R5 and R7 positions on fluoroquinolone scaffolds drive F-actin filament disruption

TL;DR

This study shows how specific chemical modifications on fluoroquinolone antibiotics can disrupt actin filaments, suggesting potential for treating actin-related neurological disorders.

Contribution

The study identifies R5 and R7 positions on fluoroquinolone scaffolds as key for actin disruption and BBB permeability.

Findings

Gen3 and Gen4 fluoroquinolones reduced actin aggregates in over 60% of yeast cells.

Fluoroquinolones alter F-actin thermal stability with minimal structural changes.

R5 amino groups and R7 bulky rings are crucial for actin disruption and BBB permeability.

Abstract

Fluoroquinolones (FQs) are widely prescribed broad-spectrum antibiotics, with newer generations capable of crossing the blood brain barrier (BBB). Previous studies have reported the actin-destabilizing effects of FQs, suggesting their potential for drug repurposing. Actin-associated neuropathies are characterised by the formation of persistent, F-actin aggregates in neurons, which impair critical cellular functions. Therefore, identifying small molecules that can disrupt these actin filaments and aggregates can provide a promising therapeutic strategy. This research aims to map the direct interaction between FQs and F-actin to identify the structural basis for actin disruption. We demonstrated that FQs irreversibly disrupt F-actin filaments in a concentration-dependent manner using scattering-based assay. Electron microscopy and gel filtration confirmed generation-dependent disruption activity. In particular, Gen3 and Gen4 FQs reduced actin aggregates in more than 60% yeast cells. FQ treatment altered the thermal stability of F-actin at both 1:30 and 1:50 molar ratios with minor secondary structural changes. To explore the molecular insights of FQs interaction with F-actin, saturation transfer difference-NMR combined with complementary molecular dynamics simulations revealed the importance of the fluorinated quinolone core, which is common to all FQs. These studies highlight the involvement of an amino group at R5, and bulky piperazine, azabicyclo rings at the R7 position in driving F-actin disruption. We would like to propose that rational modifications at R5 and R7 positions can enhance both actin-disrupting potency and BBB permeability, thereby providing a basis for developing FQs derived therapeutics against actin-related neurodegenerative disorders.

The online version contains supplementary material available at 10.1038/s41598-026-36089-x.

Figures

Structural annotation of fluoroquinolones across generations. (a) Generic quinolone core, (b) Gen1 NDA, (c) Gen2a CFX, (d) Gen2a NFX, (e) Gen2b OFX, (f) Gen2b LFX, (g) Gen3 SFX, (h) Gen4 MFX. Color-coding of functional groups are based on their dominant interaction type over 90 ns MD trajectories: hydrophobic (green), hydrogen-bonding (yellow), and van der Waals (pink). Epitopes of FQs interacting with F-actin are identified using STD-NMR (marked by red asterisks).

F-actin disruption assay using Right Angle Light Scattering (RALS) measurements. (a) Control represents F-actin (3 µM) without SFX treatment. F-actin (3 µM) treated with SFX at various concentrations (3-900 µM) at the indicated time points up to 72 h are represented by different colour bars. Data represents mean and the standard error of mean (mean ± SEM), *n* = 3. (b) Percentage F-actin disruption measured immediately upon treatment (0 h) with increasing molar ratios of the indicated FQs. The dashed lines represent 50% and 90% disruption respectively. Data represents mean and the standard err

Size-exclusion chromatography (SEC) profiles of 30 µM F-actin treated with the indicated fluoroquinolones (FQs) at a 1:15 molar ratio. Elution was monitored by absorbance at 280 nm (mAU). Retention volumes (mL) corresponding to the peak maxima are indicated for each chromatogram.

Transmission electron microscopy (TEM) images of negatively stained actin were acquired to capture filament morphology. (a) 5 µM G-actin control, (b) 5 µM F-actin control, (c) 5 µM F-actin treated with SFX (1:30 molar ratio), (d) 5 µM F-actin treated with SFX (1:50 molar ratio). Images were captured at 10,000× (scale bar: 500 nm), 25,000× (scale bar: 200 nm), and 40,000× (scale bar: 100 nm).

Circular-dichroism spectra of 3 µM F-actin treated with increasing SFX molar ratios at 1:5 (red), 1:10 (cyan), and 1:30 (olive). Untreated F-actin control is shown in grey.

Differential scanning calorimetry (DSC) thermograms of the G-actin (50 µM) and F-actin (50 µM) controls are represented by black and red dashed curves respectively. F-actin (50 µM) treated with (a) 1:30 and (b) 1:50 molar ratios of FQs are represented by baseline corrected distinct solid curves.

Effect of FQ compounds on actin aggregates in yeast cells. (a) Fluorescence microscopy images depicting *∆end3* cells stained with Rh-phalloidin post-treatment with FQ compounds. Scale bar − 5 μm. (b) The bar graphs represent the percentage of cells showing aggregated and diffused/punctate F-actin phenotype upon treatment with FQs. For quantification, 50 random yeast cells were counted from the acquired images of one experiment. Two such separate experiments were considered for performing statistical analysis using a one-way ANOVA (non-parametric) test. Data represent mean ± SEM, *n* = 2 (p <

Saturation-transfer difference (STD) NMR of SFX bound to F-actin. (a) STD spectrum and (b) Reference spectrum (off resonance) of SFX in presence of 30 µM F-actin. The compound was added to 1.5 mM final concentration to result in a 50-fold excess of the SFX in the F-buffer containing 10% D_2_O. The functional groups of SFX involved in binding with F-actin are marked with red asterisks in the molecular structure of the compound.

Surface representation of the pentameric F-actin filament (PDB ID: 8A2T) showing the binding interfaces of fluoroquinolones (FQs). FQs compounds under study (NDA, CFX, NFX, OFX, LFX, SFX, and MFX) are depicted as coloured spheres at their respective binding pockets. To highlight the interface regions where FQ binding occurs, actin subunits are labelled as chains A-E and coloured light blue, yellow, pink, cyan, and grey, respectively.

Interaction dynamics of F-actin in complex with SFX. (a) RMSD plot of the pentameric F-actin–SFX complex with F-actin (black) and SFX ligand (red). (b) Heatmap representation of frequently interacting amino acids across all trajectories. The non-interacting residues carried the score of 0 (white) and the most interacting residues were scored at 1 (dark green). (c) SFX, shown as sticks in cyan, bound at the interface of the pentameric F-actin represented as a cartoon with chain A (light blue), chain B (yellow), and chain C (pink). (d) Amino acid residues of F-actin (shown as sticks) interacting

Full text

Introduction

Neurodegeneration is characterised by the progressive loss of neurons and abnormalities in neuronal synapses often appearing in middle or later stages of life^1,2^. The neurodegenerative disorders (NDs), which impact over 50 million people worldwide, are often linked to synaptic dysfunction, disruptions in neural networks, and the accumulation of abnormal protein variants in the brain^2,3^. Common NDs include Alzheimer’s disease, Parkinson’s disease, Amyotrophic lateral sclerosis (ALS), motor neuron disease, Huntington’s disease, spinal muscular atrophy, prion diseases, and spinocerebellar ataxia^3–7^.

