Discovery of N8: a novel IKKε inhibitor with potent anticancer activity via cytotoxicity, migration suppression, and autophagy modulation.

Introduction
Cancer poses a serious threat to human health, and remains a leading cause of deaths worldwide1. Nearly 20 million new cancer cases and 9.7 million cancer-related deaths were reported in 2022 alone2. Targeted drugs exhibit higher efficacy and specificity and lower toxicity against a variety of malignant tumours than conventional cancer treatment modes such as chemotherapy; as such, they have become the mainstay of cancer therapy over the last decade3. Kinases play significant regulatory roles in the formation and progression of numerous tumours and have been successfully targeted to treat cancer4. The first small-molecule protein kinase inhibitor, imatinib, was approved for use in patients with cancer by the United States Food and Drug Administration (FDA) in 20015. Since then, 71 small-molecule kinase inhibitors have been approved by the FDA6. Over the last 5 years, the FDA has approved 37 small-molecule kinase inhibitors, which accounts for approximately 15% of all new drugs approved by the FDA, demonstrating the promising potential of these therapeutic agents7. Despite these advances in the discovery of small-molecule kinase inhibitors, few have proved effective in treating tumour metastasis owing to toxicity and drug resistance. Thus, there is an urgent need for a new generation of safe but effective targeted therapeutic agents.
Inhibitor of nuclear factor kappa B kinase ε (IKKε) is a serine/threonine protein kinase belonging to the nonclassical IKK family. IKKε is involved in the regulation of inflammation8 and the immune response9; it is also implicated in several metabolic diseases10. Recently, IKKε was shown to act as an oncogenic protein in various cancers, including breast cancer11, hepatocellular carcinoma12, glioblastoma13, pancreatic ductal adenocarcinoma14, and prostate cancer15. Previously, our study found that IKKε regulates cancer metastasis and correlates with poor patient outcomes16. IKKε has also been identified as a potential prognostic biomarker for patients with ovarian cancer and non-small cell lung cancer17–19. Given the roles of IKKε in cancer, the development of IKKε inhibitors as therapeutic agents has gained increasing attention. Amlexanox, a drug used for the treatment of recurrent aphthous ulcers and asthma20, and considered a promising therapeutic candidate for the treatment of obesity and non-alcoholic fatty liver disease21, has been tested as a potential IKKε inhibitor22. Although other IKKε inhibitors have been identified, like BAY-98523, most are still in the preclinical stage of development.
Developing new drugs is a costly and time-consuming process, with a minimum cost of 1 billion USD associated with taking a drug candidate from discovery to development. Moreover, a drug candidate takes an average of 15 years to reach the market24. The use of extensive databases can accelerate and lower the cost of drug discovery25. Computer-aided drug design (CADD) is a field that encompasses various computational strategies for discovering, designing, and developing novel therapeutic drugs. CADD plays a crucial role in improving the design of active ligands, discovering new drugs, and understanding biological processes at the molecular level26. Virtual screening is the most common CADD method used to identify drug targets from large molecular databases27. To date, virtual screening has facilitated the discovery of many kinase inhibitors28,29.
Therefore, the present study aimed to employ computer-aided virtual screening to identify novel inhibitors targeting the kinase domain of IKKε from various compound databases, including those focused on novel compounds, kinase-specific molecules, and tumour-related agents. Through molecular docking, we successfully identified N8 as a candidate compound with a strong binding affinity for IKKε. We then assessed the anticancer efficacy of N8 across a spectrum of cancer cell lines, including colorectal, liver, bladder, breast, lung, and cervical cancers, and further explored its underlying mechanism of action, particularly its impact on cancer cell proliferation and autophagy regulation.

Materials and methods

Cells and compounds
The colorectal cancer cell line HCT116, the hepatoma cell line HepG2, the bladder cancer cell line T24, the breast cancer cell line MDA-MB-231, the lung cancer cell line A549, and the cervical cancer cell line Hela were used in this study. All the cancer cell lines were obtained from ATCC and cultured in DMEM/high‑glucose medium (Cytiva) supplemented with 10% foetal bovine serum (FBS, Excell) and 1% penicillin‑streptomycin (Gibco), in a humidified atmosphere of 5% CO2 at 37 °C. Compound N8 (CAS. 2117748–15-7), alternatively known as benzamide, N-([(1R,9aS)-octahydro-2H-quinolizin-1-yl] methyl)-3–(2-oxo-2H-1-benzopyran-3-yl)-rel-(ACI), and BAY-985 (CAS. 2409479–29-2) were purchased from Topscience (Shanghai, China) with purity >95%. Amlexanox (CAS. 68302–57-8) was purchased from MCE (Shanghai, China) with a purity of 99.88%. Compounds were dissolved in DMSO.

