Probing peptide-mediated membrane pore formation by fluorescence, infrared spectroscopy, and atomic force microscopy

Introduction

Membrane pores and channels play an indispensable role in cell physiology by maintaining the ionic homeostasis, facilitating transmembrane signaling, and enabling directional transport of minerals, metabolites, and biomolecules (Jia et al., 2025; Pottosin et al., 2025; Zhang et al., 2026). Defects in channel function caused by mutations or other alterations result in diseases collectively known as channelopathies (Comini et al., 2024; Mӓnnikkӧ and Kullmann, 2024). Certain pathogenic bacteria produce pore forming toxins that are involved in infection and the ensuing pathology such as hemolysis or pyroptosis (Ilangumaran Ponmalar et al., 2021; Margheritis et al., 2023; Chatterjee et al., 2026). Many organisms have evolved to produce antimicrobial peptides, also known as host defense peptides, which kill the invading pathogen by membrane perforation and cell lysis (Wouters et al., 2026). Analysis of the molecular details of channel- and pore-forming proteins is therefore imperative for understanding the structural features that support their function and for designing drugs to regulate their activities. Since channels and pores are formed by membrane-embedded proteins or peptides and hence are not readily accessible to atomic resolution methods such as crystallography (Guo, 2020), multidisciplinary experimental and computational approaches have been utilized to study such systems (Margheritis et al., 2023; Gregor et al., 2025; Megta et al., 2026; Cui et al., 2026; Kissell et al., 2026; Cheng and Arcario, 2026). Significant advances have been achieved in structure-function analysis of membrane channels and pores by cryo-electron microscopy (cryo-EM), approaches based on fluorescence spectroscopy, including single molecule techniques, Fourier transform infrared (FTIR) spectroscopy, atomic force microscopy (AFM) and its combination with FTIR or Raman spectroscopy.

During the last decade, the spatial resolution of cryo-EM has reached an atomic scale, allowing protein structure determination where single atoms, including hydrogens, are discerned. The structure of apoferritin has been resolved by this method at an astounding 1.25 Å resolution (Yip et al., 2020). Structural changes that accompany the activation of a pentameric proton-gated ion channel reconstituted in lipid nanodiscs have been determined at ∼3 Å resolution (Li et al., 2025). The structures of a voltage-dependent K⁺ channel and a retinal protein that is important for vision in dim light were resolved by cryo-EM at 4.3 Å resolution that shed light on the mechanism of voltage gating and the architecture of the channel forming domains (Mandala and MacKinnon, 2025; Fabrizio et al., 2026). Other recent examples include structure determination of a human microglial K⁺ channel at 3.2 Å resolution that unveiled the mechanisms of tyrosine-assisted gating and regulation by lipids and anesthetics (Rödström et al., 2025; Riel et al., 2025) and the 3.6 Å resolution cryo-EM structure of a bacterial multidrug efflux pump that helped elucidate the molecular details of drug resistance (Gregor et al., 2025).

Fluorescence techniques with various modifications such as total internal reflection fluorescence, single molecule pull-down, and single particle tracking have allowed identification of the stoichiometry and dynamics of multimeric channels in lipid bilayers with around 20 nm resolution (Kissell et al., 2026). Single-molecule fluorescence resonance energy transfer has the power of unveiling non-synchronized conformational changes that are lost through ensemble-averaging in conventional spectroscopy (Margheritis et al., 2023). This method has been used to detect inter-subunit dynamics within ion channels and G protein-coupled receptors and to elucidate the mechanism of ligand or voltage gating and ion selectivity (Wang et al., 2019; Chowdhury et al., 2026). Moreover, the single-molecule fluorescence ligand binding technique has afforded real time tracking of protein-ligand interactions (Chowdhury et al., 2026). Fluorescence spectroscopy is widely used to detect membrane pore formation and direct transmembrane transport of fluorophores such as calcein or carboxyfluorescein based on dequenching following release into the medium (Sciacca et al., 2012; Garg et al., 2013; Sumarokova et al., 2025). Ca²⁺ transport across lipid membranes has been monitored using vesicles loaded with a Ca²⁺-sensitive fluorophore such as Quin-2 or Fura-2, which undergo increase or shift of fluorescence upon Ca²⁺ influx through protein- or peptide-induced pores and binding to the entrapped fluorophore (Sciacca et al., 2012; Kandel et al., 2017; Kandel et al., 2019).

