranes, electrostatic interactions are much stronger than in aqueous solution, strengthening the salt bridge interaction in the 2-residue-shifted antiparallel arrangement, compensating for unfavorable repulsive forces between adjacent charged amino acid residues in the in-register parallel structure. Similar amyloid fibrils are also formed in a less polar 1, 4-dioxane/water mixture.81) Trace amounts of pyroglutaminated Aβ-(3–42) enhance aggregation of Aβ-(1–42) on neuronal membranes at physiological nanomolar concentrations.82) Fibril formation in aqueous solution and membranes can be inhibited by nanogels,83) a helical Aβ analog,84) and small compounds.85)
The folding and stability of α-helical membrane proteins, such as G-protein-coupled receptors (GPCR), are governed by interactions between their constituent transmembrane helices.86,87) An attractive approach to measure transmembrane helix–helix interactions is the use of model peptides forming transmembrane helices. First, we designed the completely hydrophobic, ‘inert’ peptide (LALAAA A)3 without any of the specific sidechain interactions and showed that they form a stable transmembrane helix, which can be inserted into and dissociated from lipid bilayers.88,89) The insertion topology can be controlled by helix macrodipole–transmembrane potential interaction. Systematic investigation of the thermodynamics of transmembrane helix–helix interactions was carried out using the modified peptide (ALAALA)3 with a symmetric sequence. FRET measurements between dye-labeled peptides revealed that the helices form antiparallel dimers with a crossing angle of approx. 0° by electrostatic (enthalpic) helix macrodipole interactions. In thinner membranes, the helix termini having partial charges causing the macrodipole are exposed to the aqueous phase with a high polarity, diminishing the helix–helix interaction, whereas in thicker membranes they are buried in a less polar environment, leading to a stronger helix dimerization90,91) (Fig. 7(a)). Changes in hydrophobic thickness of dimonounsaturated PC bilayers from C14 to C22 caused an alteration in the ΔG value from −10 to −20 kJ mol−1 at 308 K. The presence of PE having a smaller headgroup than PC92) and cholesterol93) enhanced the antiparallel dimer formation, which is enthalpically driven. Interestingly, in the presence of cholesterol, the dimer is X-shaped with a crossing angle of 56°, because cholesterol with the small polar group (-OH) and large hydrophobic sterol ring imposes a high lateral pressure in the central part of bilayers.
(a) Effects of membrane thickness on the formation of an antiparallel dimer driven by electrostatic macrodipole interaction. (b) Both the presence of a GXXXG motif and cholesterol affect the formation of a parallel dimer. See the text for details.
A GXXXG motif is known to mediate interhelical packings in helical membrane proteins.94) To examine the effects of the introduction of the motif on the self-association of transmembrane helices, single-pair FRET experiments were devised using a pair of topologically controlled helices incorporated in a liposome.95) The host peptide (ALAALA)3 did not form a parallel dimer because of repulsive macrodipole interaction. In contrast, the AALALAA-AGLALGA-AALALAA peptide with the motif in the middle as the guest self-associated with a small crossing angle of 10° in PC bilayers (Fig. 7(b)). The incorporation of cholesterol destroyed this dimer structure because it is destabilized by the high lateral pressure in the hydrophobic region of the bilayer. Thus, helical assembly is controlled by not only amino acid sequences but also lipid compositions.
Membrane proteins, such as receptors and ion channels, often function as homo- or hetero-oligomers.96–98) Existing approaches have a number of problems in precisely determining the oligomeric states of membrane proteins on the plasma membranes of living cells. Destructive methods such as immuneprecipitation can detect artificial aggregation of proteins in the presence of detergents. As nondisruptive approaches, FRET and bioluminescence resonance energy transfer have been widely employed to monitor protein–protein interactions in living cells. Most studies use the genetic fusion of fluorescent or luminescent proteins to label the target proteins. However, the considerable size of the label may perturb protein functions. Furthermore, it is difficult to control the donor⁄acceptor expression ratios, which are critical to quantitatively analyze the results. In addition, signals from proteins that are not sorted to the plasma membrane hamper accurate analysis. To overcome these problems, we developed a novel labeling technique named the ‘coiled–coil labeling method’99–102) (Fig. 8). It utilizes the tight heterodimer formation between the negatively charged E3 tag (EIAALEK)3 and the positively charged K probe (KIAALKE)x (x = 3 or 4). The E3 tag is genetically fused to the protein of interest and the synthetic, fluorescently labeled K probe is added to selectively label the protein. The dissociation constants are approx. 60 and approx. 6 nM for the E3–K3 and E3–K4 pairs, respectively. Advantages of this method include quick (<1 min) and surface-specific labeling, the precise control of the donor⁄acceptor ratio in two-color labeling, and minimal effects on protein functions. Nonspecific staining of highly negatively charged cell lines such as HEK293 by the positively charged K probes can be avoided by incorporating phosphoserine residues in the probe.103) Selective covalent labeling of amine groups of cell surface proteins is also possible, guided by the coiled–coil assembly.104)
(a) Helical wheel representations of E-3 and K-3. They form a tight heterodimer by virtue of electrostatic (gray arrows) and hydrophobic (white arrows) interactions. (b) Principles of specific labeling of the target protein and the detection of oligomer formation by FRET. The E3 tag is genetically fused to the target protein, which is labeled with donor- and acceptor-attached K probes.