Neurons naturally harbour dynamic actin filaments and under certain conditions can assemble them into rod-like aggregates composed primarily of actin and cofilin^8^. When exposed to oxidative or energetic stressors, those reactive molecules disrupt the regulation of actin and its associated proteins and impair filament polymerisation dynamics that are essential for maintaining dendritic spine structure and synaptic plasticity^9,10^. As polymerisation stalls, actin and cofilin coalesce into distinctive rods. In the short term these actin-cofilin rods help protect neurons by delaying cytochrome C release from mitochondria and temporarily staving off apoptosis^11,12^. However, persistent rods obstruct intracellular transport and serve as nucleation sites for amyloid precursor protein (APP) and tau accumulation, which accelerates fibril formation^13^. Such rods have been identified in the brains of patients with Guam amyotrophic lateral sclerosis parkinsonism dementia complex, Alzheimer’s disease, and Pick’s disease^14–16^. Over time these aggregates mature into paracrystalline Hirano bodies, which contribute to neurodegeneration. As evident in transgenic models, these Hirano bodies not only serve as early markers of disease but also intriguing targets for therapeutic intervention^17–19^.

A number of marine macrolides, such as reidispongiolides, sphinxolides, aplyronines, and ulapualides bind to the barbed end of actin and disrupt actin filament, however, no actin-binding drugs have broken beyond the preclinical stage due to their extreme cytotoxicity^20^. Therefore, it is pertinent to identify the potential small molecules that could reverse aberrant actin aggregation and disrupt the actin-cofilin rods into smaller soluble forms without being cytotoxic to the cells. Pathak et al. showed that members of the tetracycline family, along with the second-generation fluoroquinolone ofloxacin, can disrupt actin aggregates and thus apart from being good antibiotics these compounds may also be repurposed for actin disruption^21,22^. Likewise, for FDA-approved Rifampicin and Colchicine actin disruption ability has been reported^23,24^. However, their poor permeability across Blood Brain Barrier (BBB) renders them unsuitable for treating actin mediated neuropathies^25,26^. As reported by Kloskowski et al., FQs were found to degrade F-actin stress fibres in various cell-lines^27^. FQs with moderate lipophilic nature, absence of charge at physiological pH, and low plasma protein binding capacity that favours blood-brain barrier penetration makes them suitable candidates for treating brain actinopathies^28,29^. Although FQ-induced actin destabilisation has been reported, detailed insights into the structure–function–activity relationship remain largely unexplored. We hypothesise that FQs interact directly with F-actin and mapping these interactions will reveal structural features that govern their ability to disrupt actin. These insights will provide a rationale for repurposing specific FQs as therapeutic candidates for actin-driven neuropathies.

To fully characterise the applicability of FQ’s for actin-driven neuropathies, this study investigates the effect of five generations (Gen) of FQs (Gen1, Gen2a, Gen2b, Gen3 and Gen4) on F-actin disruption using an interdisciplinary in vitro and in silico approach. FQs-mediated disruption of F-actin was assessed using scattering assays, further confirmed by electron microscopy, gel filtration and studies in the Saccharomyces cerevisiae ∆end3 strain. The secondary structure and thermal stability of FQ-treated F-actin samples were assessed using circular dichroism (CD) and differential scanning calorimetry (DSC) respectively. Using Saturation Transfer Difference (STD NMR) and molecular dynamics simulations (MD), we identified critical functional groups of the FQs interacting with F-actin and characterised the nature of these interactions. This enables us to suggest potential modifications to the FQs structure for improving the potency of these FDA-approved drug molecules. Moxifloxacin (MFX) and Sparfloxacin (SFX) stand out for their efficacy in disrupting actin filaments and may be repurposed for actinopathies inside and outside the brain respectively.

Results

We studied the effects of five generations of FQs (Gen1: Nalidixic acid (NDA); Gen2a: Ciprofloxacin (CFX) and Norfloxacin (NFX); Gen2b: Ofloxacin (OFX) and Levofloxacin (LFX); Gen3: Sparfloxacin (SFX); Gen4: Moxifloxacin (MFX)) on F-actin using a plethora of spectroscopic techniques.

Fluoroquinolones efficiently and irreversibly disrupt F-actin

Right Angle Light Scattering (RALS) was used to measure the F-actin disruption efficacy of the aforementioned FQs. A control sample with untreated 3 µM F-actin in the F-buffer, exhibited a high level of scattering exceeding 100 a.u. All FQs began disrupting F-actin at a 1:15 molar ratio, with concentration dependent increase in efficacy (Fig. 1 and Fig. S1). While NDA showed good efficacy at 1:100 molar ratio, Gen2a FQs required higher concentrations of 1:200 to achieve substantial activity (Fig. S1a, b). In contrast, Gen2b FQs were more effective even at 1:60 molar ratio (Fig. S1c, d). SFX, a Gen3 FQ exhibited marked F-actin disruption at a 1:15 molar ratio, with a reduction in scattering intensity exceeding 50%, indicating its effectiveness (Fig. 1a). MFX, a Gen4 FQ was the most effective F-actin disruptor among all FQs, achieving complete disruption at a 1:30 molar ratio (Fig. 1b) (Fig. S1f). We do not observe any reversibility in the treated samples, i.e., reformation of F-actin even after 72 h in case of NDA (Fig. S1e).

Fig. 1F-actin disruption assay using Right Angle Light Scattering (RALS) measurements. (a) Control represents F-actin (3 µM) without SFX treatment. F-actin (3 µM) treated with SFX at various concentrations (3-900 µM) at the indicated time points up to 72 h are represented by different colour bars. Data represents mean and the standard error of mean (mean ± SEM), n = 3. (b) Percentage F-actin disruption measured immediately upon treatment (0 h) with increasing molar ratios of the indicated FQs. The dashed lines represent 50% and 90% disruption respectively. Data represents mean and the standard error of mean (mean ± SEM), n = 3.

Heterogeneity of disrupted F-actin oligomers

To assess the extent of F-actin disruption observed in RALS and TEM experiments, untreated and FQ-treated F-actin samples were analysed using gel filtration chromatography. Untreated F-actin eluted as a broad peak at 9.2 mL, near the column’s void volume, which is consistent with the large filamentous structure (Fig. 2). F-actin treated with Gen1 NDA exhibited a sharper elution peak at 15.6 mL, indicating a more homogeneous population of disrupted filaments compared to other FQ-treated samples. Notably, treatment with Gen2a CFX and NFX, resulted in the formation of visible white precipitate thereby preventing their analysis by gel filtration chromatography. Gen2b treated F-actin began eluting near the void volume, with peak maxima at 21.6 mL for OFX and 21.3 mL for LFX, suggesting marked filament disruption. Similarly, Gen3 SFX and Gen4 MFX treated F-actin exhibited elution profiles comparable to those of Gen2b compounds, with elution starting from the void volume and peak maxima at 22.4 mL for SFX and 22.2 mL for MFX. The diversity in elution profiles suggests heterogeneity of actin oligomer populations among the treated samples, reflecting the plausible differences in their molecular disruption mechanisms.

Fig. 2. Size-exclusion chromatography (SEC) profiles of 30 µM F-actin treated with the indicated fluoroquinolones (FQs) at a 1:15 molar ratio. Elution was monitored by absorbance at 280 nm (mAU). Retention volumes (mL) corresponding to the peak maxima are indicated for each chromatogram.