FTIR spectroscopy, especially in combination with complementary computational and experimental techniques such as molecular dynamics (MD) and fluorescence, has elucidated the structural details of channel- or pore-forming proteins and peptides (Kandel et al., 2019; Tatulian et al., 1998; Tatulian et al., 2012; Karkisaval et al., 2024; Drewniak et al., 2024). The structural dynamics of lipid-incorporated human aquaporin-1 and the importance of H-bonding networks in folding and oligomerization have been probed by attenuated total reflection FTIR (ATR-FTIR) spectroscopy, fluorescence spectroscopy, and MD methods (Drewniak et al., 2024). Changes in protein’s secondary structure and the amide hydrogen-deuterium (H/D) exchange resulting from mutations and temperature were analyzed based on ATR-FTIR data, MD simulations modeled the broken H-bonds upon such alterations, and tryptophan (Trp) fluorescence indicated the degree of solvent exposure of local protein sites. Polarized ATR-FTIR spectroscopy, in combination with fluorescence, AFM, and electrophysiology recordings, unveiled specific structural and functional features of amyloid β (Aβ) peptide isoforms including their secondary structure, orientation in lipid bilayers, single channel conductance, pore dimensions, and their supramolecular morphology (Karkisaval et al., 2024).

AFM has been extensively used to study integral or peripheral membrane proteins that are either attached to or embedded in substrate-supported lipid bilayers, including channel- and pore-forming proteins (Sapra, 2019; Ewald et al., 2019; Sanganna Gari et al., 2021; Benn et al., 2021; Lansky et al., 2023; Goh et al., 2023; Sumino et al., 2023; Ye et al., 2024; Lafargue et al., 2025; Liu et al., 2025; Vázquez et al., 2025; Ren et al., 2025). Various modes of AFM, i.e., derivatization of the cantilever tip with an antibody against the channel for single molecule recognition imaging, conductive AFM using a current-conducting tip, and Kelvin probe force microscopy, have provided a wealth of information on membrane topography, channel distribution, surface electrostatic potential, and current-voltage characteristics of mouse astrocyte Ca²⁺ channels at a single molecule level (Li et al., 2024). The use of AFM tip as a conductive electrode has allowed recording of the action potential in membranes of single beating cardiomyocytes (Dong et al., 2024). A similar multimodal AFM approach has been utilized to elucidate the mechanism of apoptosis of breast cancer cells by tumor necrosis factor-α through modulation of Ca²⁺ channel distribution and activity (Wang et al., 2025). The dynamic process of protein translocation across artificial bacterial membrane, i.e., bicelles made from lipids extracted from bacteria, has been visualized using high speed AFM with time resolution of a few seconds (Kanaoka et al., 2025). AFM has been coupled with vibrational spectroscopy to assess the structure and function of ion channels. The structure of voltage-gated sodium channels in their native neuronal membrane has been analyzed at a single molecule level by combining transmission electron microscopy (TEM), AFM, and surface-enhanced Raman spectroscopy (Moretti et al., 2022). Tip-enhanced infrared nanospectroscopy, i.e., combination of AFM with FTIR, has been used to analyze light-induced conformational changes in bacteriorhodopsin that support proton transport across native purple membranes (Giliberti et al., 2019).

The aim of this article is to describe the capabilities of fluorescence, FTIR spectroscopy, and AFM in characterizing ion conducting membrane pore formation by peptides. These include the kinetics of pore formation, the pore size, secondary, tertiary, and quaternary structural features of the pore complex, the affinity between peptide molecules that governs pore formation, and the overall morphology of the functional pore. It is demonstrated how the methods traditionally considered as “low resolution” techniques are able to produce molecular models of membrane pores in great detail.

