The multiple sclerosis drug fingolimod induces pore-like membrane defects, contributing to its antimicrobial activity

Abstract

Although receptor-mediated mechanisms account for the therapeutic action of numerous FDA-approved drugs, emerging evidence suggests that many of these therapeutics have off-target antimicrobial activities. One example is fingolimod, an immunomodulator used to treat multiple sclerosis, that has been reported to have antimicrobial effects associated with membrane permeabilization. Yet the molecular mechanism by which fingolimod alters bacterial membranes remains unknown. As a cationic amphiphilic drug, fingolimod is comprised of both hydrophobic and positively charged regions that can enable membrane interactions. We show that fingolimod compromises membrane integrity in Escherichia coli and Pseudomonas aeruginosa, contributing to its antimicrobial activity. To determine how fingolimod disrupts membrane integrity, we used planar lipid bilayer electrophysiology with phospholipid compositions mimicking Escherichia coli membranes. Using gramicidin A channels as molecular biosensors, we show that fingolimod alters both mechanical properties and surface charge of lipid bilayers at concentrations that have antimicrobial effects. At higher concentrations, fingolimod induces pore-like defects, as revealed by conductance measurements. The Bilayer Overtone Analysis suggests an autocatalytic mechanism for the exchange of fingolimod between opposing lipid leaflets. Molecular dynamics simulations correlate fingolimod’s preference for pore-favoring curvature with its strong interactions with lipids and trans-leaflet translocation. These findings point to a molecular mechanism for fingolimod's off-target activity and provide a basis for understanding how some cationic amphiphilic drug structures can contribute to membrane-specific effects that compromise bacterial physiology.

Fingolimod (FTY720, FL), an FDA-approved treatment for relapsing multiple sclerosis, acts as an antagonist of sphingosine-1-phosphate receptors, leading to lymphocyte sequestration and immunomodulation (1). Structurally, fingolimod belongs to a class of therapeutics known as cationic amphiphilic drugs (CADs), which have a positively charged amine group and a hydrophobic region. These characteristics enable adequate solubility in body fluids while allowing penetration into lipid membranes (2). Beyond its intended therapeutic mechanism, fingolimod, like other CADs, may have off-target interactions with cellular membranes that remain poorly understood, and some members are known to induce phospholipidosis (2, 3, 4). Our previous work, along with studies by other labs, shows that antidepressants, which are also CADs, display antimicrobial effects (5, 6).

Recent studies have revealed antimicrobial properties of fingolimod, proposed to be associated with perturbation of bacterial quorum-sensing systems and membrane permeabilization (7, 8, 9). Furthermore, fingolimod displays synergistic effects with antibiotics such as doripenem and colistin, enhancing their efficacy against resistant bacteria, including carbapenem-resistant Escherichia coli (E. coli) and Klebsiella pneumoniae (10, 11). For example, fingolimod combined with doripenem significantly reduced the expression of resistance-related genes, inhibited bacterial motility, and improved survival in infection models (10). Fingolimod combined with colistin disrupted bacterial membranes, increased oxidative stress, and altered fatty acid metabolism, leading to significant bactericidal effects both in vitro and in vivo (11). Furthermore, fingolimod has been reported to exhibit antibacterial activity against Gram-positive bacteria, including Staphylococcus aureus and Staphylococcus epidermidis, and was shown to inhibit the in vitro growth of Clostridium perfringens (11, 12).

Although fingolimod displayed antimicrobial activity and was shown to disrupt bacterial membranes, the biophysical mechanisms underlying its membrane-permeabilizing effects remain poorly understood. Importantly, there are no known bacterial receptors that directly mediate the activity of fingolimod, suggesting that targeting of lipid membranes may play a central role. Given their amphiphilic nature, high lipophilicity, and protonatable amine group (pKa ∼8–9), CADs are expected to partition into lipid bilayers and interact electrostatically with negatively charged membrane surfaces, potentially altering membrane structure and function. Thus, a key unresolved research question is whether fingolimod interacts with lipid membranes to modify their biophysical properties and induce permeabilization, thereby contributing to its antimicrobial activity.

We first identified fingolimod in a small-scale screen of CADs with bactericidal activities. To test the hypothesis that fingolimod interacts with lipid bilayers through its amphiphilic structure, we used a combination of biophysical approaches and found that fingolimod indeed alters the properties of planar lipid membranes and induces membrane permeability at concentrations that exhibit antimicrobial activity. Molecular dynamics (MD) simulations further supported our experimental results, suggesting that fingolimod induces metastable membrane pores that catalyze its own exchange between membrane leaflets. These findings suggest a molecular mechanism contributing to fingolimod's antimicrobial activity and identify potential strategies for developing and repurposing new membrane-targeting antibiotics based on the fingolimod scaffold.

Results

A screen of 25 CADs reveals antimicrobial activity of fingolimod

In previous work, it was reported that antidepressants and fingolimod displayed antimicrobial activity (6, 67, 7, 9, 10). Given their characteristic CAD structures, we performed a small-scale screen of primary, secondary, and tertiary amine CADs to assess their effects on E. coli growth. We assembled a library of 25 compounds (Table S1), treated E. coli with 40 μM of each compound, and measured the growth profile over a 20 h period. We observed 24 drugs (upper curves) with ∼10% decrease in total growth compared to untreated cells. However, fingolimod was extremely toxic and prevented the growth of cells completely (Fig. 1A, lower red curve). We selected fingolimod for follow-up experiments and tested growth as a function of several concentrations of the compound. Even at lower concentrations (10 μM), a modest reduction of growth was observed (Fig. 1, B and D). To determine whether this growth effect was specific to E. coli or more broadly to other bacterial species, we tested effects on the growth of Pseudomonas aeruginosa (P. aeruginosa), another clinically relevant gram-negative bacterium. Similar to the E. coli, treatment with increasing concentrations of fingolimod (10 and 40 μM) led to a progressive reduction in P. aeruginosa growth (Fig. 1, C and D). These findings demonstrate that fingolimod exerts antimicrobial activity in a concentration-dependent manner.