Morphological changes in disrupted filaments

We used TEM imaging to characterise the morphologies of the disrupted F-actin filaments. Actin exhibits distinct morphologies when negatively stained in G-buffer and F-buffer. G-actin primarily consists of actin monomers and oligomers (Fig. 3a), while polymerised F-actin is a filamentous structure in solution (Fig. 3b). F-actin treated with Gen1 NDA at 1:30 and 1:50 molar ratios resulted in the disruption of filaments, with larger oligomers predominating in the solution (Fig. S2). In contrast, treatment with Gen2a CFX and NFX at the same molar ratios (1:30 and 1:50) produced perturbed filaments with distinct morphologies compared to the control F-actin. These perturbed filaments showed multiple kinks, bends, and discontinuities, though complete filament disruption was not observed (Fig. S3, S4). Similarly, Gen2b LFX and OFX treatment at a 1:30 molar ratio also resulted in perturbed F-actin morphology, similar to Gen2a treated samples, with kinks, bends, and oligomers present, but without complete filament disruption. However, at a higher molar ratio of 1:50, OFX treatment resulted in the complete loss of filamentous structures (Fig. S5). Surprisingly, LFX, despite being from the same generation, did not show this effect (Fig. S6). Treatment with Gen3 SFX and Gen4 MFX led to complete disruption of filaments at both 1:30 and 1:50 molar ratios (Fig. 3c, d, and Fig. S7). Furthermore, the oligomers generated post-filament disruption were predominantly smaller in size in comparison to earlier generations of FQs.

Fig. 3. Transmission electron microscopy (TEM) images of negatively stained actin were acquired to capture filament morphology. (a) 5 µM G-actin control, (b) 5 µM F-actin control, (c) 5 µM F-actin treated with SFX (1:30 molar ratio), (d) 5 µM F-actin treated with SFX (1:50 molar ratio). Images were captured at 10,000× (scale bar: 500 nm), 25,000× (scale bar: 200 nm), and 40,000× (scale bar: 100 nm).

Fluoroquinolones induce minor secondary structural changes in actin

Treatment of F-actin with FQs disrupts the filaments as observed in RALS, SEC and TEM measurements. To confirm the possible secondary structural changes of the resulting populations of disrupted F-actin filaments, far UV CD spectroscopic measurements were performed. The two negative bands at 208 nm and 222 nm are characteristic of α-helical structures while the negative band at 218 nm is associated with β-sheet structures^30^. Actin is a rather rigid protein molecule and has secondary structure conformations of 30% α-helix, 23.5% β-sheet, 12.2% turns, and 34.3% random coil. These values closely match the previously reported secondary structure composition of actin protein^31^.

Upon treatment of F-actin at a 1:30 molar ratio, Gen1-NDA increased the α-helical content by 4.1% and decreased the β-sheet content by 2.6% (Fig. S8e). Gen2a CFX treatment of F-actin decreased the α-helical content by 3% and increased the β-sheet content by 3.5% with minor changes in the turns and other conformations such as random coils (Fig. S8a). Gen2a NFX treatment of F-actin increased the α-helical content by 2.2% and decreased the β-sheet content by 4.2% (Fig. S8b). Gen2b OFX treatment of F-actin decreased the α-helical content by 4.6% and increased the β-sheet content by 3% (Fig. S8c). Similarly, Gen2b LFX treatment of F-actin decreased the α-helical content by 7.6% and increased the β-sheet content by 4.8% (Fig. S8d). Gen3 SFX treatment of F-actin induced only minor change in the α-helical content but decreased the β-sheet content by 1.3% with only marginal alterations in turns and random-coils (Fig. 4). Gen4 MFX treatment of F-actin induced minor change in the α-helical content but decreased the β-sheet content by 3.9% leaving turns and random-coil fractions virtually intact (Fig. S8f). The complete secondary structure data are provided in Supplementary Table S1.

Fig. 4. Circular-dichroism spectra of 3 µM F-actin treated with increasing SFX molar ratios at 1:5 (red), 1:10 (cyan), and 1:30 (olive). Untreated F-actin control is shown in grey.

Thermal stability of fluoroquinolones treated F-actin samples

The primary goal of DSC measurements was to assess the thermal stability of F-actin filaments treated with different generations of FQs. In the case of thermal denaturation of biological samples, a difference of 1 °C in melting temperature (T_m_) is considered significant^32^. To ensure that the observed changes in thermal parameters were due to drug effects and not concentration-dependent artifacts, the actin concentration was kept constant at 50 µM for all DSC experiments. The denaturation peaks of G-actin (T_m_ = 59.0 °C) and F-actin (T_m_ = 69.8 °C) are consistent with previously reported literature values^33^. After FQ treatment the F-actin undergoes disruption as observed in RALS, SEC, and TEM experiments. Therefore, we expect FQ mediated F-actin disruption should also destabilise protein structure. The disrupted oligomers should be more globular compared to the compact F-actin structure thus reducing the melting temperature of F-actin treated with FQs (Fig. 5). After the FQ treatment at 1:30 and 1:50 molar ratios, a marked reduction in T_m_ and ΔH_cal_ of treated F-actin was observed (Table 1).

Table 1. Thermal parameters of the denaturation of native (both G- and F-actin) and FQ-treated F-actin.T_m_ (°C)ΔH_cal_ (kJ.mol^− 1^)T_m_ (°C)ΔH_cal_ (kJ.mol^− 1^)G-actin (native)59.0480.9 ± 48.9F-actin (native)69.81012.0 ± 48.3F-actin treatedwith FQs1:30 molar ratio1:50 molar ratioNDA66.8591.5 ± 23.261.5445.4 ± 9.3CFX65.6666.4 ± 21.565.4620.8 ± 15.0NFX67.5771.8 ± 28.867.0690.6 ± 28.7OFX68.1714.3 ± 29.665.7688.3 ± 21.7LFX68.0617.4 ± 22.868.3753.0 ± 26.2SFX66.9667.1 ± 21.964.1619.7 ± 19.6MFX69.1837.6 ± 40.568.2793.3 ± 30.7

Fig. 5. Differential scanning calorimetry (DSC) thermograms of the G-actin (50 µM) and F-actin (50 µM) controls are represented by black and red dashed curves respectively. F-actin (50 µM) treated with (a) 1:30 and (b) 1:50 molar ratios of FQs are represented by baseline corrected distinct solid curves.

Fluoroquinolone mediated actin disaggregation in Δend3 S. cerevisiae

In vivo effects of FQs on actin bundles were investigated in the S. cerevisiae ∆end3 strain. This strain has been widely used in studies involving actin-binding small molecules as the F-actin dynamics are reduced during the stationary phase of growth and their larger cell size facilitates visualisation of actin aggregates^34^. After the cells reached the stationary phase, they were treated with 300 µM FQ compounds and stained with Rh-phalloidin to assess morphological changes in actin aggregates using fluorescence microscopy.

In untreated control cells, large aggregated patches of F-actin were observed whereas FQ-treated cells exhibited diffused/punctate F-actin morphology (Fig. 6a). 81% of the untreated yeast cells displayed aggregated F-actin patches (area ranging between 155 and 530 nm^2^), while only 19% showed diffuse/punctate morphology (area ranging ≤ 90 nm^2^, Fig. 6b). Upon FQ treatment, the F-actin distribution was significantly altered. In cells treated with the Gen1 NDA, 31% cells retained the aggregated actin, whereas a significantly greater proportion of cells (69%) exhibited diffused/punctate F-actin. Similarly, treatment with Gen3 SFX and Gen4 MFX resulted in a marked increase in the proportion of cells (67% and 63% respectively) exhibiting diffused/punctate F-actin compared to the control. Cells treated with Gen2 compounds NFX, OFX, and LFX also showed a significantly higher reduction in aggregates than the untreated group (62%, 59% and 58% respectively) (Fig. 6b). In CFX-treated group, the percentage of cells with aggregated F-actin remained relatively higher (63%) compared to the other FQ-treated groups. Nevertheless, CFX treatment still led to a higher proportion of cells (37%) displaying diffused/punctate F-actin compared to the untreated control (19%) (Fig. 6b).