Fluorescence spectroscopy

The basic principles of fluorescence spectroscopy and its use in protein structure analysis can be found in the highly acclaimed book by Lakowicz (2006). Fluorescence spectroscopy has been used to obtain detailed information on membrane pore or channel formation by peptides or proteins and to elucidate the structural features and the mode of membrane interaction of the pore complex. Two frequently utilized approaches are discussed in this section. The first method involves loading lipid vesicles with a fluorophore at self-quenching concentration and addition of the pore forming agent. The gradual release of the fluorophore through the newly formed pores results in dequenching, which allows recording of the kinetics and other details of pore formation. This method has been applied to detect membrane pores formed by the C-terminal 20 amino acid residue peptide of Bax protein (Bax-C20: VTIFVAGVLTASLTIWKKMG). Bax is a B cell lymphoma 2 family protein involved in apoptosis via outer mitochondrial membrane permeabilization, release of cytochrome c, and caspase-mediated cell death, and its C-terminal α-helix plays a key role in membrane binding and insertion (Sekar et al., 2022; Clifton et al., 2023; Nguyen et al., 2024). A 21-amino acid residue sequence corresponding to Bax C-terminus was able to form pores in lipid vesicles allowing efflux of carboxyfluorescein (Torrecillas et al., 2007). A similar fluorescence-based study showed that Bax-C20 indeed possesses potent pore forming capability where the two lysines play an important role (Garg et al., 2013). Unilamellar lipid vesicles composed of 100% 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) or 70 mol% POPC and 30 mol% 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG) were loaded with calcein at a self-quenching concentration (10 mM). The Bax-C20 peptide, solubilized in 6 mM dodecylphosphocholine (DPC) micelles, was added to the vesicles so that DPC concentration dropped 10-fold and the detergent dispersed below its critical micelle concentration of 1.5 mM, allowing Bax-C20 to partition into the vesicle membrane. Gradual increase in calcein fluorescence following addition of the peptide indicated calcein release and dequenching (Figures 1A,B). Addition of 0.6 mM DPC exerted negligible effect while solubilization of the lipid by 0.05% Triton X-100 resulted in maximal fluorescence due to total degradation of the vesicles and dilution of calcein. It was shown in separate light scattering experiments that 0.6 mM DPC did not affect vesicle integrity while Triton X-100 converted the vesicles to detergent-lipid micelles (Garg et al., 2013).

The kinetics of fluorescence after addition of three different concentrations to the peptide to vesicles are presented in Figures 1A,B. The rise of fluorescence was described by double exponential curves comprising relatively fast and slow components with time constants ∼0.2 min and ∼5.0 min, respectively. Higher peptide concentrations as well as the negative charge of the membranes favored more efficient calcein release, the latter apparently due to the two cationic amino acid residues (lysines) in the peptide. Further analysis of the data allowed determination of peptide-peptide affinity constants (10⁶-10⁷ M^-1) and the number of peptide molecules in the pore complex (4-8 peptides) (Garg et al., 2013).

Fluorescence experiments were conducted to determine the degree of membrane insertion of the peptide based on differential quenching of the single tryptophan (Trp₁₆) by lipids brominated at different positions along the sn-2 acyl chain (Tatulian et al., 2012). Trp₁₆ was located at ∼9 Å from membrane center for POPC membranes. The anionic charge of membranes contributed by 30% POPG strongly influenced the depth of membrane insertion, placing Trp₁₆ at ∼3 Å from the membrane center. These results, combined with polarized ATR-FTIR data on the secondary structure and orientation of the membrane-embedded peptide (vide infra), allowed construction of a pore model where eight peptide molecules form an annular structure, named α/β ring, with a central pore large enough to allow diffusion of calcein (Figures 1C,D) (Tatulian et al., 2012).