Figure 1.

Fingolimod exhibits strong antimicrobial activity and inhibits growth of E. coli and P. aeruginosa. A growth inhibition screen of 25 CADs (Table S1) against E. coli. Bacterial cultures were treated with 40 μM of each compound and OD₆₀₀ was measured over a 20 h period. While many CADs had minimal to moderate effects, fingolimod showed the most potent growth inhibition (red curve), nearly abolishing bacterial growth across the time course. B and C, concentration-dependent inhibition of bacterial growth by fingolimod. E. coli (B) and P. aeruginosa (C) were grown in LB medium supplemented with increasing concentrations of fingolimod (0, 10, and 40 μM). OD₆₀₀ was recorded every 10 min over 20 h. Fingolimod inhibited growth in a dose-dependent manner in both bacterial strains, with E. coli showing slightly greater sensitivity. D, end-point optical density analysis of bacterial growth after 20 h in the presence of fingolimod based on panel B and panel C. Bar graphs show mean OD₆₀₀ values ± SD for E. coli and P. aeruginosa at each concentration tested. A clear dose-dependent reduction in bacterial growth is observed, consistent with the kinetic growth curves. Note: The differences between the 40 μM fingolimod curves in panels A and B reflect variability between independent experiments; error bars in panel D indicate the standard deviation of replicate measurements within the same experiment.

Gramicidin A as a molecular biosensor to study CAD membrane interaction

Gramicidin A (grA) serves as an extremely sensitive molecular biosensor of lipid bilayer mechanics (13). GrA is a short linear peptide (15 amino acids) that forms ion channels upon dimerization of its two monomers incorporated into opposing membrane leaflets (14). The dimeric grA channel (∼2.2 nm) (15) is shorter than the typical bilayer thickness (16), resulting in a hydrophobic mismatch (17, 18, 19, 20). Therefore, channel formation requires local membrane bending, making grA highly responsive to bilayer mechanical properties (15, 21). The schematic in Figure 2A illustrates how cationic amphiphiles can alter lipid bilayer properties, thereby modifying grA channel conductance, lifetime, and opening frequency. GrA channel conductance is defined by the concentration of K⁺ ions at the channel entrance, which, in turn, depends on the membrane surface charge and the bulk electrolyte concentration (22). When amphiphilic fingolimod is added to the planar membrane, it partitions into the bilayer and reduces membrane negative surface charge via its positively charged amine group, repelling the K⁺ ions from the bilayer surface (Fig. 2A). This reduces the cation concentration near the channel entrance, thereby decreasing channel conductance (22). The lifetime of the grA conducting dimer depends on several membrane parameters, such as bilayer thickness, spontaneous curvature, and lipid packing stress (13, 23, 24). Multiple studies have shown that grA channel parameters can report on amphiphilic drug-induced modifications of bilayer electrostatics and hydrophobic core properties (23, 25, 26).

Figure 2.

Fingolimod decreases GrA conductance and increases lifetime.A, schematic of GrA channel formation and recording. GrA is a short linear peptide that forms ion channels upon dimerization of two monomers from opposing lipid bilayer leaflets. Cationic amphiphiles can alter bilayer properties, thereby modifying GrA channel parameters: conductance and lifetime. The inset trace shows GrA current recording, highlighting measured channel parameters: conductance (current amplitude) and lifetime (channel duration). Created in BioRender.com. B, representative current records of the GrA channels reconstituted into a planar lipid bilayer made from lipid mixture of DOPE/DOPG/CL (0.75:0.2:0.05 w/w) before (control) and after addition of 0.5 μM and 1 μM fingolimod to the cis side of the membrane. The membrane bathing solution contained 0.1 M KCl buffered by 5 mM Hepes at pH 7.4. A voltage of 100 mV was applied. The dashed black line indicates the zero current level; blue dashed lines indicate single-channel amplitudes. Addition of fingolimod decreases channel conductance and increases lifetime. For data analysis, the current records were digitally filtered at 5 kHz using a low-pass Bessel (8-pole) filter. C, GrA single-channel conductance normalized versus control conductance at different fingolimod concentrations. D, GrA lifetime as a function of fingolimod concentration. Lifetimes were normalized to control values in the absence of fingolimod. Data points in C and D represent mean ± SD of three independent experiments. Experimental conditions are as in (B). CL, cardiolipin; DOPE, dioleoyl-phosphatidylethanolamine; DOPG, 1,2-dioleoyl-sn-glycero-3-phosphoglycerol; GrA, gramicidin A

Fingolimod alters planar membrane properties

To investigate the effect of fingolimod on bilayer membrane properties, we measured grA channel conductance and lifetime at increasing concentrations in planar lipid bilayers composed of an E. coli-mimicking phospholipid mixture (75% PE, 20% PG, 5% CL), representing the inner membrane and inner leaflet composition of the outer membrane (Fig. 2B). The representative current traces of grA channels obtained from the same planar membrane made of this lipid mixture are shown in Figure 2B. It can be seen that addition of fingolimod (0.5 and 1 μM) results in significant and dose-dependent decrease of channel conductance and increase of its lifetime.

Normalized grA channel conductance shows a progressive reduction with increasing fingolimod concentration (0.5–2 μM) in 0.1 M KCl solution (Fig. 2C). This reduction is due to greater surface charge neutralization by the positive charges of fingolimod and a decrease in the K⁺ concentration near the channel entrance (22). Normalized grA lifetime increases in a dose-dependent manner (Fig. 2D). These results indicate that the membrane hydrophobic core properties are modified upon fingolimod partitioning into the membrane, as shown in the schematic in Figure 2A, in a manner that stabilizes the grA conducting dimer, resulting in a prolonged lifetime compared to the control.

