Megalin-Aminoglycosides Interaction as a Mechanism of Ototoxicity: Insights from In Silico Modeling and Cell Functionality Assays

1. Introduction

Streptomycin was first isolated eight decades ago from Streptomyces griseus [1], initiating the development of aminoglycoside (AG) antibiotics; since then, thousands of AGs have been either naturally isolated (produced by actinobacteria such as Streptomyces or Micromonospora) or obtained semi-synthetically (chemically modified derivatives of natural aminoglycosides). AG antibiotics are widely used to treat severe bacterial infections caused by Gram-positive and Gram-negative bacteria [2–4}. Despite their efficacy, their clinical use is frequently limited by adverse effects, mainly nephrotoxicity, neurotoxicity, and ototoxicity, which may affect the vestibular and/or cochlear systems. Vestibular toxicity occurs when the vestibular cranial nerve branch responsible for balance is damaged, and cochlear toxicity occurs due to damage to the cochlear Hair Cells (Inner Hair Cells -IHC- and Outer Hair Cells -OHC-) and the cranial nerve branch leading to sensorineural hearing loss [5]. The likelihood of ear damage occurring depends on the dose and duration of treatment with aminoglycosides and genetic susceptibility [6}.

Once AGs reach the inner ear fluid endolymph (fluid surrounding IHC), AG antibiotics enter IHC and OHC. Several mechanisms of AG antibiotics transport into IHC and OHC have been described. The three most important are: 1) via mechano-transduction channels, located at the tips of stereocilia which open and close in response to hair bundle deflection and allow non-specific cation influx {7], 2) by endocytosis, where AG antibiotics entry occurs through cell membrane invagination [8], and 3) by ion channels, like transient receptor potential channels, which are large calcium-permeable cation channels [9].

Like their antibacterial mechanism, AGs target the ribosomes within mammalian inner ear cells, triggering a cascade of deleterious intracellular events. Once aminoglycosides are inside hair cells, increased generation of reactive oxygen species (ROS) or free radicals, activation of stress kinases and proteolytic enzymes have been observed. Through in vivo experiments, it has been shown that AG antibiotics administration leads to activation of the JNK signaling pathway, which triggers pro-apoptotic signals [6,10]. Furthermore, it is suggested that inhibition of mitochondrial protein synthesis by AG antibiotic exposure leads to decreased ATP. Reduced energy production compromises mitochondrial integrity and predisposes to cytochrome C release and caspase activation, suggesting that hair cells undergo caspase-mediated apoptosis [4,11,12,13].

1.1. Megalin

Megalin, also known as Low-Density Lipoprotein Receptor-Related Protein 2, is a multiligand scavenger receptor belonging to the LDL receptor family. It is predominantly expressed by epithelial cells, including those of kidneys (renal proximal tubules) and the cochlea. In the kidney, megalin is essential for the reabsorption of filtered proteins. It interacts with various ligands, including aminoglycosides like gentamicin [14].

In the inner ear, the function of megalin remains less understood. Its expression in the cochlea suggests a potential role in the uptake of specific ligands, which could be relevant to auditory physiology. Structurally, the human megalin receptor is a large glycoprotein whose extracellular domain is composed primarily of complement-type repeat (CR) domains essential for ligand binding. Each CR domain comprises about 40 residues and contains a calcium-binding site and three disulfide bonds [15] . This work focused on the 10th CR domain from the human megalin receptor determined by Dagil et al. 2013. Recently, Goto et al 2024 elucidated the full architecture of megalin purified from rat kidneys using cryoelectron microscopy (Figure 1) [16].

Seven AG antibiotics (Gentamicin, Tobramycin, Kanamycin, Hygromycin, Streptomycin, Neomycin, and Paromomycin) were selected to explore the molecular recognition between megalin-AG antibiotics; all except hygromycin are approved by the Food and Drug Administration (FDA) and are available for clinical use¹ Drew RH. Aminoglycosides. 2024, http://www.uptodate.com. Some characteristics and chemical structures of the seven AG studied here are shown in Table 1 and Figure 2.

1.2. Protonation States and pKa Considerations

The protonation state of aminoglycoside amino groups is relevant in receptor binding, given their contribution to electrostatic interactions. At physiological pH, these amino groups of aminoglycoside antibiotics (NH₃⁺) are positively charged, enabling strong electrostatic interactions with negative charges on the receptor's surface. In this work, all aminoglycosides were modeled with protonated amino groups based on prior findings that such interactions contribute significantly to binding affinity and, consequently, to the ototoxic effects [14,17,18]. Furthermore, Zhou et al. (2012) showed a similar behavior for L-dopa, which holds two protonatable groups, the carboxylic moiety (pKa= 2.3) and the amino group (pKa= 8.11), where NH₃⁺/NH₂ can be protonated or deprotonated between pH 7.0- 8.0 (17). In addition, unpublished data from our group indicate that non-protonated amino groups might limit or impede the binding of aminoglycosides to megalin, highlighting the importance of pH-dependent protonation equilibria.