The above-described method can detect relatively large pores that are able to transport calcein molecules but miss smaller pores or channels. A second fluorescence-based method that detects small ion transporting pores is based on sequestration of a calcium-sensitive fluorophore and Ca²⁺ ions by the vesicle membrane and addition of the peptide to the vesicle suspension. Pore formation by the peptide results in transmembrane Ca²⁺ transport and combination with the fluorophore, producing gradual increase in fluorescence intensity. An 11 amino acid stretch of the amyloid β peptide Aβ_25–35 (GSNKGAIIGLM) is highly cytotoxic and capable of ion-conducting pore formation in cellular and artificial lipid membranes (Kandel et al., 2017; Kandel et al., 2019; Fantini et al., 2014; Peters et al., 2016). The mechanism of membrane pore formation by Aβ_25–35 has been studied by using lipid vesicles loaded with a Ca²⁺-sensitive fluorophore Quin-2 in a Ca²⁺-less buffer. After removal of Quin 2 from the extravesicular space by column chromatography, CaCl₂ was added externally so Ca²⁺ ions and Quin-2 were separated from each other by the vesicle membranes. Aβ_25–35 peptide oligomers were prepared by stirring in buffer for 2.5 h and were added to the vesicle suspension at different concentrations. This resulted in concentration-dependent increase in Quin-2 fluorescence, indicating Ca²⁺ influx and binding to Quin-2 (Figure 1E) (Kandel et al., 2017). Maximum increase in fluorescence was produced by a non-fluorescent Ca²⁺ ionophore 4-Br-A23187. Data analysis, using the kinetics and the relative increase in fluorescence at each peptide concentration, allowed determination of the rate constant of pore formation, the affinities between the peptide molecules, and the number of peptide molecules in the pore complex. Further studies using circular dichroism and ATR-FTIR spectroscopy provided information on the secondary structure and the orientation of the membrane-embedded peptide, which led to a molecular model of the pore complex where six β-barrel-like hexamers of Aβ_25–35 form an annular structure containing total 36 peptide monomers with a ∼13 Å wide central pore (Figure 1F) (Kandel et al., 2019).

The intrinsic fluorescence of membrane proteins has been used to probe the degree of membrane insertion. One widely used technology is determination of membrane insertion depth based on quenching of Trp fluorescence by lipids brominated at distinct positions of their acyl chains (Tatulian et al., 2012; Pande et al., 2006). Quenching of Trp fluorescence by water-soluble or membrane-soluble quenchers such as acrylamide or 10-doxylnonadecane, respectively, provides qualitative information on membrane insertion (Pande et al., 2006). Dependence of Trp fluorescence on the polarity of the environment, i.e., a ∼10 nm blue shift upon transfer of Trp from the aqueous medium to the membrane, is an effective method to estimate membrane recruitment of a protein without the need of extrinsic fluorophores (Pande et al., 2006; Qin et al., 2004). Tyrosine (Tyr) does not exhibit environment polarity dependence of the emission wavelength, making this latter approach ineffective for proteins that are Trp-less but have one or more Tyr residues. It has been shown that Tyr fluorescence can undergo splitting due to partial excited-state deprotonation of the phenolic hydroxyl group assisted by a nucleophilic buffer component or water itself (Karkisaval et al., 2024). Since deprotonated Tyr emits at significantly longer wavelengths (∼340 nm vs. ∼307 nm for the protonated state), this effect results in a split emission band as shown for the Aβ_pE3-40 peptide that has a single Tyr and no Trp (Figure 1G, cyan, green, and gray lines) (Aβ_pE3-40 is the Aβ_1-40 peptide with the first two residues removed and glutamic acid 3 converted to pyroglutamic acid.) Interestingly, the splitting did not take place in the presence of lipid vesicles (Figure 1H), suggesting membrane insertion of the peptide and sequestration from the solvent that contains the proton acceptor. Thus, Tyr fluorescence can be used as an effective means to probe membrane insertion of Trp-less proteins.

Like any experimental technique, fluorescence spectroscopy has its limitations. It is not applicable to proteins or peptides with no aromatic (fluorescent) amino acids. In such cases, a Trp or Tyr can be incorporated at a desired site by chemical synthesis or recombinantly, but that is likely to affect the protein’s/peptide’s structure and function. For proteins with multiple fluorescent amino acids, site-specific information such as conformational changes, solvent accessibility, or the depth of membrane insertions cannot be obtained accurately. In cases when the use of intrinsic fluorescence is impracticable, protein labeling by exogenous probes offers solutions (Kyrychenko and Ladokhin, 2024; Marshall and Sutherland, 2025). Labeling with relatively large probes as resonance energy transfer donors or acceptors, which can be as big as the green fluorescent protein, can perturb the molecular structure under study (Marshall and Sutherland, 2025; Raina et al., 2012). Probes of small size or targeted incorporation of unnatural amino acids with spectrally shifted fluorescence properties have provided site-specific structural information on proteins including ion channels (Kyrychenko and Ladokhin, 2024; Marshall and Sutherland, 2025; Infield et al., 2024). These techniques utilize nonsense suppression or protein semisynthesis (Infield et al., 2024). The former involves insertion of a stop (nonsense) codon at the site of interest of recombinant DNA, e.g. TAG (UAG in mRNA), and use of a nonsense suppressor tRNA with an anticodon complementary to the stop codon, CUA in this case, loaded with an unnatural amino acid. The latter is based on chemical synthesis of a peptide (part of the target protein) harboring the modified amino acid(s) and ligation with the rest of the recombinantly expressed and purified protein. These procedures are challenging yet quite attainable and produce fully functional proteins for biophysical analysis of their function such as domain rearrangements related to ion channel gating (Infield et al., 2024).