Fingolimod permeabilizes E. coli and P. aeruginosa membranes

Previous work with fingolimod and sphingosine, a structural analog of fingolimod, has demonstrated their membrane-permeabilizing activity against S. aureus and E. coli, respectively (7). Here, we tested propidium iodide (PI) uptake, a membrane-impermeable dye whose fluorescence quantum yield increases markedly upon DNA binding (27). Therefore, bright fluorescence signals indicate membrane permeabilization, as PI gains entry into the cell and is able to bind to genomic DNA. E. coli or P. aeruginosa grew in the presence of increasing doses of fingolimod (0, 10, 20, and 40 μM) for 1 hour, followed by assays for propidium uptake. As shown in Figure 3, both E. coli and P. aeruginosa displayed a dose-dependent increase in PI fluorescence, indicating progressive membrane permeabilization with increasing concentrations of fingolimod. Notably, E. coli exhibited slightly higher PI uptake compared to P. aeruginosa at the highest (40 μM) tested concentrations, suggesting a subtle species-specific preference for E. coli.

Figure 3.

Fingolimod induces membrane permeabilization in E. coli and P. aeruginosa.E. coli and P. aeruginosa cultures were treated with increasing concentrations of fingolimod (0, 10, 20, and 40 μM) for 1 h at 37 °C, followed by incubation with propidium iodide to assess membrane integrity. PI uptake was quantified fluorometrically as an indicator of membrane permeabilization. A concentration-dependent increase in PI fluorescence was observed in both E. coli and P. aeruginosa, indicating enhanced membrane disruption. Bars represent mean ± SD from three biological replicates.

Fingolimod permeabilizes planar lipid membranes by forming lipidic pore-like defects

The fingolimod dose-dependent response in propidium uptake in E. coli and P. aeruginosa (Fig. 3) is consistent with a recent observation that fingolimod can disrupt S. aureus bacterial membrane integrity (7). These results suggest that the antimicrobial activity of fingolimod is associated with bacterial membrane perturbation, but the exact mechanism remains unknown.

To determine whether fingolimod can directly compromise the barrier function of lipid membranes in the absence of grA or other peptides, we performed experiments using planar lipid bilayers of the same lipid composition as in grA experiments, using micromolar drug concentrations. The representative current traces of the fingolimod-induced transient conductance events, obtained at three different concentrations, under application of 60 mV voltage, are shown in Figure 4A. At 4 μM of fingolimod (left panel), the current amplitude is mostly low, ∼2 to 7 pA, suggesting relatively small conductive defects. Increasing the fingolimod concentration to 8 μM (middle panel) resulted in higher current fluctuations up to 200 to 250 pA, whereas 16 μM fingolimod (right panel) induced a large-scale, up to 750 to 800 pA, membrane permeabilization, suggestive of sustained formation of lipidic pore-like defects. This pore formation is a transient and reversible process, with bursts of current (whose amplitude scales with fingolimod concentration) that temporarily disappear and then reappear again in a random manner during the recording time. At a fingolimod concentration > 8 μM, the membrane is inevitably ruptured after some time (5–15 min). Notably, in terms of average conductance and overall noisy character, fingolimod-induced transient permeabilization events compare well with those observed in protein-free lipid membranes under high-voltage stress (28).

Figure 4.

Fingolimod permeabilizes planar lipid membranes at a higher concentration.A, representative current records obtained on the same planar membrane at increasing concentrations of fingolimod as indicated, at 60 mV of applied voltage. At 4 μM fingolimod, transient and small-amplitude pore-like events (<7 pA) were observed, indicative of minor bilayer destabilization. At 8 μM, sustained large-conductance pore formation was evident (∼250 pA), suggesting transient lipidic pore formation. At 16 μM, fingolimod induced high-amplitude, irregular current bursts (∼800 pA), consistent with large-scale membrane permeabilization. Experimental conditions as in Figure 2. B, violin plot showing the distribution of fingolimod-induced conductance at 4, 8, and 16 μM of fingolimod. The width of each violin reflects the probability density of measured conductance values, with white dots marking the median and boxes indicating interquartile ranges in six independent experiments. A progressive increase in both median and range of conductance is evident with increasing drug concentration.

To show the broad range of membrane conductance across different voltages, we compiled single-event data from multiple traces and summarized them in a violin plot (Fig. 4B), revealing concentration-dependent conductance heterogeneity. Collectively, these findings suggest that fingolimod compromises the lipid bilayer integrity through promoting pore-like lipidic defects by disrupting stabilizing interactions between the membrane lipids, rather than through an oligomerization-like mechanism (28, 29, 30). The similarity in conductance, duration, heterogeneity, and stochastic behavior to voltage-induced membrane permeabilization events (28) suggests that the structure of metastable fingolimod-induced pores is not well-defined. Judging by the amplitude of the stepwise transitions seen in Figure 4A and the average conductance data shown in Figure 4B, the “effective radius” of the pore-like defects depends on fingolimod concentration, increasing from several Angstroms to several nanometers, though some kind of synchronized opening/closing of smaller pores organized into clusters cannot be excluded.

Fingolimod binding to the planar lipid membranes studied by Bilayer Overtone Analysis

Bilayer overtone analysis (BOA) is a method that allows direct measurements of the binding kinetics of various compounds to planar lipid membranes (31). BOA provides a real-time readout on the intramembrane potential difference, Ψ, by measuring the second harmonic response to the periodic electrical signal, caused by asymmetrical charge distribution between two lipid leaflets (31, 32, 33, 34). Fingolimod, which carries a net positive charge due to the presence of a primary amine group, is expected to create an asymmetry in the intramembrane potential when bound to one side of the membrane.