Fig. 6. Effect of FQ compounds on actin aggregates in yeast cells. (a) Fluorescence microscopy images depicting ∆end3 cells stained with Rh-phalloidin post-treatment with FQ compounds. Scale bar − 5 μm. (b) The bar graphs represent the percentage of cells showing aggregated and diffused/punctate F-actin phenotype upon treatment with FQs. For quantification, 50 random yeast cells were counted from the acquired images of one experiment. Two such separate experiments were considered for performing statistical analysis using a one-way ANOVA (non-parametric) test. Data represent mean ± SEM, n = 2 (p < 0.05, p < 0.01, *p < 0.001, ***p < 0.0001).

Differential binding of fluoroquinolones to actin using STD NMR

STD NMR was used to study the interaction between actin protein and FQs at a 1:50 molar excess ratio. The actin-FQ mixtures contained 10% D₂O in the F-buffer at 25 °C for acquiring on-resonance and off-resonance spectra. The presence of signals in the STD spectrum provided clear evidence of FQs binding to actin protein. The binding epitopes of the FQ compounds (proton-containing functional groups) were identified by comparing the STD spectra with the corresponding ¹H NMR spectra of the compounds, thereby providing atomic-level insights into their interaction with the actin. STD NMR spectra for all FQs exhibited signals, indicating that each compound readily interacted with the F-actin. The characteristic signals of protons from the aromatic quinolone bicyclic core, methyl (-CH₃), and methylene (-CH₂-) groups of the FQs were prominent. Due to rapid exchange with water protons, signals from the OH and NH functional groups were not readily detectable.

Group epitope mapping studies using STD NMR

Gen3 SFX exhibited stronger STD intensities than the other FQs. The H12 and H15 protons from the methyl groups attached to the piperazine ring showed a 100% STD effect, while H10 and H14 protons from the same ring contributed 7.0% (Fig. 7). The H17 and H18 protons from the cyclopropyl group exhibited a moderately weak STD effect of 17.6%. NH₂ group (4.2%) protons attached to the quinolone core and the proton from the NH group (7.3%) of the piperazine ring showed weak STD effects. Putative NH signals in the STD NMR spectrum were unexpected; therefore, these signals were confirmed by acquiring and comparing spectra of SFX in pure D₂O. These signals are weak and typically not visible in aqueous solutions due to rapid exchange with water. The only aromatic proton, H3 (6.7%), exhibited a weak STD effect. Signals in the STD spectrum indicate that, in conjunction with the quinolone core, piperazine ring also contributes to the binding interactions between SFX and actin protein.

Figure S9 shows Gen1 NDA in complex with actin protein. H12 protons from the methyl group, had a 100% STD effect. STD intensities of other signals in the spectrum were calculated relative to this strongest signal by comparing their intensities. Methyl H9 protons showed a moderate STD effect of 20.9%, whereas the aromatic protons from the naphthyridine core, such as H3 (7.1%), H6 (4.2%), and H8 (4.8%), showed very weak STD effects.

Fig. 7. Saturation-transfer difference (STD) NMR of SFX bound to F-actin. (a) STD spectrum and (b) Reference spectrum (off resonance) of SFX in presence of 30 µM F-actin. The compound was added to 1.5 mM final concentration to result in a 50-fold excess of the SFX in the F-buffer containing 10% D_2_O. The functional groups of SFX involved in binding with F-actin are marked with red asterisks in the molecular structure of the compound.

Gen2a CFX and NFX displayed weak overall STD effects. These were also the consistently weakest binders among all FQs tested in other assays. For CFX, the H15 and H16 protons from the cyclopropyl group showed the highest STD effects at 100% and 91.3%, respectively. Aromatic protons from the quinolone core, such as H2 (21.9%), H8 (9.8%), and H4 (11.5%), exhibited moderate to weak STD effects (Fig. S10). NFX, which belongs to the same generation, showed only four signals in the STD NMR spectrum, confirming its lower binding activity as observed in earlier data. H15 proton from the methyl group exhibited a 100% STD effect, while the aromatic protons from the quinolone core, such as H8 (12.2%), H6 (10.0%), and H3 (11.7%), showed weak STD effects similar to CFX (Fig. S11).

Gen2b OFX and LFX showed 100% STD effect for the H10 methyl proton. H15 protons from the methyl group attached to the piperazine ring showed a moderate STD effect of 45.7% in OFX and 40.7% in LFX, indicating good contact with actin protein and participation in binding interactions. The aromatic protons from the quinolone core of OFX exhibited weak STD effects, as indicated by H8 (13.0%), H9 (6.9%), and H1 (11.7%) (Fig. S12). Similarly, LFX showed weak STD effects for H8 (9.2%), H9 (6.7%), and H1 (9.7%) (Fig. S13).

Gen4 MFX also exhibited strong STD effects but followed a binding mode similar to CFX. The H19 and H20 protons from the cyclopropyl group contributed the highest STD effects at 100%. Protons from azabicyclo group, such as H10 and H16, together exhibited a strong STD effect of 79%. Additionally, NH protons (34.1%) and H14 (48.4%) showed moderate STD effects, indicating the prominent contribution of the azabicyclo group to actin binding. Aromatic protons from the quinolone core, such as H6 (32.6%) and H7 (18.8%), displayed moderate to weak STD effects (Fig. S14). Based on the number of signals and their STD effects in the spectrum, MFX demonstrates strong binding to actin protein.

Atomic level structural insights into binding of fluoroquinolones to F-actin

The binding poses and type of interaction of FQs with a pentameric F-actin model were characterised using molecular docking (Fig. 8, Fig. S21, Supplementary Table S2) and further assessed for stability and dynamics by independent molecular dynamics (MD) simulations.

Fig. 8. Surface representation of the pentameric F-actin filament (PDB ID: 8A2T) showing the binding interfaces of fluoroquinolones (FQs). FQs compounds under study (NDA, CFX, NFX, OFX, LFX, SFX, and MFX) are depicted as coloured spheres at their respective binding pockets. To highlight the interface regions where FQ binding occurs, actin subunits are labelled as chains A-E and coloured light blue, yellow, pink, cyan, and grey, respectively.

In the MD simulations, both the Gen2a CFX and NFX deviated away from the docked sites. CFX exhibited a steadily increasing RMSD (Fig. S16a), reaching over 5.0 nm by the end of simulation over the period of ~ 38 ns (Supplementary movie SM1). This progressive deviation suggests major conformational and spatial rearrangements of CFX, leading to its “diffusion away” from the initial binding site on the protein, implying an unstable interaction with relatively stable F-actin over time. On the other hand, Gen2a NFX did not bind at the interface and was observed to interact with an ATP/ADP binding site (Fig. S17a). Gen2b OFX (Fig. S18a), showed a similar outcome in one of the replicas. Along with CFX and NFX, it was excluded from further investigation, as these compounds were categorised as either weak binders or ligands requiring refinement in future docking studies.

Gen1 NDA, Gen2b LFX, Gen 3 SFX, and Gen4 MFX maintained low and stable RMSD values (~ 0.2–0.6 nm) over ~ 90ns in the three replicas (Fig. S15a, Fig. S19a, Fig. 9a, and Fig. S20a, Supplementary movie SM2, SM6, SM7, and SM8), with minimal deviation from the initial poses, with negligible structural perturbation to F-actin.