FTIR spectroscopy

The capabilities of FTIR spectroscopy in elucidating the structure and function of membrane proteins, including pore- and channel-forming proteins, are numerous. These include the analysis of the protein’s secondary structure, including local, site-specific structure using isotope labeling, the rate of amide H/D exchange and hence the degree of solvent exposure and the compactness of the protein’s tertiary fold (Tatulian et al., 1998; Goldblatt et al., 2015; Tatulian, 2026). Polarized ATR-FTIR studies on proteins embedded in lipid membranes deposited on an internal reflection element (IRE) have allowed determination of the orientation of both α-helices and β-sheets relative to the membrane normal thereby elucidating the functionally important mode of membrane interaction of membrane proteins (Kandel et al., 2019; Tatulian et al., 2012; Karkisaval et al., 2024; Tatulian et al., 2005). In an ATR-FTIR experiment, one or both surfaces of the IRE, such as a germanium plate, are covered with a lipid bilayer with incorporated protein or peptide (Figure 1I). The sample can be a lipid multilayer rather than a single bilayer; the technical details have been described elsewhere (Tatulian, 2026). The infrared light is plane polarized by a polarizer and directed by two mirrors into the IRE where it undergoes a series of internal reflections and then exits the IRE and reaches the detector via two other mirrors. At each internal reflection, an evanescent wave is created at the IRE surface where the lipid-protein sample is located, resulting in absorption of part of the light and hence allowing recording of the absorbance spectrum. Spectra at two mutually perpendicular plane polarizations allow determination of orientations of α-helical or β-sheet structures within the protein.

The secondary structure of the Bax-C20 peptide that forms calcein-transporting pores in lipid membranes was determined by ATR-FTIR to be 40% α-helix, 32% β-sheet, 22% turn, and 6% irregular (unordered) structure (Tatulian et al., 2012). The tilt angle of the helical axis relative to the membrane normal was 60° for POPC membranes and 42° for POPC/POPG (7:3 molar ratio) membranes. The orientation of β-strands with respect to the β-barrel axis (angle β) also depended on the membrane charge: ∼30 and ∼40° for POPC and POPC/POPG membranes, respectively. Considering that the pore radius is inversely proportional to cosβ (Kandel et al., 2019), more efficient (wider) pores are expected to form in membranes containing the anionic lipid POPG, as confirmed by data shown in Figures 1A,B. The diameter of the Bax-C20 pore formed by peptide backbones was estimated based on geometric constraints derived from polarized ATR-FTIR data to be 28–32 Å (Tatulian et al., 2012). The actual pore diameter shown in Figure 1D is ∼25 Å due to the layer formed by the side chains. This is an example of how fluorescence and FTIR studies can complement each other in producing a membrane pore structure like the α/β-ring shown in Figures 1C,D.