Subsequent additions of fingolimod to the cis side of the planar membrane formed from the same lipid mixture used in the experiments with grA, resulted in a stepwise increase in ΔΨ, indicating an asymmetric accumulation of positively charged species on the cis side of the membrane (Fig. 5A). Fingolimod’s cationic headgroup can mediate interactions with anionic lipids 1,2-dioleoyl-sn-glycero-3-phosphoglycerol (DOPG) and cardiolipin (CL), creating an asymmetric charge distribution across the bilayer. Interestingly, the ΔΨ response kept increasing gradually up to 4 μM of fingolimod and sharply declined following fingolimod addition to the 8 μM final concentration (Fig. 5A). This behavior indicates a transition from fingolimod binding to (and staying at) the cis side of the membrane to its translocation to the opposite, trans side bilayer, thus eliminating the charge asymmetry and, consequently, bringing ΔΨ down to zero at > 4 μM of fingolimod. This process was accompanied by a rapid increase in membrane conductance, which was simultaneously measured by BOA (Fig. 5B) on the same membrane. The simultaneous drop in intramembrane potential difference and increase in conductance indicate the translocation of fingolimod to the opposite, trans side of the bilayer through the formation of the transient pores. We call this reaction “autocatalytic exchange” of fingolimod between the opposing membrane leaflets. Indeed, strong attractive interactions between fingolimod and lipids in the leaflet (see results of MD simulations below) do not allow for any measurable exchange. Only the formation of lipidic pores alters the leaflet topology, enabling fingolimod to diffuse to the opposite leaflet. Though pore-like conductive defects may require additional biophysical techniques to be established as the only fingolimod translocation pathways, this mechanistic model is strongly supported by the combination of BOA and conductance data shown in Figure 5.

Figure 5.

Fingolimod membrane binding measured by BOA suggests its autocatalyzed translocation.A and B, representative time courses of the intramembrane potential difference (A) and membrane conductance (B) upon fingolimod addition to the cis side of the same planar membrane made from DOPE/DOPG/CL (0.75:0.2:0.05 w/w) in 0.1 M KCl, pH 7.4. Stepwise additions of fingolimod are shown by black arrows; the final concentrations of fingolimod are indicated. Increasing concentrations of fingolimod produced a progressive positive shift in the relative intramembrane potential difference ΔΨ, with its sharp decrease after 4 μM fingolimod was added. After the subsequent addition of the drug to the 8 μM concentration, the membrane ruptured as indicated by the red arrow. Relative membrane conductance remained stable at ∼ 15 to 25 pS up to 4 μM fingolimod addition and sharply increased after 8 μM fingolimod addition, indicative of pore formation (B). C, average ΔΨ as a function of fingolimod concentration. Data points represent the mean ΔΨ values at each fingolimod concentration, with error bars indicating ± SD for 15 independent experiments. A progressive increase in ΔΨ is observed with increasing fingolimod concentration up to ∼4 μM, followed by a reverse ΔΨ to a baseline at higher fingolimod concentrations, which suggests fingolimod equilibration between the cis and trans lipid leaflets of the membrane. D, schematic cartoons and diagrams of fingolimod-generated potential illustrating fingolimod incorporation into the cis leaflet of the membrane (panel ∗) and its subsequent translocation to the trans leaflet (panel ∗∗). A fingolimod-induced lipidic pore connects the opposing leaflets, allowing diffusion-driven equilibration of the drug between them. This is witnessed by the drop in the intramembrane potential difference measured by BOA. BOA, bilayer overtone analysis. CL, cardiolipin; DOPE, dioleoyl-phosphatidylethanolamine; DOPG, 1,2-dioleoyl-sn-glycero-3-phosphoglycerol.

The events of permeabilization are followed by further destabilization of the membrane and its subsequent rupture (indicated by the red arrows in Fig. 5, A and B). The average binding curve for fingolimod is shown in Figure 5C, demonstrating the transient character of fingolimod-membrane interaction when the binding regime at low fingolimod concentrations is followed by the translocation regime at the increased fingolimod concentration (see Fig. 5D).

Fingolimod-induced formation of pore-like defects is pH-dependent

Studies of gram-positive bacteria by Shang et al. (7), suggested that fingolimod can compromise the structural integrity of bacterial membranes in a pH-dependent manner. Another recent study has demonstrated that CADs tend to accumulate within acidic cell organelles such as lysosomes, as the low pH environment favors the protonation of their amine groups (35).

In turn, we investigated the impact of fingolimod ionization on its membrane-binding behavior using BOA and conductance measurements at acidic (pH 5.5) and basic (pH 9.5) conditions, considering the reported fingolimod pKa ∼8.0. At pH 5.5, when fingolimod is mostly protonated, subsequent additions of drug (0.25–16 μM) to the planar membrane made from the same lipid mixture as in the above planar membrane experiments, induced a concentration-dependent increase in the intramembrane potential difference, reaching over 10 mV after 2 μM of drug addition (Fig. 6A). The reverse of ΔΨ back to zero, followed by addition of 8 and 16 μM fingolimod (Fig. 6A), was accompanied by a sharp increase in ion conductance (Fig. 6C), consistent with membrane destabilization and lipidic pore formation (Fig. 4). In contrast, at pH 9.5, fingolimod, being in its neutral, deprotonated form, influenced neither the ΔΨ (Fig. 6B) nor membrane conductance (Fig. 6D) upon drug addition up to 16 μM, revealing the absence of fingolimod-membrane interaction or perturbation.

Figure 6.

Fingolimod-membrane binding is pH-dependent. Representative time courses of the intermembrane potential difference and conductance changes measured by BOA at different fingolimod concentrations at pH 5.5 (A, C) and pH 9.5 (B, D). A and B, sequential additions of fingolimod (0.25–16 μM) at pH 5.5 produced progressive shifts in the positive ΔΨ in a dose-dependent manner up to 4 μM, followed by progressive reversal in ΔΨ toward a baseline with further additions of fingolimod (A). At pH 9.5, the addition of fingolimod up to 16 μM did not induce a substantial change in ΔΨ (B). C and D, relative membrane conductance increased up to 700 pS after the addition of 8 μM fingolimod at pH 5.5 (C) and remained essentially unchanged at pH 9.5 (D). All other experimental conditions are as in Figure 4. BOA, bilayer overtone analysis.