2. Results and Discussion

2.1. Molecular Dynamics Simulations and Cluster Analysis

The molecular dynamics simulations of the 10 th CR domain of human megalin was carried out for 200 ns by triplicate, yielding 600 ns of simulation data [19]. Root Mean Square Deviation (RMSD) values for Cα atoms were calculated for all systems (data not shown). Based on the RSMD behavior, we analyzed the final 50 ns from each replica. Cluster analysis allowed us to select the central structure of each replica. The most populated cluster of each experiment contained more than 75% of the analyzed frames. Central structures from the most populated clusters of each cluster were visually compared to the 20 structures determined by NMR from 2M0P [14]. As shown in Figure 3, the most representative structure from each cluster closely resembles the NMR structures.

To identify potential binding sites on the surface of the 10th CR domain of human megalin, blind molecular docking of seven aminoglycosides (Gentamicin, Kanamycin, Hygromycin, Paromomycine, Streptomycin, Tobramycin, and Neomycin) was performed using AutoDock Vina. The three representative Megalin structures (from each cluster) were used for docking experiments. Each ligand was docked 100 [20] independent runs against each cluster. The resulting poses were clustered using a 2 Å RMSD cutoff to obtain the most probable binding sites for each megalin cluster. Binding affinities and binding site occupancy percentages are summarized in Table 2.

2.2. Molecular Docking Analysis

Docking results for the three megalin structures showed two major aminoglycoside binding sites with favorable ΔG_b values (Table 4). Figure 4a shows the major bong sites of cluster 1. Figure 4b-4h show representative aminoglycoside-meglin (cluster 1) interactions. Gentamicin formed three hydrogen bonds with Thr 1130, Asn 1132, and Glu1140 (Figure 4b). Tobramycin formed two hydrogen bonds with Thr 1130 and Glu 1140 with Gly 1137 (Figure 4c). Neomycin established four hydrogen bonds: two with Glu1140, one with Asp 1136 and Ser 1145 (Figure 4d). Paromomycin engaged six hydrogen bonds: two with Asp 1111 and one with Cys 1116, Asp 1117, His 1119, and Lys 1124 (Figure 4e). Hygromycin established two hydrogen bonds with Asn 1132 and Glu 1140 (Figure 4f). Streptomycin presented one hydrogen bond with Glu1140 (Fig 4g). Finally, Kanamycin shows five hydrogen bonds: one with Ala 1107, Cys 1109, two with Asp 1111, and one with Gln 1113 (Fig 4h). The amino acids highlighted in bold are consistent with those reported by Dagil et al., 2013 [14]. These authors also performed molecular docking studies of the megalin receptor with gentamicin. Notably, Thr 1130 and Asn 1132 were identified as part of one of the binding sites in both studies {14]. The Supplementary Material presents additional docking results for clusters 2 and 3, where overlapping binding sites were observed across ligands.

Table 3 highlights conserved residues forming the binding sites in each megalin cluster, essential for interacting with the various aminoglycosides.

2.3. Binding Free Energy Determinations

Binding free energy values (ΔG_b) for each megalin-aminoglycoside complex was estimated for each representative cluster derived from molecular dynamics simulations (Table 4). The computational results suggested that the binding process of each megalin cluster with antibiotics was governed by electrostatic interactions, mainly by Coulombic forces, and a minor proportion by hydrophobic interactions. It seems important to emphasize that Megalin-Gentamicin, Megalin-Tobramycin, and Megalin-Neomycin systems presented the most favorable ΔG_b values due to important electrostatic contacts in the binding, suggesting a stronger binding affinity that may contribute to their higher ototoxic potential.

2.4. Cell Functionality Assays

Resazurin assays in HEK 293 cells revealed that gentamicin reduced cell functionality at concentrations above 10 μM (Figure 5). This correlates with the high binding affinity of Gentamicin observed in docking experiments. Cell functionality decreased in a concentration-dependent manner.

The action mechanism of gentamicin in kidney cells has been reported. This mechanism consists of the antibiotic's internalization via the megalin-cubulin complex, followed by intracellular calcium accumulation. This causes endoplasmic reticulum stress and generation of reactive oxygen species (ROS), which consequently trigger necrosis and the intrinsic apoptosis pathway. It has been reported that the proliferator–activated receptor alpha (PPAR-α) is an important regulator of ROS within the cell. In tubular kidney cells, PPAR-α is less expressed when exposed to gentamicin, which may further enhance sensitivity to its toxicity [16,23].

Notably, gentamicin is frequently prescribed to pediatric patients, and it is reported that 5% of the administered dose of antibiotics accumulates, resulting in non-oliguric acute kidney injury (AKI) for long exposures to this antibiotic [24].