NMR studies on Bax C-terminal 27-mer peptide in DPC micelles identified α-helical dimer formation stabilized by intermolecular nonpolar contacts and hydrogen bonding (Zhang et al., 2025). MD simulations on the same peptide demonstrated its ability to embed in POPC bilayers and form a transmembrane pore lined by 8 or 16 peptide units in mostly (∼55%) α-helical conformation with 8.5 Å and 15.3 Å pore diameter, respectively (Jiang and Zhang, 2019). MD results on Bax lacking both N-terminal and C-terminal helices α1 and α9 produced a pore model in distearoylphosphatidylcholine membranes consisting of six dimers with pore diameter of 48 Å (Zhang et al., 2017). Simulations on the full-length Bax in membranes formed by mitochondrial lipids resulted in a 40–70 Å-wide pore model consisting of eight dimers with the α9 helix serving as a transmembrane anchor (Zhang et al., 2025). Despite the significant divergence in pore size, which may stem from differences in protein or peptide construct and distinct lipid environments, all results point to formation of large membrane pores by Bax and its C-terminal helix.

Polarized ATR-FTIR studies were instrumental in constructing a structural model for membrane pores formed by the Aβ_25–35 peptide. The β-strands of this peptide were tilted at ∼22° with respect to the central axis of 6- or 8-stranded β-barrel structures for which the pore diameter was estimated to be 10.2 Å and 13.3 Å, respectively (Kandel et al., 2019). The internal cavity of the barrel-like structure has been shown to accommodate the inward oriented side chains, leaving little room for the formation of an ion-conducting pore or channel by a single β-barrel. The ion-conducting pore was proposed to form by a supramolecular structure: a hexamer of 6-stranded β-barrels with a ∼13 Å-wide central pore, as shown in Figure 1F. MD analysis of Aβ_25–35 in POPC/POPG bilayers produced octameric or decameric β-barrels (Dutta et al., 2023). The octameric barrel was congested with the side chains in contrast to an earlier MD structure of an 8-stranded β-barrel of Aβ_25–35 in POPG membranes with a 3.5–4.0 Å-wide inner pore (Chang et al., 2011), whereas the decameric barrel was more relaxed and stable with a backbone-to-backbone diameter of 19 Å (Dutta et al., 2023).

ATR-FTIR spectra of Aβ_1-42 embedded in lipid multilayers at p- and s-polarizations are presented in Figures 1J,K. The spectral range covers the lipid carbonyl stretching mode and the peptide’s amide I region. The lipid carbonyl mode is fitted with two components corresponding to hydrated and dehydrated carbonyls. The amide I band of the peptide is fitted with components corresponding to turn, α-helix, irregular (unordered) structure, and β-sheet structure. These data were used to determine the secondary structure content in the peptide, which produced 46% β-sheet and smaller fractions of other structures. Spectra at two different polarizations were further used to determine the orientations of the β-strands: ∼30° relative to the barrel axis. Using this tilt angle, a pore diameter of around 11 Å can be estimated for a 6-stranded β-barrel formed by Aβ_1-42 (Kandel et al., 2019). Note, however, that this is the diameter of a cylinder formed by the peptide backbone in a β-barrel conformation; the inside oriented protein side chains will occupy a significant part of the channel resulting in a much narrower (and irregular) passageway that may hardly support efficient ion transport. Yet the peptide did form ion conducting channels in lipid membranes with single channel conductivity of around 100 pS, as shown in Figure 1L. Using this conductivity, combined with the channel length based on ATR-FTIR data (57 Å) and a 5-fold enhanced resistivity of the salt solution inside the channel relative to its bulk value, a channel diameter of ∼5.6 Å was estimated (Karkisaval et al., 2024), which is consistent with a pore diameter fluctuating between 4 Å and 8 Å obtained by MD simulations for Aβ_1-42 embedded in lipid bilayers mimicking neuronal lipid composition (Nguyen et al., 2019). Although this result is in accord with the interpretation of the above estimate, the ion conducting unit may be a more complex structure than a single β-barrel-like structure, such as a supramolecular assembly of barrels or other oligomers (Viles, 2023; Durell et al., 2026), making such comparison irrelevant. It should also be noted that Aβ_1-42 channels formed in cell membranes had significantly larger conductivity (340–630 pS) and correspondingly larger dimensions, ranging from 17 Å to 24 Å in diameter (Bode et al., 2017). A possible explanation of different channel conductivities in lipid bilayers and cell membranes is the presence of additional components in the cell membrane such as the cellular prion protein (Viles, 2023) or GM1 ganglioside (Durell et al., 2026) that act as accessory elements for membrane binding and pore formation by Aβ_1-42.

