A structural feature common to many previously developed NK1R antagonists is the 3,5-bis-trifluoromethylphenyl (TFMP) group (Supplementary Fig. 1). While this motif is known to enhance receptor binding affinity, it may also influence drug metabolism, pharmacokinetics or receptor interactions, thereby influencing therapeutic efficacy16,17,18. Despite this, no systematic studies have investigated whether modifying or removing TFMP could improve the antidepressant potential of NK1R antagonists. Although the relationship between compound structure and clinical efficacy remains to be established, evaluating structurally distinct NK1R antagonists offers an initial step toward exploring how chemical modifications may influence biological outcomes. To address this, we utilized computational approaches to identify NK1R antagonists with alternative structural features, evaluated their antidepressant-like effects in preclinical models of depression using behavioral assessments and analyzed markers of neuroinflammation at the mRNA level. Our findings suggest that the structurally distinct antagonists identified in this study exhibit antidepressant-like effects, providing renewed evidence for further exploration of NK1R antagonism as a therapeutic strategy for MDD.
A dataset containing ChEMBL ID and activity measurements, including half-maximal inhibitory concentration (IC₅₀), half-maximal effective concentration (EC₅₀), inhibition constant (Kᵢ), and dissociation constant (K) values, was obtained from the ChEMBL database (https://www.ebi.ac.uk/chembl/, 5,997 results) and NCBI PHAROS database (https://pharos.nih.gov, 1,632 results). After removing duplicates, 2,499 ligands were included in the study. The pChEMBL value was calculated as -log(effective value) for prediction. Dimensionality reduction of ligand structures was conducted by t-distributed stochastic neighbor embedding with the Rtsne package (v 0.16) in R 4.2.2.
The Enamine Screening Collection, which contains 2,681,264 molecules, was obtained from https://enamine.net/compound-collections/screening-collection and used for virtual screening.
QSAR models were developed to predict the pChEMBL value using KNIME (version 4.4.4) with the Schrödinger extension (version 21.4.135) and DeepChem in Maestro 12.8 (Schrödinger 2021-2 version). Four different approaches were used: random forest, simple regression, AutoQSAR and DeepChem. The dataset was divided into a training set (70%, 1,749 ligands) and a test set (30%, 750 ligands). For AutoQSAR and DeepChem, the SMILES (Simplified Molecular-Input Line-Entry System) string for each ligand was converted to .Mae file format. For other methods, the SMILES strings were transformed into Morgan fingerprints (radius, 2; number of bits, 1,024) using RDkit nodes in KNIME. Predictions were evaluated using the coefficient of determination (R) and root mean square error/deviation (RMSE/RMSD).
Ligands were clustered based on Murcko scaffolds and ECFP4 fingerprints using Tanimoto coefficients, and hierarchical clustering was performed with a distance threshold of 0.75 to form 35 clusters. Fifteen ligands were manually selected for in vitro experiments considering cluster assignment, prediction score and core structure.
Similar ligands to compound #1 were searched using infiniSee (BioSolveIT) in the Enamine REAL space containing 3.4 × 10 ligands and the KnowledgeSpace with 2.9 × 10 ligands. The predicted pChEMBL value of the top 200,000 similar ligands was calculated with the generated DeepChem model. Similar ligands were filtered with the predicted pChEMBL score (>9.5). Among 40 ligands, 8 ligands were selected on the basis of core structure.
Molecular docking experiments were performed using Schrödinger Glide software (version 2022-4), and molecular dynamics (MD) simulations were conducted using Desmond on the Schrödinger platform. All calculations were executed in a Linux environment.
The crystal structure of the protein (PDB ID: 6HLO) was imported into Maestro and prepared for docking using the Protein Preparation Workflow with the OPLS4 force field. The receptor grid for docking was defined based on the position of the bound ligand, Aprepitant, without further modification. The compounds were prepared for docking using the LigPrep function, which generated all possible stereoisomers and states at a target pH of 7.0 ± 2.0.
The prepared compounds were docked to the protein structure using the Glide software with extra precision (XP) settings. This docking process generated protein-ligand complexes for further analysis and exploration.
The protein-ligand complexes obtained from the molecular docking were utilized to build the system for MD simulations. The simulation setup included the placement of a POPC (300 K) membrane model and the addition of counter ions to neutralize the system, with TIP3P as the solvent system. The MD simulation was performed for a duration of 500 ns using the ensemble class NPγT at a temperature of 300.0 K. The simulation trajectory was subsequently analyzed using the Simulation Interactive Diagram.