3. Materials and Methods

3.1. Molecular Dynamics Simulations

Molecular dynamics simulations (MDS) of the 10th CR domain of human megalin (PDB ID: 2M0P) were performed using the GROMACS 2019.2 software. Nuclear magnetic resonance (NMR) resolved the human megalin structure, and then the first model, out of twenty-structure ensembles, was used as the initial structure (14). CHARMM-GUI server was assigned hydrogens and charges with CHARMM36 force field [25,26]. Each system was surrounded by 10 Å of TIP3P water from any protein atom to a cubic water box; 0.15 M KCl was added to the simulation box to neutralize the water. All simulations were conducted in triplicate at 310.15 K and 1 atm of pressure for 200 ns, and inputs were used as suggested by the CHARMM-GUI server [19,27].

Trajectories were analyzed using GROMACS 2019.2 utilities. The last 50 ns of each simulation were clustered using the GROMOS method for clustering with a 0.75 Å cutoff (using this value was enough to obtain more than 75% of the structures on the most populated cluster).

3.2. Molecular Docking Assays

The central structure of the most populated cluster was used as the receptor model for blind molecular docking with aminoglycosides (Gentamicin, Kanamycin, Hygromycin, Paromomycin, Streptomycin, Tobramycin, and Neomycin) using Autodock Vina 1.2.5 [20]. Aminoglycosides were obtained from Pubchem Database [28]. Pubchem Database for each ligand were: Gentamicin (3467), Kanamycin (6032), Hygromycin (35766), Paromomycin (165580), Streptomycin (19649), Tobramycin (36294) and Neomycin (8378). Each aminoglycoside was loaded into MarvinSketch, and the best conformer was saved as mol2 and pdb file. Each ligand was docked in 100 blind docking experiments on the surface of each megalin cluster. The resulting poses were clustered based on a 2 Å RMSD threshold. Molecular interactions maps were generated using Maestro from Schrodinger and Visual Molecular Dynamics [22].

It is essential to point out that in this work, all the amino groups of aminoglycosides were considered in their protonated form (NH₃⁺), since it has been shown that strong electrostatic interactions are established, which contributes to hearing loss [14,17,18].

3.3. Computational Binding Energy (ΔG_b) Calculations

The free energy change of binding (ΔG_b) between megalin and its ligands was calculated according to the method described by Baker et al.. using equation 1 [29];

ΔG_b= ΔG_solv + ΔG_Coul +ΔG_non-elec,

(1)

where the free energy change of solvation (ΔG_solv) and coulombic free energy changes ΔG_Coul were calculated using the Adaptive Poisson-Boltzmann Solver (APBS) program (http://www.poissonboltzmann.org/) [21].

In Nathan Baker´s script, we assigned dielectric constants of 78 and 4 for water and protein, respectively, for electrostatic determinations. Furthermore, the parameters were taken from the PDB2PQR server (http://server.poissonboltzmann.org/pdb2pqr) [30] as ionic radii and atomic charges for megalin, which were assigned from the CHARMM force field [31] at pH 7.0 [32]. Ligands' atomic charges were assigned from the force field implemented in the Autodock Vina program [20].

Regarding the ΔG_non-elec is the non-electrostatic contribution to the binding energy and was determined by multiplying the change in solvent accessible surface (ΔASA) upon binding by the coefficient interfacial tension (γ=5 cal•mol^-1 Å^-2) [33]. ASA values were determined by the VMD 1.9.1 program (https://www.ks.uiuc.edu/Research/vmd/) [22]:

ΔG_non-elec = γ (ASA_{megalin-ligands complex}− ASA_megalin – ASA_ligands),

(2)

3.4. Resazurin Cell Functionality Assays

The blue non-fluorescent dye Resazurin (7-Hydroxy-3H-phenoxazin-3-one 10 oxide), which is irreversibly reduced to the pink-colored fluorescent resorufin by dehydrogenase enzymes in metabolically active cells, was used to determine cytotoxicity on HEK293 cells. These cells could be used as a cellular model of ototoxicity since it has been reported that ototoxicity by aminoglycosides is related to other tissue damages such as the kidney (34). Cells were cultured with Dulbecco's Modified Eagle Medium (DMEM), supplemented with 10% Fetal Bovine Serum, and maintained at 95% humidity and 5% CO₂. We used antibiotics with the most favorable affinity energy for the treatments. The concentrations were 10, 25, 50, and 75 mM. Not-treated cells (NT) were positive controls, while cells treated with a stable peroxide solution were negative controls. All data were analysed with one-way ANOVA and post-test Tukey (P < 0.05).

4. Conclusions

The molecular docking results revealed favorable binding interactions with ΔGb<0 between aminoglycosides with the megalin receptor, indicating spontaneous complex formation. Binding affinity was mainly governed by electrostatic interactions and hydrogen bonding, with minor contributions of non-electrostatic forces. Gentamicin, Tobramycin, and Neomycin exhibited the most favorable ΔGb values, suggesting stronger binding to megalin than Kanamycin, Hygromycin, Paromomycin, and Streptomycin. Gentamicin was identified as the aminoglycoside with the strongest binding affinity, and therefore, it was selected for cell functionality assays. HEK293 cells are susceptible to gentamicin by reducing its functionality, supporting its dual role as a potent antibiotic and a potential contributor to organ toxicity when not carefully monitored.