Despite the exquisite sensitivity of FTIR spectroscopy to secondary structural changes of proteins, it comes with certain limitations. FTIR determines the global, rather than site-specific, structure of a protein. In addition, there is significant overlap between the amide I components assigned to distinct secondary structures as seen in Figures 1J,K. Furthermore, the peak-fitting procedure does not uniquely determine the fractions of secondary structures in a protein as a multitude of distinct combinations of amide I components can result in a nearly identical amide I lineshape (Tatulian, 2026). Local structural information can be obtained by labeling targeted sites of a protein with stable isotopes such as ¹³C and ¹⁵N. Such labeling can be applied to a single amino acid or a stretch using the above-mentioned nonsense suppression or peptide ligation (protein semisynthesis) approaches (Tatulian et al., 2005; Tatulian, 2024). The problem with single site labeling is that the signal from one amino acid can hardly be reliably resolved from the whole amide I band. In addition, a single amide I oscillator has little diagnostic value in structural terms because the characteristic, structure-specific frequency is produced by normal modes resulting from vibrational coupling between adjacent amide units. Even when several amino acids are labeled and generate a distinctive spectral feature, the overlap between the amide I components generated by labeled and unlabeled parts of the protein reduce the precision of structure determination (Tatulian, 2024). Direct visualization of membrane pores or channels by various forms of microscopy is required to shed light on the actual pore structure, as described in the following section.

Atomic force microscopy

AFM is a scanning probe technique widely used in biomedical research to unveil the structural and morphological features of various samples such as proteins, nucleic acids, membranes, cells, and tissues with a sub-nanometer resolution (Sapra, 2019; Benn et al., 2021; Ye et al., 2024; Lafargue et al., 2025; Ren et al., 2025; Dokukin and Sokolov, 2017; Almonte and Colchero, 2017; Shen et al., 2025). As physical contact of the probe with the soft biomaterial can damage the latter, a feedback system is utilized to maintain a gap, e.g. ∼ 0.2 nm, between the sample and the tip of the probe (cantilever) corresponding to a constant force of interaction. The lateral resolution is proportional to the width of the force profile and depends primarily on the radius of curvature of the tip. Sharper tips, with radius of curvature around 2–5 nm, produce higher resolution images. The scan rate is another experimental parameter affecting the image quality—better resolution can be achieved with slower scan rates. However, slow scan rates are associated with long scan time; dynamic structural/morphological transformations of the sample during the experiment can produce images with temporally (and hence structurally) asynchronous parts. The properties of the sample such as structural stability, stiffness, elasticity and the strength of its attachment to the atomically flat substrate, such as mica, are equally important factors determining the image quality. By careful adjustment of experimental and software parameters, images with lateral resolution of around 0.5 nm and vertical resolution of 0.1 nm are obtained (Sapra, 2019; Shen et al., 2025), making AFM a powerful technique for structural analysis of membrane proteins. The image acquisition can be done on samples in vacuum, in natural atmosphere (air), or under aqueous buffer, the latter providing a near-physiological environment. The shortcomings of AFM compared to other structural methods are that the attachment of the sample to a solid surface can potentially modulate the native structural features of biomolecules and prevent actual 3-dimensional structural analysis as only the exposed surface is being scanned.

To reduce the perturbation of the sample by the cantilever tip due to the lateral force that is present in the continuous contact scanning mode, the “tapping mode” (a.k.a. “intermittent contact mode”) has been introduced where the tip vibrates in vertical direction at a frequency in the ∼10–500 kHz range with a ∼10–100 nm amplitude thus temporally minimizing the tip-sample contact and hence curtailing the potentially damaging effect of the cantilever on the sample (Sapra, 2019; Goh et al., 2023; Li et al., 2024). In PeakForce tapping mode, the impact of the tip on the sample is further moderated by limiting the maximum normal force via a feedback mechanism to a certain value (e.g., PeakForce setpoint ≤80–400 pN) and the image is constructed based on cantilever deflections due to periodic tapping along the scanning trajectory (Karkisaval et al., 2024; Ren et al., 2025; Dokukin and Sokolov, 2017; Gamage and Pan, 2025; Wei et al., 2025).