Gen3 SFX was found to bind at the multimeric interface of chain A, B and C (Fig. 9c). The frequently interacting amino acid residues from different subunits of F-actin with SFX are shown in Fig. 9b. The prominent residues of the interface such as Phe266, Pro172, Ile267, Lys191 and Ile267 form strong and persistent contact (dark green). Additionally, the other interface-forming residues Tyr188, Phe375, His173, Gly268, Ile175, Arg256, Ser265, Cys374, Leu110, Ile192, and His40 interact with SFX, as visualised in Fig. 9d. To identify the highly interacting functional groups of SFX with F-actin, atom-specific contact frequencies were analysed over the full MD trajectory (total interactions: 838,311). The analysis revealed that atoms N4, O1, O2, O3, and F2, located on the quinolone core of SFX (as mapped in Fig. 9e), exhibited the highest interaction frequencies compared to other functional groups (Fig. 9f).

Similarly, interaction analysis has been performed onto all FQs- CFX (Fig. S16), NFX (Fig. S17), OFX (Fig. S18), LFX (Fig. S19), MFX (Fig. S20), including NDA (Fig. S15), respectively. Gen1 NDA interacted with the interface formed by chain C and D (Fig. S16c). Prominent interacting amino acids included Pro172, Leu110, Phe375, Pro109, Lys113, Asn111, Arg116, Cys374 of chain C and Glu195, Lys191, Tyr188, Ile192, Ile267, Arg256, Phe266 of chain D. These interactions remained consistent across the three replicas (Fig. S16b and Fig. S16d). Atoms O2, O3, O1 and C6 (Fig. S16e) of the quinolone core showed the highest interaction frequencies with actin.

Gen4 MFX was found to bind at the interface of chain C and D (Fig. S20c). Frequently interacting amino acid residues with MFX are shown as the dark green patch in the heatmap (Fig. S20b). Prominent residues interacting with MFX, are Lys284, Asn280, Ile175, His173, Met176, Arg177 of chain C as well as Lys191, Asp187, Ile267, Thr194, Phe200, Met190 of chain D, as visualised in Fig. S20d. The total number of interactions between MFX and F-actin during the simulation was calculated to be 555,792. The most frequently interacting atoms were O1, O3, O4, and F of the quinolone core (Fig. S20f).

Fig. 9. Interaction dynamics of F-actin in complex with SFX. (a) RMSD plot of the pentameric F-actin–SFX complex with F-actin (black) and SFX ligand (red). (b) Heatmap representation of frequently interacting amino acids across all trajectories. The non-interacting residues carried the score of 0 (white) and the most interacting residues were scored at 1 (dark green). (c) SFX, shown as sticks in cyan, bound at the interface of the pentameric F-actin represented as a cartoon with chain A (light blue), chain B (yellow), and chain C (pink). (d) Amino acid residues of F-actin (shown as sticks) interacting with SFX as seen in the heatmap. (e) 2D annotated structure of SFX. (f) Bar graph representing top 10 frequently interacting atoms of SFX (all trajectories) with F-actin.

Discussion

Actin-associated neuropathies are characterised by persistent F-actin aggregates in neurons, which impairs essential cellular functions, making their disruption a promising therapeutic approach. Since the development of new therapeutic compounds is an immense challenge, repurposing FDA-approved drugs provides an attractive alternative strategy. FQs have previously been reported as potential candidates to destabilise actin, yet their structure–function–activity relationship remains undefined. This study investigates the direct interaction of multigeneration FQs with F-actin to elucidate the structural basis of filament and aggregate disruption. Our assays were designed to evaluate whether FQs can actively disrupt stable, pre-formed F-actin filaments, rather than merely inhibit polymerisation or act under depolymerising conditions. This filament disruption in a stabilising environment provides stronger evidence for the potent disruptive activity of these compounds. While our purified F-actin system attempts to simulate the conditions, it does not completely replicate the complex actin-cofilin aggregates observed in neuro-actinopathies. To address this, we adopted a complementary experimental design wherein in vitro F-actin disruption assays using stable, purified filaments under polymerising conditions were employed to establish the direct interaction of FQs with F-actin. Using the yeast Δend3 cellular model, we demonstrated that FQs effectively resolve pre-existing aggregates in a cellular context.

Our in-vitro biophysical RALS measurements confirm that Gen4 MFX exhibited the most potent activity, effectively disrupting F-actin at an equimolar ratio. This was followed by Gen3 SFX, which required a fivefold molar excess. Interestingly, the smallest molecule, Gen1 NDA, showed moderate disruption activity. Notably, F-actin disruption induced by all FQs was found to be irreversible, with the exception of NDA. At lower concentrations, NDA-treated samples showed increased scattering intensity at 72 h (Fig. S1e), suggesting that insufficient drug concentration may permit re-bundling of smaller F-actin fragments, ruling it out as a potential candidate for further clinical investigation. Gen2b (OFX, LFX) displayed weak activity, while Gen2a (NFX, CFX) were the least effective, with CFX requiring a sixty-fold excess to induce observable disruption which is clearly less potent than MFX and SFX. Despite differences in potency, all FQs maintained smaller actin fragments post-disruption for at least 72 h, supporting their irreversible mechanism of action. The clear shift of F-actin towards lower molecular-weight oligomers, confirmed effective filament disruption in the gel filtration profiles for Gen2b, Gen3 and Gen4 FQs. These profiles exhibited broad peaks, reflecting the heterogeneity of the fragmented actin population. In contrast, Gen1 NDA yielded larger, more homogenous actin fragments, suggesting a different mode or efficiency of disruption.

TEM analysis further provides morphological validation of F-actin disruption. Gen1 NDA, Gen3 SFX, and Gen4 MFX caused complete filament fragmentation at both thirty- and fifty-fold drug excess. Gen2b OFX disrupted filaments into larger oligomers, whereas Gen2a CFX and NFX, and Gen2b LFX, only showed minimal morphological changes/structural perturbations, even at higher concentrations. Polymerisation-driven stabilisation increases the F-actin melting temperature by ~ 11 °C when compared to the G-actin^33^. Interestingly, treatment of F-actin with all the FQs, shifted the melting profile towards that of G-actin, confirming disruption of filament integrity to a varying extent (Fig. 5). Despite F-actin disruption upon treatment with FQs, only minor changes in α-helical, β-sheet, and random coil content are seen in CD spectra. This indicates that the FQs do not alter the secondary structure of actin.

Isothermal titration calorimetry (ITC) failed to yield conclusive binding interactions (data not shown), indicating a weak binding between FQs and F-actin. Due to this weak interaction, Saturation Transfer Difference (STD) NMR was employed to identify the key interacting protons of FQs. STD NMR highlights the protons involved in hydrophobic and van der Waals interaction. Protons at position 2 and 5 of the core quinolone rings (Fig. 10a) were observed to be consistently interacting with F-Actin among all FQs, thereby emphasizing their vital role in drug binding. These findings were corroborated by MD simulation studies, which showed that position 2 engages in van der Waals interactions, whereas position 5 forms hydrophobic contacts (Fig. 10). NDA exhibited additional interactions via proton at position 6, which is absent in other FQs due to fluorine substitution (Fig. 9b). CFX and NFX displayed weak STD signals and unstable MD trajectories, consistent with their poor experimental performance. In contrast, SFX and MFX showed strong and sustained interactions, attributed to bulky ‘R7’ substitutions such as dimethyl piperazine (SFX) and azabicyclo ring (MFX) (Fig. 10g, h). It is important to note that due to the large size of the F-actin polymer, the protein’s signals are not detectable in STD-NMR experiments. Instead, this technique excels at identifying the binding epitopes of the ligand-that is, which parts of the FQ molecule make direct contact with the protein. Our STD-NMR and ITC data suggest that FQs bind weakly to the F-actin filament. Such weaker interactions compared to those of classical actin-binding toxins may actually be advantageous. Strong binders often cause irreversible cytotoxicity, which has hindered their therapeutic application. In contrast, FQs retain the ability to destabilise actin filaments despite weaker interactions, indicating minimal off-target toxicity.