Fingolimod-membrane interaction studied by MD simulations

To clarify the molecular details of fingolimod’s membrane activity, we used MD simulations. First, MD simulations provided us with the structural features of fingolimod/1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayers (20% fingolimod by mole), as well as the sensitivity of fingolimod to local curvature. The coupling of fingolimod to curvature is critical because the pore structure we hypothesize is highly curved, and the curvature of its edge in the center of the membrane is net negative (concave, Fig. 7A) for small pores. The approach is described in more detail in the supplemental material (Fig. S1). Briefly, the curvature preferred by the lipid is determined by where it dynamically enriches on thermally excited membrane undulations. This spontaneous curvature explains how it supports the finite-sized pores observed in planar lipid bilayers. Pores are routinely simulated in lipid bilayers (36, 37, 38), typically by fixing boundary conditions to prevent collapse. Here, we report the redistribution of fingolimod on a tension-stabilized model pore to determine its affinity for the pore edge; high affinity would be consistent with stabilizing pore expansion beyond the conduction seen in experiments.

Figure 7.

Hypothesized small lipidic pore.A, top view with negative (concave) curvature shown in red, bringing headgroups closer together. B, sideview of positive (convex) curvature indicated in blue, bringing headgroups apart. Interleaflet continuity facilitating exchange is displayed in purple. Fingolimod molecules, substituting lipids, are shown at the pore wall center where negative curvature is dominating for the small-radius pores. All-atom planar membrane simulations: C, the transverse distribution of the fingolimod charged amine (orange) relative to hydrophobic regions (green/blue) and the POPC phosphate (red); (D) 2D (projected) radial distribution functions (RDFs) of the fingolimod amine group and the POPC phosphate group, indicating charged fingolimod has strong interactions with nearby anionic moieties. All-Atom simulation of a large lipidic pore: E, overhead view of a large, tension-stabilized POPC/fingolimod pore; F, radial density of fingolimod and POPC, with fingolimod scaled by the ratio of POPC to fingolimod in the simulation. The increased lipid density is at the pore edge due to the inwardly oriented lipids projected into the plane. Fingolimod is depleted at the positively curved pore edge, consistent with its sensitivity to negative curvature. POPC, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine.

Second, the simulations demonstrated that model fingolimod strongly favors the hydrophobic bilayer. Figure 7C shows the location of the fingolimod charged amine group (orange) relative to the POPC phosphate (red). The fingolimod amine is closer to the bilayer center by ca. An average value of 2 Å, consistent with its analogy to the sphingosine backbone. At no point in the replicas of either the protonated or unprotonated models did fingolimod leave the membrane or flip leaflets in intact bilayers. This finding underscores the importance of the fingolimod-induced transient structural defects in the lipid bilayer (Figs. 4 and 5B) that alter leaflet topology, thereby catalyzing fingolimod inter-leaflet exchange by diffusion (Fig. 5A).

Figure 7D shows the radial distribution functions (RDFs) between the fingolimod amine (N atom) and the anionic phosphate moiety of POPC. RDFs indicate correlated positions between two molecular centers, where strong positive correlations indicate a frequently observed interaction and thus strongly suggest stabilization of a complex. The RDF is shown projected into the XY plane for both protonated (solid line) and unprotonated (dashed line) fingolimod models. The protonated form is strongly enriched at an in-plane distance of 4 Å, consistent with a strong electrostatic interaction between the oppositely charged groups. In contrast, the neutral form is moderately depleted (dashed line). Figure S2 shows the full three-dimensional RDF, meaningful only at short distances, indicating the extreme enrichment of the charge pair.

Third, it was established that fingolimod prefers negative leaflet curvature and is not a pore line-actant. Spontaneous curvature was computed from a simulation using the SPEX method (39, 40) as implemented in the MembraneAnalysis software package (41) (Supporting Information). The spontaneous curvature of both the protonated and unprotonated forms of fingolimod is negative for the model simulated here. Fingolimod interacting with neighboring POPC phosphate groups has spontaneous curvature −0.027 ± 0.002 Å⁻¹ with this interacting form present 81% of the simulation time in the protonated model. In a typical phospholipid bilayer, a pore is energetically unfavorable due to the high curvature of bending lipids from one leaflet to another (Fig. 7, A and B). Radial expansion of the pore increases the amount of bent lipids linearly, as they line the circumference of the pore. A pore line-actant reduces the line tension by stabilizing this interface. While there is substantial negative curvature in small pores (Fig. 7A), the net curvature is positive for large pores, as the convex curvature (Fig. 7B) does not decrease with pore radius. Surfaces can be curved in two directions, such that they have principal curvatures $J_1$ and $J_2$ as shown in Figure 7, A and B. Pores, whether fusion/fission pores or single leaflet pores, have both total curvature ($J_1+J_2$) that can be positive or negative and Gaussian curvature ($J_1\times J_2$) that is always negative. Figure 7E shows an overhead view of a single-bilayer pore created by removing a cylindrical section of a planar membrane and fixing the projected simulation area such that the pore cannot collapse. The leaflet at the pore edge has positive curvature in the radial direction (Fig. 7B), where the leaflet curves from inside to outside. The magnitude of this curvature does not change with pore radius, although the area of the leaflet surface with this curvature increases proportionally to the radius. In the plane of the pore (the red circle in Fig. 7A), curvature is negative. As the pore size decreases, this curvature becomes more negative and can dominate the total local curvature of the lipid in the center of the pore wall. Lipids that prefer positive curvature will enrich at the edges of large pores, consistent with a model of a “line actant” decreasing the effective line tension.