Membrane pore formation by the Aβ_1-42 peptide has been directly visualized by cryo-electron tomography, TEM, and AFM (Karkisaval et al., 2024; Viles, 2023; Durell et al., 2026; Lin et al., 2001; Quist et al., 2005; Mrdenovic et al., 2022; Lu et al., 2026). These studies have identified annular structures with outer diameter of 80–120 Å composed of 4-6 oligomeric units, each containing around 6 peptide monomers in a β-barrel-like conformation. These structures were evidently responsible for ionic currents recorded in lipid bilayers and cell membranes treated with Aβ_1-42 (Karkisaval et al., 2024; Viles, 2023; Bode et al., 2017; Lin et al., 2001; Quist et al., 2005). TEM micrographs of unilamellar lipid vesicles with reconstituted Aβ_1-42 show peptide assemblies approximately 80–100 Å wide both embedded in vesicle membranes and scattered in the buffer as well as holes of similar dimensions in the membranes (Figure 1M) (Lu et al., 2026). These observations suggest that the peptide forms aggregates that are able to insert into the membrane and leave the membrane thereby extracting lipid and degrading membrane integrity, as described earlier (Ewald et al., 2019; Viles, 2023; Mrdenovic et al., 2022). Recent PeakForce tapping mode AFM studies revealed the oligomeric pore-like structure of Aβ_1-42 composed of four units assembled around a conspicuous central pore (Figure 1N). This structure has been modeled by MD simulations as a 24-mer composed of four hexameric β-barrels with a pore diameter of 17 Å (Durell et al., 2026) consistent with earlier estimates (Viles, 2023; Lin et al., 2001; Quist et al., 2005; Durell et al., 2022). Various types of β-barrel structures, including regular, single-walled and more complex concentric β-barrels, have been constructed for Aβ_1-42 using MD (Nguyen et al., 2019; Durell et al., 2026; Durell et al., 2022; Shafrir et al., 2010; Strodel et al., 2010; Sepehri and Lazaridis, 2023). NMR studies combined with MD yielded an octameric β-sandwich composed of two juxtaposed 6-stranded β-sheets, each comprising four peptide units, that interacts with lipid headgroups through the polar side chains of the flanking β-strands and thereby forms channels for “edge conductivity” (Ciudad et al., 2020; Matthes and de Groot, 2023). Membrane spanning molecular models have been constructed by MD approaches for other Aβ species such as Aβ_1-40 and truncated or modified (e.g., pyroglutamylated) peptides (Dutta et al., 2023; Chang et al., 2011; Nguyen et al., 2019; Sepehri and Lazaridis, 2023; Jang et al., 2010a; Jang et al., 2010b; Connelly et al., 2012; Di Scala et al., 2014a; Di Scala et al., 2014b; Gillman et al., 2014; Jang et al., 2016; Dai et al., 2022; Drajkowska and Molski, 2026). These studies reported a multitude of pore shapes and dimensions including β-barrel-like and helical bundle structures that reflect differences in the choice of the initial structure, the specific lipid environment and other details. Determination of the actual pore structure and the internal pore dimension is challenging due to multiple factors including the intricate inner architecture of the pore and the limited resolution of electron or atomic force microscopy. Modeling the pore as a perfect cylindrical passageway is an oversimplification. Cutting edge structural approaches with superior resolution coupled with electrophysiological data and MD simulations hold the key to deciphering the molecular details of membrane channels and pores.

Conclusion

Significant progress has been made during the last decade in elucidation of the structure and function of membrane pores and channels. As a single method cannot provide a comprehensive structural characterization, multidisciplinary approaches need to be employed to achieve this goal. This is especially true for proteins or peptides that are resistant to crystallization for diffraction experiments, such as the amyloid proteins. Biophysical techniques that are traditionally considered as low resolution methods, such as fluorescence, infrared spectroscopy, and AFM have provided invaluable information on the structural details of membrane pores and channel. Combination of these and other structural techniques with measurements of single channel conductivities may resolve the molecular details of channel- or pore-forming proteins.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Amita Rani Sahoo, Case Western Reserve University, United States

Reviewed by: Nikhil Bharambe, Nanyang Technological University, Singapore

Gašper Šolinc, National Institute of Chemistry, Slovenia