All FQs, except NFX (Supplementary Movie 4 NFX), bind to the interface of F-Actin filaments as observed in Fig. 8. In corroboration to our experimental findings, the potent F-actin disruptors that are SFX and NDA exhibit lower standard deviation (Fig. S22b, Supplementary Movie 7 SFX, Supplementary Movie 2 NDA) across all the three replicas which depicts stable binding with the filamentous actin and hence induce notable structural deviations in actin polymer (Fig. S22a). On the other hand, the experimentally less effective OFX, LFX and CFX bind to F-Actin for smaller duration (with higher standard deviation), thereby tend to get dislodged (Fig. S22b, Supplementary Movie 5 OFX, Supplementary Movie 6 LFX, Supplementary Movie 3 CFX) from the complex resulting minor deviations in the actin filament (Fig. S22a, Supplementary Movie 1 F-actin).

Fig. 10. Structural annotation of fluoroquinolones across generations. (a) Generic quinolone core, (b) Gen1 NDA, (c) Gen2a CFX, (d) Gen2a NFX, (e) Gen2b OFX, (f) Gen2b LFX, (g) Gen3 SFX, (h) Gen4 MFX. Color-coding of functional groups are based on their dominant interaction type over 90 ns MD trajectories: hydrophobic (green), hydrogen-bonding (yellow), and van der Waals (pink). Epitopes of FQs interacting with F-actin are identified using STD-NMR (marked by red asterisks).

Further, analysis of the Molecular Dynamics Simulation data captured the important amino acids of actin involved in interactions with FQ molecules. Lys191, Leu110, Ile175, and Asp187 are the key amino acids mediating the interaction with FQs, irrespective of the chains it binds. These residues form the conserved core of the FQ binding pocket, providing stable polar and hydrophobic contacts across all ligands. As depicted in Fig. 8, despite MFX, NDA, OFX and SFX binding at different regions or chains of F-actin, His173 is consistently involved in interactions with all four compounds, suggesting its vital role in F-actin filament disruption. Gen1 NDA and Gen3 SFX were observed to be effective disruptors when compared to the other FQs as both interact with Arg256 of F-actin (PDB ID: 8A2T). Arg257 and Glu196 in the γ-actin isoform are the direct sequential and functional equivalents of Arg256 and Glu195 mature α-actin structure^35^. The salt bridge formed between Arg256 and Glu195 in mature α-actin (PDB ID: 8A2T) is essential for actin filament stability^35,36^. Arg257Cys mutation reported by Ceron et al. and Chiappori et al. suggest that this mutation either destabilises or depolymerises the filamentous actin^35,37^. In light of the known data both NDA and SFX appear to disrupt the salt bridge that Arg256 would be forming with Glu195. The Gen3 SFX quinolone core atoms such as N4, O1, O3, F1 and F2 formed hydrogen bonds with Arg256 (Fig. 9f) as observed in simulation data. For Gen1 NDA, atoms O2, O3, N2 were the prominent ones forming hydrogen bonds with Arg256 (Fig. S15f). It is pertinent to note that SFX exhibits greater potency compared to NDA, due to the presence of an -NH₂ group at position ‘R5’ and two fluorine atoms at ‘R6’ and ‘R8’, which enable more effective hydrogen bonding with Arg256. However, owing to the small size of NDA, effective F-actin disruption observed in preceding experiments could be achieved at the higher concentrations. It is likely that SFX and NDA mediated abolition of the salt bridge and hydrophobic contacts destabilise lateral interstrand interactions thus explaining the destabilisation and disruption of F-actin filaments.

The yeast cell assay clearly revealed that Gen1 NDA, Gen3 SFX and Gen4 MFX exhibited the strongest effect, whereas Gen2a NFX, Gen2b OFX and LFX demonstrated only moderate disruption of actin aggregates. The oxygen atoms from carboxylic and the keto groups of the FQs core, in combination with fluorine at ‘R6’ position formed hydrogen bonds with the F-actin (Fig. 10). This signifies the role of common core of FQs in forming the hydrogen bonds that form the primary mode of interaction with F-actin. SFX additionally contains -NH_2_ present at position ‘R5’.

Another important position on FQs implicated in disruption of F-actin is ‘R7’ (Fig. 10). The piperazine ring at ‘R7’ position in Gen2a doesn’t have much interaction as observed in STD-NMR and MD simulations (Fig. 10c, d). The methylation of the same piperazine in Gen2b FQs increases its steric bulk and the introduced methyl group forms hydrophobic interactions with the protein, which may explain their improved efficacy (Fig. 10e, f). The bulkier dimethylated piperazine occupying ‘R7’ position in Gen3 SFX has considerable overall interaction thus enhancing the compound’s effectiveness (Fig. 10g). At the same position, the azabicyclo ring in Gen4 MFX enhanced the F-actin interaction compared to Gen2 FQs therefore this binding may facilitate better F-actin disruption (Fig. 10h).

CFX, SFX and MFX contain a cyclopropyl group at the ‘R1’ position, and the interaction of cyclopropyl group with F-actin is evident in STD NMR spectra, majorly driven by hydrophobic and Van der Waals interactions (Fig. 10). The actin disruption activity of SFX and MFX is far greater than CFX regardless of the presence of cyclopropyl group thereby indicating that this cyclopropyl ring is important in interaction but not in the disruption of actin filaments. Likewise, for moderately effective OFX and LFX, the presence of oxazine ring-structure spanning positions 1 and 8 does not appear to have a pronounced effect on reduction in size of actin aggregates when compared to the stronger activity of Gen3 and Gen4 FQs.

It is important to note that, for NDA, the ‘X’ position (Fig. 10b) is a nitrogen, placing it in the naphthyridone family rather than among fluoroquinolones. Among all the compounds investigated in this study, Gen3 SFX and Gen4 MFX seem to have the best disruption activity of actin aggregates. Also, the smaller bulk of the Gen1 NDA may facilitate better uptake by the yeast cells and thereby contributing to significant actin aggregate disruption. However, NDA exhibits activity only at higher concentrations, as indicated by RALS and gel filtration experiments.

FQs have been reported to have a better Blood Brain Barrier (BBB) permeability compared to other antibiotics^38,39^. Gen1 NDA, Gen2a CFX and NFX, and Gen3 SFX are known to have poor BBB permeability, whereas Gen2b OFX and LFX have moderate permeability^40^. Gen4 MFX is known to have better crossing ability through BBB^28^. Based on the presented experimental data Gen3 SFX and Gen4 MFX emerge as the promising candidates towards the F-actin disruption.

The presented data highlight the roles of functional groups at positions ‘R5’ and ‘R7’ of already existing FQs in F-actin disruption. The lipophilicity conferred to the MFX by the functional group at ‘R7’ position is vital for its ability to cross BBB and also found to be effective towards F-actin disruption. This makes MFX a potential candidate for the treatment of actin-mediated neuropathies. However, the equally effective SFX may be better suited for targeting aberrant actin aggregation in peripheral tissues of the body. By modifying the lipophilicity at the ‘R7’ position, even SFX could be developed into a potent therapeutic candidate for neurological disorders.