Figure 7F shows the density of fingolimod and POPC in the tension-stabilized pore as a function of distance from the center, with the fingolimod density multiplied by 4.1 so that the average densities are expected to be the same, all else being equal. For this relatively large pore, fingolimod is depleted at the positively curved pore edge, consistent with its preference for negative curvature and contrary to its possible role as a line actant.

Discussion

In this report, we show that fingolimod, an FDA-approved immunomodulatory drug for multiple sclerosis, exhibits antimicrobial activity that is associated with increased membrane permeability in E. coli and P. aeruginosa. To understand the molecular mechanism of this effect, we used several biophysical approaches to probe the interaction between fingolimod and lipid membranes. Our data support the hypothesis that fingolimod not only forms metastable lipidic pores but also, through an autocatalytic mechanism, permeates to the opposite membrane leaflet at sufficiently high concentrations. Using grA channels as biosensors in planar lipid bilayers mimicking the phospholipid composition of E. coli, we demonstrate that fingolimod partitions into membranes and modulate their properties, as reported by dose-dependent changes in grA channel conductance and lifetime (Fig. 2). Importantly, fingolimod also permeabilizes unmodified (that is, not containing grA or any other peptide) membranes in a concentration-dependent manner, inducing transient, relatively small-conductance fluctuations at 4 μM and high-conducting metastable pores of 5 to 15 nS at higher concentrations (8–16 μM) (Fig. 4). Accordingly, BOA measurements (Fig. 5A) revealed that binding of fingolimod to one side of the membrane generates an intramembrane potential difference that increases with fingolimod concentration. However, at higher drug concentrations, the potential difference decreases, indicating fingolimod translocation to the opposite membrane leaflet via the formation of pore-like lipidic membrane defects. These findings are consistent with our hypothesized lipidic pore formation as a molecular mechanism contributing to fingolimod's antimicrobial effects. To our knowledge, this observation is the first report of fingolimod’s ability to form pore-like defects in purely lipidic membranes. These results may explain previously reported antimicrobial activities, synergistic effects with conventional antibiotics, and pH-dependent bactericidal properties of fingolimod (7, 12).

The fingolimod concentrations used in our study span two levels of membrane activity. At sub-micromolar to low-micromolar concentrations, fingolimod altered membrane properties without permeabilization, as shown by grA assays at 0.5 to 2 μM and BOA measurements starting at 0.25 μM (Figs. 2, 5, and 6). Testing at higher concentrations, we observed that fingolimod produced membrane-disruptive effects. In protein-free planar lipid bilayers, 4 μM caused only small transient conductive defects, whereas 8 to 16 μM generated pore-like conductance events and membrane rupture. Similarly, BOA experiments point to permeabilization occurring within this higher concentration regime. In bacteria, this higher concentration range overlapped with increased PI uptake and antimicrobial activity, with growth inhibition observed at 10 μM and complete inhibition at 40 μM (Figs. 1 and 3). These results suggest that low concentrations of fingolimod alter lipid bilayer properties, whereas higher concentrations promote pore-like defects that increase permeabilization and contribute to antimicrobial activity. However, we note that this overlap reflects an association rather than a direct demonstration of causation.

Due to the cationic nature of fingolimod, its membrane activity is expected to depend on protonation state. CADs characteristically partition into lipid bilayers through their hydrophobic scaffold, while protonation of the amine enhances electrostatic interactions with negatively charged lipids and promotes accumulation in acidic compartments (42, 43, 44). This framework is consistent with our finding that fingolimod's membrane activity is pH-dependent: protonated fingolimod forms pores, whereas neutral fingolimod shows no measurable membrane activity (Fig. 6). Such pH dependence may be relevant physiologically, as acidic microenvironments occur in inflamed tissues and within endosomal/lysosomal compartments, where protonation-dependent membrane interactions of CADs are favored (45, 46).

All-atom molecular modeling (Fig. 7) indicated that fingolimod molecules are stable within the bilayer, with no fingolimod molecules leaving the bilayer or flipping leaflets on the microsecond timescale. RDFs indicate that protonated fingolimod strongly interacts with neighboring anionic moieties, leading to enrichment of this configuration relative to neutral fingolimod. It is the charge-paired interacting configuration of fingolimod that strongly prefers negative curvature, whereas noninteracting fingolimod is enriched at positive curvature. The net-negative curvature preference of fingolimod is consistent with its depletion in large, tension-stabilized bilayer pores (which have net positive curvature). Small-radius pores, in which the concave principal curvature is greatest, have net negative curvature and thus the simulations predict such pores would be supported by fingolimod. We note that our simulation results do not yield a small stable pore. This is likely due to the timescale for formation, which we hypothesize requires fluctuations of localized fingolimod to coincide with transient water penetration defects, yet the microsecond scale limitation of the simulations is inadequate to sample these rare events. Fluctuations in the measured current (Fig. 4A) indicate wait times for formation far exceeding the simulation timescale. Simulation results are consistent with our experimental results, indicating small, relatively long-lived pores, in that they predict fingolimod’s support for small-pore negative curvature as well as strong electrostatic structural coupling. Nevertheless, the simulations alone are insufficient to predict the concentration at which fingolimod would form a pore. Protonated fingolimod shows strong interactions with neighboring anionic moieties, forming complexes that prefer negative curvature consistent with a small pore. By changing leaflet topology, that is, establishing continuity between them, a pore allows simple diffusion of fingolimod between the outer and inner leaflets, as observed in the experiments on planar lipid membranes upon 4 μM fingolimod addition. Analogous to mechanisms suggested for pore formation by amphipathic antimicrobial peptides in cell membranes (47), we propose a tentative model where transient membrane defects in which lipid headgroups line the pore interior alongside fingolimod molecules lead to lipidic pore formation (Fig. 7, A and B). These lipidic pores differ fundamentally from protein and peptide-formed channels (48, 49) in that they lack a well-defined protein or peptide scaffold. Instead, they are dynamic, metastable, self-assembled structures stabilized by the curved lipid-water interface.