The FLIPR calcium 6 assay (Molecular Devices) was used for evaluating IC and the dose-response curve of each ligand. hTACR1-HEK293 cells were seeded in 96-well black clear-bottom plates at a cell density of 3 × 10 cells per well. The following day, the cells were loaded with the FLIPR calcium 6 dye mixed at 1:1 to medium with 2.5 mM probenecid. The assay plate was wrapped with aluminum foil and incubated for 2 h at 37 °C in a humidified CO incubator. Each concentration of chemicals was dissolved in 1× Hank's Balanced Salt Solution (plus 20 mM HEPES buffer, pH 7.4) and added to the well. The plate was incubated for 30 min at 37 °C before measuring. The final concentration of dimethyl sulfoxide was 0.5% in each well. Then, 1 nM SP-induced intracellular calcium mobilization was examined by monitoring the fluorescence intensity (excitation at 485 nm, emission at 525 nm, cutoff filter set at 515 nm) using a Flex Station 3 multimode microplate reader. Changes in fluorescence signal after SP addition were normalized to its starting signal and denoted as F/F.
All experimental procedures with animals were approved by Korea University and Konkuk University Institution of Animal Care and Use Committee (study approval numbers KOREA-2021-0202 and KU22186) and were performed according to university and the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines. Seven-week-old male C57BL/6 mice (Orient Bio) were group housed under a 12-h light/dark cycle and given ad libitum access to food and water. After arrival, animals were allowed 1 week to habituate to the facility before experiments and randomly divided into experimental and control groups.
The ligands were dissolved at a dose of 10 mg/kg in the vehicle (10% dimethyl sulfoxide and 10% Kolliphor in 0.9% saline) and intraperitoneally injected into each animal in experimental groups.
To generate the lipopolysaccharide (LPS) model, 0.25 mg/kg of LPS from E. coli O26:B6 (Sigma, L3755) was dissolved in 0.9% saline and injected intraperitoneally. The control group was treated with 0.9% saline instead of LPS injection. The mice received injections of either vehicle, compound #1, compound #15 or Aprepitant (Tocris Bioscience) on day 1, 1 h before LPS injection and on day 2.
To induce depressive-like behaviors in mice, restraint stress (RS) was applied using 50-ml conical tubes with many holes for 20 days, 2 h per day. From day 11 to 20, mice received injections of either vehicle or compound #15.
To generate the social isolation (SI) stress model, the animals were single-caged for 14 days. From day 6 to 13, mice received injections of either vehicle or compound #15.
Four hours after LPS injection, the frontal lobe and hippocampus were isolated from the animal brain and placed in TRIzol solution (Ambion). Total RNA sample (2 μg) was reverse-transcribed into cDNA using Moloney Murine Leukemia Virus reverse transcriptase (M-MLV RT; Promega) and oligo (dT) primer (Novagen). Quantitative reverse transcription polymerase chain reaction (qRT-PCR), which quantifies gene expression levels, was performed using 0.5 μg of the reverse-transcription generated cDNA and specific primer sets (Supplementary Table 3). PCR amplification with iQ SYBR Green Supermix was performed in triplicate using the CFX96 Touch-Time System (Bio-Rad). The final products of qRT-PCR were electrophoresed on 2% agarose gels and visualized with SafeView Nucleic Acid Stain (G108, Applied Biological Materials). Cycle threshold (Ct) values at which the fluorescent signal exceeded the background were determined by qRT-PCR, and expression values for each gene were normalized to expression values of GAPDH. Relative quantification to calculate fold change was performed using the comparative Ct method (ΔΔCt).
On day 4 after LPS administration, the tail suspension test (TST) was conducted. A four-chamber apparatus divided by nontransparent acrylic partitions was used, and the mice were suspended in each chamber by the tail. A video was recorded for 6 min, and the last 4 min were manually scored for immobility time.
On day 21 in the RS model and day 12 in the SI model, the forced swim test (FST) was performed. During each trial, mice were placed in a test cylinder (diameter 15 cm, height 25 cm) filled with water regulated at 25 ± 1 °C, and a video was recorded for 6 min. Immobility time was manually scored during the last 4 min of each trial.
On day 24 in the RS model and day 13 in the SI model, the open field test (OFT) was conducted to measure the general locomotor activity of mice. Each mouse was individually placed in an open box (40 × 40 × 40 cm) for 10 min. The locomotion of mice was tracked using EthoVision XT 11.5 (Noldus). The total distance moved and movement duration were measured for analysis.
Statistical analysis was conducted with IBM SPSS Statistics 24 for Windows software (IBM). Comparisons between groups were analyzed using one-way analysis of variance (ANOVA), followed by Tukey's honestly significant difference (HSD) as a post-hoc test. The harmonic mean of the group sizes was used for unequal-sized groups. P < 0.05 was considered significant, and values were expressed as mean ± s.e.m.