Cofilin-containing pathological aggregates are hallmarks of actinopathies. The effect of FQs on these aggregates remains an important future direction which is beyond the purview of the present study. The study comprehensively focuses on establishing the disruptive activity of FQs on F-actin filaments in vitro and validating their relevance in a cellular aggregate-clearance using the yeast model. Based on our findings we propose that all FQ’s in general utilise the common protons at positions 2 and 5 of the core quinolone rings to interact with F- actin at the lateral inter-strand leading to disruption of filaments. Thus, FQs may have potential to resolve the larger, pathological aggregates that are derived from these filaments.

It should be noted that the biophysical assays with purified proteins often require higher ligand concentrations to produce clear and robust signals detectable by the instruments. Therefore, the concentrations used in the study were selected to definitively establish and characterise the proof of principle that FQs can directly interact with and disrupt F-actin filaments/actin aggregates, rather than to replicate a clinically relevant dose.

Based on our data we conclude that the fluoroquinolone potency in F-actin disruption is strongly governed by substitutions at the ‘R5’ and ‘R7’ positions. The lipophilic ‘R7’ moiety contributes critically to binding strength, efficient filament disruption and blood brain barrier permeability. The findings provide a foundation for the development of more potent FQ derived actin disruptors with a superior BBB permeability and will be required to undergo a rigorous pharmacological evaluation. The data reported in this study establishes the structural basis of FQ-actin interactions and identify these novel interaction sites at both the protein and drug levels for the first time.

Materials and methods

All chemicals and drug compounds were of the highest purity grade obtained from Sigma Aldrich, Hi-media, TCI Chemical, and SD Fine chemicals, unless otherwise mentioned. Note that Actin and FQ concentrations were adjusted to suit the sensitivity of the experiments while preserving the Actin to FQ compound molar ratio.

Actin protein purification

Actin was purified from the cytoskeletal muscle acetone powder using G-buffer (2 mM tris.HCl pH 7.5, 0.2 mM CaCl_2_, 0.5 mM ATP, 0.5 mM DTT, 1 mM NaN_3_) as per Pardee & Spudich protocol with some minor modifications to obtain high purity levels for the protein^41^ (Fig. S23). The F- actin filaments were resuspended in F-buffer (10 mM tris. HCl pH 7.5, 50 mM KCl, 2 mM MgCl_2_, 1 mM ATP, 1 mM NaN_3_) and were stored at -80 °C. The actin protein concentration was determined spectrophotometrically (Implen nanophotometer NP80) at 290 nm (note longer wavelength) corrected for background scattering at 340 nm^42,43^. The purified protein was identified by Peptide Mass fingerprinting using MALDI TOF MS (Bruker corporation, USA) at TIFR, Mumbai.

Right angle light scattering (RALS) measurements

RALS measurements were performed using the Cary Eclipse Fluorescence Spectrophotometer (Agilent technologies, USA) using a 1 cm x 1 cm quartz cuvette with 1 cm path length (Starna Scientific, UK). The excitation and emission wavelength were both set at 350 nm, the excitation and emission slit width was kept 5 nm, the excitation and emission filter were set to auto and PMT voltage set to medium. The concentration of actin protein was fixed at 3 µM and was incubated with varying concentrations of FQs from 0 h to 72 h in F- buffer.

Size exclusion chromatography (SEC)

SEC experiments were performed using Biorad NGC (Biorad, USA) at Department of Chemical Sciences, TIFR Mumbai, India. The 30 µM F-actin protein was treated with FQs at protein-to-compound molar ratios of 1:15 and incubated for 1 h in F-buffer. The running F-buffer and the FQs treated F-actin protein solution was degassed prior to applying the sample onto the Superdex 200 10/300 GL column (GE Healthcare, USA). The chromatogram was analysed and plotted with Origin 9.0 software.

Transmission electron microscopy

Protein concentration of 5 µM F-actin was used for these studies. F-actin was treated with FQ compounds at protein-to-compound molar ratios of 1:30, and 1:50 in F-buffer. 5 µL of F-actin control and FQs treated F-actin were adsorbed on 400 mesh formvar/carbon coated copper grids for an hour before negative staining with 1% uranyl acetate, the excess stain solution was removed with a filter paper as previously described^44^. The negatively stained grids were imaged on JEOL JEM-1400 PLUS electron microscope (JEOL, Japan) equipped with EMSIS Tengra CCD camera and LaB_6_ filament operating at an accelerating voltage of 120 kV at ACTREC Mumbai, India.

Circular dichroism spectroscopy

Circular Dichroism spectra were acquired on Jasco J-1500 spectrometer (Jasco Corporation, Japan) between 260 nm and 190 nm in 0.2 cm pathlength J/21 quartz cells (Jasco Corporation, Japan). The measurements were performed with a 1 nm step size with the digital integration time of 4 s at a scan rate of 50 nm/min and at least three accumulations were averaged for each sample. The baseline scan of the F-buffer containing FQs was subtracted from the F-actin treated with FQs. The protein concentration was maintained at 3 µM and treated with FQs at protein-to-compound molar ratios of 1:5, 1:15, and 1:30. The spectra from 260 nm to 200 nm were analysed using JASCO spectra analysis software and protein secondary structural content was estimated using CD Multivariate SSE program provided by JASCO Corporation. The CD spectrums are expressed in the mean residue molar ellipticity [θ] calculated from the following equation: [θ] = θ_obs_/10.n.c.l (deg.cm^− 2^.dmol^− 1^); where θ_obs_ is the observed ellipticity in degrees, n is the number of amino acid residues in the protein, c is the final molar concentration of the protein, and l is the path length in cm^31^.

Differential scanning calorimetry

The experiments were performed on a Nano DSC instrument (TA instruments, USA). All measurements were performed using F-buffer and G-buffer by replacing Tris-HCl with HEPES. The protein was dialysed against their respective buffers and FQ compounds solutions were prepared by dissolving the compounds in the dialysed buffer. A protein concentration of 50 µM was measured at 25–90 °C, with a scan rate of 2 °C per minute at the constant pressure of 3 atm^45,46^. The DSC thermograms were analysed using NanoAnalyze (TA instruments, USA) processing software to obtain the temperature-dependent calorimetric enthalpy change (ΔH_cal_) and the transition temperature (T_m_).

Actin depolymerisation dynamics in Saccharomyces cerevisiae ∆end3 yeast assay

Media and growth conditions

The Saccharomyces cerevisiae ∆end3 strain YNL084C (MATα; his3∆1; leu2∆0; lys2∆0; ura3∆0) purchased from Dharmacon Inc. was used in this study. A loop full of cells from a YPD plate (1% yeast extract, 2% Bacto-peptone, 1% glucose, and 2% agar) were inoculated in 25 mL liquid YPD medium (1% yeast extract, 2% Bacto-peptone, and 1% glucose) and were grown overnight to saturation in a shaker incubator at 30 °C and 200 rpm. The cells were subsequently shifted to 0.1 OD_600nm_ in fresh YPD medium and grown until they reached stationary phase (24 h)^22^. The stationary phase cells were treated with 300 µM FQs and allowed to grow for another 6 h.