Membrane permeabilization by fingolimod may enhance the entry of coadministered antibiotics into bacterial cells, improving their efficacy even against resistant strains. This synergy may be particularly relevant for gram-negative bacteria, where the outer membrane serves as a major permeability barrier limiting antibiotic access. Fingolimod-induced pore-like defects could compromise this barrier, sensitizing gram-negative pathogens to antibiotics that would otherwise be excluded. This provides a molecular mechanism linking fingolimod's chemical structure to its antimicrobial activity. Fingolimod’s amphipathic structure, having a flexible hydrophobic tail similar to a fatty acid and an ionizable cationic headgroup, enables fingolimod to insert into lipid bilayers with a preference for curved lipid-water interfaces, mediating pore-like formation. Consistent with this interpretation, the variability across initial CAD screening results (Fig. 1A) indicates membrane activity likely depends not only on cationic charge and lipophilicity, but also on structural features such as molecular geometry and headgroup accessibility. Our screen sampled only 25 compounds and is not sufficient to define broad structure–function principles for CADs. Identifying fingolimod as a particularly potent hit provides only a starting point and structure-activity analyses across a larger panel of analogs will be required to establish how specific molecular features contribute to membrane activity.

Although our planar lipid bilayer studies provide detailed biophysical characterization of fingolimod-membrane interactions, several questions remain. The lipid composition of the model membranes mimics that of the E. coli membrane. However, it does lack the full complexity of bacterial membranes, including proteins, lipopolysaccharides (LPS) (in gram-negative bacteria), peptidoglycan-associated lipids, and membrane microdomains (50, 51, 52, 53). The negatively charged LPS are major components of the outer leaflet of the gram-negative bacterium outer membrane. Prior studies indicate that LPS can serve as both an initial electrostatic binding site for cationic amphiphiles and as a permeability barrier restricting access to the phospholipid bilayer (54, 55), as demonstrated earlier for polymyxin B, a cationic amphiphilic peptide (54). Therefore, we hypothesize that while inner-membrane-mimicking bilayers used in our study are relevant for assessing direct interaction with fingolimod, the presence of LPS in vivo might facilitate initial surface association through electrostatic interactions with negatively charged LPS, while also potentially enhancing the energetic barrier to deeper insertion and subsequent pore formation. Understanding how these components influence fingolimod pore formation in native bacterial membranes will require future work. Our studies are also limited to symmetric bilayers, whereas bacterial membranes exhibit trans-bilayer lipid asymmetry, which may also affect pore formation (56, 57, 58, 59). The PI uptake assay and planar lipid bilayer measurements demonstrate that fingolimod disrupts membrane integrity at micromolar concentrations that also inhibit bacterial growth, but we acknowledge that fingolimod may have additional intracellular targets that contribute independently or synergistically to bacterial killing. Importantly, PI uptake should be interpreted as evidence of increased membrane permeability as this assay cannot distinguish among transient nanoscale defects or secondary permeability changes associated with cell death. Future studies will be needed to more directly establish whether membrane permeabilization is the primary lethal mechanism or one of several contributing factors.

Mammalian membranes, enriched in cholesterol, sphingomyelin, and zwitterionic lipids, are generally more ordered, rigid, and less susceptible to disruption than bacterial membranes enriched in anionic lipids (60, 61). This leaves the question of whether fingolimod selectively targets bacterial membranes over mammalian ones open and will be an interesting subject for future research.

Fingolimod is part of a larger class of CADs that includes tricyclic antidepressants, selective serotonin reuptake inhibitors, and other antipsychotic compounds (3, 4). Fingolimod’s ability to perturb membrane structure and form lipidic pores raises the possibility that pore formation may occur with other CADs that have antimicrobial activity. Future studies are needed to determine whether other CADs exhibit lipid-pore-forming activity and, therefore, represent a generalizable property of CAD subgroups or a unique feature of fingolimod. Understanding the structure-activity relationships that govern membrane interactions and pore-forming activity could inform rational drug design strategies. For human-targeted therapeutics, this knowledge could aid in modifications that retain desired therapeutic activity while minimizing unwanted effects on bacterial membranes. We observed that fingolimod’s antimicrobial and membrane-permeabilizing effects occur at micromolar concentrations. The recommended clinical dose for multiple sclerosis results in plasma levels that are orders of magnitude lower (62, 63). Without additional measurements, our findings should be interpreted primarily as mechanistic insights into how fingolimod interacts with lipid membranes, rather than as evidence that standard therapeutic dosing regimens will produce clinically meaningful antimicrobial effects in vivo. However, as fingolimod is administered orally, it may be possible that the gut concentrations reach micromolar levels similar for some drugs (e.g. NSAIDs such as naproxen) that have been empirically determined (64) and thereby, have the potential to directly act on some gut bacterial species or alter microbiota composition through local membrane interactions. The present work helps establish a biophysical framework for understanding CAD membrane activity more broadly, which may guide the rational design of fingolimod analogs with improved antimicrobial potency at concentrations relevant to clinical or topical applications. Conversely, in antimicrobial drug development, the pore-forming mechanism could be exploited by designing new membrane-targeting antibiotics based on the fingolimod structure or by using pore-forming molecules to enhance the membrane permeability of other antimicrobial agents. Taken together, our results support a model in which fingolimod-induced pore-like membrane permeabilization contributes to its antimicrobial activity.

Experimental procedures

Bacterial growth assays

Overnight cultures of E. coli and P. aeruginosa were diluted to OD₆₀₀ = 0.1 in LB medium. For the drug screening, E. coli was treated with 40 μM of each 25 CADs (Table S1) in 5 ml cultures. For dose-response assays, cultures were treated with 0, 10, and 40 μM fingolimod. Cultures were plated in 96-well plates (200 μl per well), and OD₆₀₀ was measured using a Tecan Infinite 200 Oro Plate reader as described previously (5).