Using this approach combined with a spectral imaging technique (in-cell FRET), we have elucidated the oligomeric states of various membrane proteins in living cells. Regarding membrane receptors, the metabotropic glutamate receptor forms a dimer,105) whereas class-A GPCRs including β2-adrenergic receptor,105) chemokine-CXCR4, dopamine-D2, and prostaglandin-EP1 receptors are monomeric.106) Epidermal growth factor receptors mainly exist as monomers with a small fraction of predimers in the absence of its ligands, whereas approx. 70% of the receptors form dimers after being stimulated with the ligand.107) The self-association of other classes of membrane proteins was also investigated. The integral matrix protein M2 of influenza A virus forms proton-conducting dimers at a neutral pH and these dimers are converted to tetramers at an acidic pH.108) The antiviral drug amantadine hydrochloride inhibits both tetramerization and channel activity. The activity of the dimer is sensitive to the level of cholesterol. These observations are contradictory to the existing model, in which M2 forms a tetrameric channel that opens at an acidic pH and the drug clogs the channel.109) The model has been proposed using truncated M2 proteins (in most cases, transmembrane regions) incorporated in lipid bilayers and detergent micelles. Another example is glycophorin A, which has been proposed to form a dimer. We found that it is monomeric in cell membranes and cholesterol removal is needed to induce its dimerization.104)
The coiled–coil method can also be used to monitor receptor internalization. β2-Adrenergic receptor was doubly labeled with a mixture of pH-sensitive fluorescein and pH-insensitive tetramethylrhodamine.110) Agonist-induced receptor internalization, which accompanies a reduction in endosomal pH, can be detected as an increase in the tetramethylrhodamine-to-fluorescein fluorescence intensity ratio. Furthermore, the screening of agonists and antagonists is possible by utilizing the translocation of the receptor from cell-surface to intracellular regions and the acidification in endosomes.111)
Elucidation of interactions among peptides, proteins, and lipids in membranes is intractable because they are complex and time-dependent. We have clarified various physicochemical phenomena in the membrane milieu by the combined use of spectroscopic, thermodynamic, kinetic, chemical, and molecular biological approaches. Molecular mechanisms proposed by liposomal studies were confirmed by cellular experiments as much as possible. Especially, the involvement of peptide/protein–lipid interactions in innate immunity (AMPs) and the pathogenesis of AD was revealed in detail. Live-cell FRET experiments using the coiled–coil technique suggested the importance of nondisruptive, less perturbing methods to investigate protein–protein interaction in cell membranes. The author sincerely hopes that our achievements contribute to the further understanding of membrane-staged biological phenomena and the development of potent anti-infective agents and disease-modifying drugs for AD.
The author expresses sincere gratitude to the past and present staff and students of my laboratory, including Prof. Yoshiaki Yano (currently at Mukogawa Women’s University), Prof. Masaru Hoshino, and Prof. Kenichi Kawano, as well as collaborators, both in Japan and overseas. I would like to also thank the late Professor Emeritus Masayuki Nakagaki of Kyoto University, the late Professor Emeritus Hiroshi Terada of Tokushima University, the late Professor Emeritus Tetsuro Fujita of Kyoto University, the late Professor Emeritus Koichiro Miyajima of Kyoto University, Professor Emeritus Yukio Sugiura of Kyoto University, Professor Emeritus Tetsurou Handa of Kyoto University, Professor Emeritus Nobutaka Fujii of Kyoto University, Professor Emeritus Yasunori Kozutsumi of Kyoto University, President Emeritus Dr. Katsuhiko Yanagisawa of National Center for Geriatrics and Gerentology, and Professor Emeritus Joachim Seelig of the University of Basel, Switzerland. These studies have been financially supported by the Japan Society for the Promotion of Science (Grants-in-Aid for Scientific Research (B) and Priority Areas), The Mitsubishi Foundation, The Takeda Science Foundation, The Naito Foundation, The Uehara Memorial Foundation, Kato Memorial Bioscience Foundation, The Novartis Foundation (Japan) for the Promotion of Science, Mochida Memorial Foundation for Medical and Pharmaceutical Research, Ono Medical Research Foundation, Inamori Foundation, The Research Foundation for Pharmaceutical Sciences, Life Science Foundation of Japan, Japan Securities Scholarship Foundation, Daiichi Sankyo Foundation of Life Science, and the Shimizu Foundation for Immunology and Neuroscience.
The author declares no conflict of interest.
This review of the author’s work was written by the author upon receiving the 2021 Pharmaceutical Society of Japan Award.