Rhodamine-phalloidin staining of actin aggregates

Actin staining in ∆end3 cells was performed as described in Pathak et al., 2021. Briefly, cells were fixed using 4% formaldehyde for 10 min at 30 °C. These cells were harvested by centrifugation at 3000 rpm for 5 min and the pellet was resuspended in 6 mL phosphate-buffered saline (136.89 mM NaCl, 2 mM KCl, 10.14 mM Na_2_HPO_4_, 1.76 mM KH_2_PO_4_) (pH 7.4) and were fixed by adding 4 mL 4% formaldehyde for 1 h at room temperature. Cells were washed twice with PBS and finally resuspended in 0.5 mL PBS. 0.07 µM Rhodamine-phalloidin (Ex/Em: 540/565nm; Themofisher Scientific, USA) was added to 50 µL of this cell suspension followed by incubation for 1 h in the dark. The cells were washed five times with PBS and subsequently visualised using an Olympus IX83 inverted microscope equipped with a DP80 CCD camera and a 100x/1.4 NA oil immersion objective. A pE-300white CoolLED light source was used for sample illumination. Z-stack images were acquired from top to bottom at 0.31 μm intervals. The merged Z-stacks were processed by subtracting background noise. For quantification, 50 randomly selected cells from one experiment were analysed and data from two such experiments is represented. The actin phenotype observed was categorised as aggregated (F-actin present as large clumps) and diffused/punctate (F-actin present as smaller punctate structures throughout the cell) based on the area obtained using Image J. Images were acquired, processed, and analysed using CellSens Dimension software (Olympus). Final images were assembled using Adobe Photoshop.

Saturation transfer difference (STD) NMR

STD NMR studies were performed with a 30 µM solution of F-actin and 1.5 mM FQs compounds in F-buffer. The solution contained a 50-fold molar excess of FQs over actin, with 10% D₂O added to the actin-FQs mixture. NMR spectra were acquired using a Bruker Ascend 400.3 MHz spectrometer (¹H nucleus) equipped with an HR-BBO400S1-BBF/H/D-5.0-Z SP probe (¹H, ¹³C, ³¹P, ¹⁵N) and pulsed-field gradients along the z-direction at a temperature of 298 K. The spectrometer was operated with Bruker TopSpin 4.1 software for data collection and processing. The standard STD NMR pulse sequence stddiffesgp.3 was utilised to obtain the spectra. This pulse sequence applies a selective train of Gaussian-shaped pulses with a 1% truncation, a pulse length of 49 ms, and a 2 ms separation between pulses^47^. A total of 256 scans were acquired, with a saturation time of 2.0 s for the STD studies. The deuterated water (HOD) signal was suppressed using gradient-tailored water suppression encoded in the standard STD NMR pulse sequence. Protons of actin protein were selectively irradiated for saturation, with the on-resonance frequency set at 1.22 ppm after multiple experiments and the off-resonance frequency set at -40 ppm. The STD effect percentage was calculated by subtracting the on-resonance spectrum from the off-resonance (reference) spectrum, yielding the STD (difference) spectrum^48^.

In-silico studies of actin and actin-fluoroquinolone interactions

The pentameric Cryo-EM structure of the F-actin protein (PDB ID: 8A2T) was used for all the in-silico studies^36^. Missing residues were modelled using the COOT software^49^. CHARMM-GUI^50^ was then used to prepare the molecule for GROMACS2022^51^ to perform molecular dynamics (MD) simulations. CHARMM-GUI v3.7 Solution Builder^52^ module was used to solvate the protein in an octahedral water box, with an edge distance of 10 Å. NaCl ions were added to neutralise the system. Electrostatics were treated using Particle Mesh Ewald algorithm^53^.

The protein was modelled using the CHARMM36m force field^54^ and water molecules were modelled using the TIP3P model^55,56^. Forcefield parameters for ADP and FQ’s compounds were generated using CHARMM-GUI. Energy minimisation was performed using the steepest descent algorithm until the maximum force acting on the system dropped below 1000 kJ.mol^− 1^.nm^− 1^. Subsequently, temperature equilibration was carried out to 303.15 K in 250 ps using the V-rescale^57^ thermostat with a τ_t_ of 1 ps. Solute and solvent groups were coupled to independent thermostats. Pressure equilibration followed, where the system was equilibrated to 1 atm in 500 ps using the C-rescale barostat^58^ with isotropic pressure scaling, τ_p_ of 1 ps and compressibility factor of 4.5 × 10^− 5^ bar ^− 1^. During both equilibration steps, heavy atoms were subjected to position restraints with a force constant of 1000 kJ.mol^− 1^.nm^− 2^. The structure obtained after pressure equilibration was used as the starting point for the production runs.

Three independent production simulations for F-actin (pentameric actin with ADP and Mg^2+^) were performed, each lasting around 250 ns with a 2 fs time step. A velocity rescaling thermostat^57^ regulated the temperature at 303.15 K (τ_t_: 1 ps), while pressure was maintained at 1 atm using the C-rescale barostat^59^ with a τ_p_ of 5 ps. Electrostatic interactions were handled via the Particle Mesh Ewald method^53^, and all h-bonds were constrained using the LINCS algorithm^60^. Structural snapshots were recorded every 20 ps for further analysis. Equilibrated actin structure with ADP and Mg^2+^ were used for further docking steps and simulations with FQ compounds.

Fluoroquinolone compound preparation for docking

FQ compounds Nalidixic acid (NDA), Ciprofloxacin (CFX), Norfloxacin (NFX), Ofloxacin (OFX), Levofloxacin (LFX), Sparfloxacin (SFX), and Moxifloxacin (MFX) were retrieved from the PubChem Database. Energy minimisation was performed using ArgusLab (http://www.arguslab.com/arguslab.com/ArgusLab.html) with the BFGS (Broyden–Fletcher–Goldfarb–Shanno) algorithm^61^ and the Universal Force Field^62^.

Molecular docking and simulations

From the energy minimised structure of HACT, Mg^+ 2^ and ADP were removed to perform molecular docking. Hydrogen atoms and Kollman charges were added using AutoDock v1.5.7^63^. FQs compounds were assigned Gasteiger charges before performing blind docking using AutoDock Vina v1.1.2^64,65^. The best pose of the FQs was selected on the basis of the lowest binding affinity. The affinity values (kcal.mol^− 1^) for all the FQs are mentioned in Supplementary Table S2.

FQs were parameterised using the CHARMM General Force Field (CGenFF)^66^ via the Ligand Reader & Modeller module in CHARMM-GUI v3.7. Solution Builder was then used to generate CHARMM force field topologies for the ligand-parameterised protein complexes and simulations were set up as described above. For each protein-ligand complex, three independent production-run simulations of 80ns each were performed. While all three replicas for a given complex began from the identical energy-minimised and equilibrated structure, each was assigned a different set of randomly generated initial velocities for the atoms (Supplementary Table S3). MD data was analysed using GROMACS tools in combination with in-house python scripts and PyMol 3.1.0 was used for visualisation^67^.

Statistical analysis

Quantitative data are expressed as mean ± SEM. Statistical analysis was performed using the GraphPad Prism Version 8.4.2 (GraphPad Software, Inc., La Jolla, CA, USA). Statistical differences were evaluated using one-way analysis of variance (ANOVA) where P-values less than 0.05 were considered to be significant. Specific information on the statistical tests are provided in the respective figure legends.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1

Supplementary Material 2

Supplementary Material 3

Supplementary Material 4

Supplementary Material 5

Supplementary Material 6

Supplementary Material 7

Supplementary Material 8

Supplementary Material 9