GrA measurements

1,2-dioleoyl-sn-glycero-3-phosphoglycerol (DOPG), dioleoyl-phosphatidylethanolamine (DOPE) and cardiolipin (CL) were purchased from Avanti Polar Lipids, Inc. Alabaster, AL. GrA was generously gifted by O. S. Andersen, Cornell University Medical College. Planar lipid bilayers were formed using DOPE/DOPG/CL (0.75:0.2:0.05 w/w) from the apposition of two monolayers across a ∼125 μm aperture in a 15-μm-thick Teflon partition separating two ∼1.5-mL compartments as described elsewhere (65) after Montal and Mueller (66). Membrane bathing solutions contained 0.1 M KCl buffered with 5 mM Hepes at pH 7.4. GrA dissolved in ethanol solutions was added to both compartments of the chamber. Fingolimod was prepared as a 6.6 mM stock solution dissolved in dimethyl sulfoxide was added to the cis side of the membrane. Single-channel measurements were acquired using an Axopatch 200B amplifier (Axon Instruments, Inc, Foster City, CA, USA) in the voltage clamp mode. Sampling was done at 200 Hz frequency with a low-pass Bessel filter at 0.05 kHz. Later, the records were digitally filtered at 0.1 kHz using the Bessel-8 pole filter using the Clampfit 10.7 software as previously described (65). GrA channel lifetimes and mean conductance were analyzed with the previously reported protocol (65, 67).

BOA measurements

BOA measurements were conducted as described earlier (31) using a Stanford Research Systems 830 lock-in amplifier. The intramembrane potential difference Ψ (defined as the electrical potential drop across the hydrocarbon core of the bilayer) reports on the asymmetry between the two leaflets of the lipid bilayer, including their modification by membrane-bound compounds. In our experiment, Ψ was quantified by analyzing the second harmonic wave of bilayer current in response to an application of sinusoidal voltage across the lipid bilayer (34). Data are calculated as ΔΨ = Ψ − Ψ_{t = 0}, where a positive value of ΔΨ is analogous to the increase of positive charge on the cis side of the membrane. Membrane capacitance and ΔΨ were recorded at one-minute intervals using custom-built software written in Python 3.11 (31). ΔΨ values were averaged over 10 subsequent measurements following a steady-state level, operationally defined by ΔΨ variations within ± 0.5 mV.

PI uptake assay for bacterial membrane permeabilization

Overnight cultures of E. coli and P. aeruginosa were diluted to an OD₆₀₀ of 0.2 in fresh LB medium and grown for 2 h at 37 °C. Cells were then collected at OD₆₀₀ ∼0.3 and incubated with increasing concentrations of fingolimod (0, 10, 20, and 40 μM) for 1 h at 37 °C. Following treatment, cells were washed three times with 10 mM Hepes buffer (pH 7.4) to remove residual drug. For each sample, 100 μl of washed cells were mixed with 100 μl of Hepes buffer containing 2.5 μM PI, a membrane-impermeant DNA-binding dye. Samples were transferred to black-walled 96-well microtiter plates, and fluorescence was measured using a Tecan Infinite 200 Pro plate reader (excitation/emission 535/617 nm).

PI uptake was calculated as.

$$(F_{sample+PI}-F_{sample-PI})/(F_{buffer+PI}-F_{buffer-PI})$$

where F refers to fluorescence intensity.

Membrane conductance measurements in the absence of grA

Planar lipid membranes were formed as described for grA measurements. Data were filtered by a low-pass 8-pole Butterworth filter (Model 900 Frequency Active Filter, Frequency Devices) at 15 kHz, digitized with a sampling frequency of 50 kHz, and analyzed using pClamp 10.7 software (Axon Instrument). For data analysis, digital filtering using a 0.5 kHz low-pass Bessel filter was applied. After membrane was formed, recordings were taken in the absence of fingolimod for ∼30 to 40 min, then fingolimod was added to the cis side. A voltage in the range of 30 to 100 mV was applied to observe membrane permeabilization.

MD simulations

Planar membrane simulations

All simulations were performed at 30 °C using the Amber (22/PMEMD) simulation package (68) run on a single GPU. Simulations were built on the CHARMM-GUI web platform, including the parameterization of fingolimod (using the CHARMM-GUI Ligand Reader & Modeler module using the CGenFF forcefield). Initial relaxation used the standard CHARMM-GUI “equilibration” steps, while dynamics used the recommended parameters consistent with the forcefield and which are included as input files from CHARMM-GUI (version 3.7). Four replicas of both the protonated and unprotonated fingolimod were built and simulated for at least 4 μs. An additional pure POPC bilayer (200 lipids/leaflet) was simulated for 1 μs to compare the area-per-lipid with and without fingolimod (Supporting Information). The neutral surface atom and curvature sensitivity were computed using the MembraneAnalysis Julia package (41). Interacting fingolimod/POPC pairs were identified by first computing the 3D fingolimod-N/POPC-P RDF (Fig. S2), then choosing an upper limit (5 Å) that captured the highly enriched first peak.

Pore simulations

For the pore simulations, a patch of the planar protonated fingolimod/POPC bilayer was removed following at least 50 ns of simulation from each planar replica. With the box area fixed, the pore was unlikely to collapse and so persisted through the simulation, allowing the enrichment or depletion of fingolimod at the pore edge to be quantified. Nevertheless, a pore in one replica collapsed, yielding three metastable pores with lifetimes up to 3 μs. To quantify the distribution of fingolimod in the pore, the bilayer was first centered to place the pore at the origin, then fingolimod and POPC were counted in a histogram starting from the origin, using the neutral surface atom as quantified by Membrane Analysis for the planar simulations above.

Data availability

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

Supporting information

This article contains supporting information (69).

Conflict of interest

This work was prepared while Ekaterina Nestorovich was employed at The Catholic University of America and the National Institute of Child Health and Human Development (NICHD).

Reviewed by members of the JBC Editorial Board. Edited by George M. Carman