IKKε kinase domain model construction
The crystal structure of human IKKε (UniProt ID: Q14164) was unavailable at the time of this study. The predicted structure of the full-length IKKε (residues 1–716) was obtained from AlphaFold database (https://alphafold.ebi.ac.uk/entry/Q14164, version 30)30. The N-terminal kinase domain of IKKε (residues 1–310) was extracted for further analysis, with other regions excluded. Model quality was assessed using the per-residue confidence score (pLDDT)provided by AlphaFold.The model of IKKε kinase domain was then optimised for virtual screening and molecular dynamics (MD) simulations were performed in the presence of C35, a known co-crystal inhibitor of IKKε homologue23. To validate the structural reliability, the predicted structure was aligned to the closest homologous experimental structure, TBK1 (PDB ID: 5W5V, sequence identity: 73%), and the root-mean-square deviation (RMSD) was calculated. Additionally, Ramachandran plot analysis of the predicted IKKε kinase domain was performed using Ramachandran Plot Server (https://bu.wenglab.org/rama/).

Large-scale virtual screening against the IKKε kinase domain
A collection of commercially available compound libraries (containing ∼12 million compounds) was used for virtual screening (Supplementary Table 1). The via RDKIT31 and Open Babel32 tools were used to convert the compound data within the library into 3D structures, which were then saved in the PDBQT file format for use in the docking analysis. Meanwhile, the IKKε protein model was prepared using AutoDock Tools33. The CUDA-accelerated AutoDock-GPU34 was used for this large-scale virtual screening; the 1-GPU server was equipped with 4 Nvidia RTX-2080-Ti cards. Each library compound was docked into the active site of IKKε, defined by compound C35. The receptor was kept rigid during the docking process, whereas the ligands were allowed to be flexible. For each ligand, 20 poses were scored using the inherent scoring function of the software; the pose with the best docking energy score was then selected for further analysis.
After the docking experiments were completed, 5,000 compounds with the lowest docking energy scores were selected for further analysis. The results were then filtered using SMARTS (https://www.surechembl.org/knowledgebase/169485-non-medchem-friendly-smarts) and the rule-of-five (300 ≤ MW ≤ 500; −2 ≤ logP ≤ 7; HBD ≤ 5; HBA ≤ 10; TPSA ≤ 150; NRB ≤ 10) rules; the remaining compounds were assigned QED scores. 500 compounds were then selected according to their scaffold diversity, docking energy score, and QED score. After pharmacophore filtering and visual inspection, 27 compounds were selected as final hits.
The 2D protein-ligand interactions were analysed using LigPlot+ (version 2.2.4)35. The 3D images and plots were generated using VMD 1.9.136.

In vitro IKKε kinase activity assay
The ADP Glo assay (Promega) was used to detect the degree of reactivity between N8 (10, 50, or 100 μM) or amlexanox (100 μM) and IKKε (50 ng), the synthetic peptide substrate (200 μM), and ATP (100 μM). The synthetic peptide substrate had the following amino acid sequence: ADDDYDSLDWDAKKK37.

Cytotoxicity assay
The Alamarblue assay was used to measure cytotoxicity. Depending on the extent of cell growth, 3,000 or 5,000 cells/well were seeded in 96-well plates and cultured for 24 h. The cells were then treated with N8 (6.25, 12.5, 25, 50 μM), amlexanox (12.5, 25, 50, 100, 200 μM), or solvent control DMSO (0.5% v/v) for 72 h. Alamarblue solution (5%, SAICHI) was then added to each well, and the cells were incubated at 37 °C for a further 2–4 h. The fluorescence was measured at a wavelength of 540 nm on the BioTek Epoch 2 microplate reader.

Colony formation assay
Cells (200 cells/well) were seeded into a 12-well plate and cultured with N8 or amlexanox for 14 days. The resulting clones were fixed in 4% polyformaldehyde solution (Biosharp) for 30 min and stained with crystal violet solution (Beyotime) for 30 min. After washing with water, the colonies were imaged and quantified. The cell colonies were counted using ImageJ software.

Transwell migration assay
HCT116 cells (2 × 105) were placed into the upper chamber of the Transwell insert (8-mm pore size; Corning) containing serum-free medium. Medium containing 10% FBS was added to the lower chamber to act as a chemical attractant. After a 48-h incubation at 37 °C, 5% CO2, the Transwell chamber was removed. The medium was discarded, and the cells were washed with phosphate-buffered saline (PBS). The cells were then fixed with a 4% paraformaldehyde solution for 30 min and stained with crystal violet for another 30 min. Finally, the non-migrating cells in the upper chamber were gently detached with a cotton swab and counted under a microscope.

Western blotting
Cells were lysed using RIPA buffer, and the total protein concentration was determined using the Bradford method. After adjusting the protein concentration of each sample, the proteins were denatured for 5 min at 100 °C in the presence of 5× loading buffer. The expression of target proteins was determined by Western blotting. Antibodies targeting the following proteins were used: LC3B (3868S, CST), p62 (PM045, MBL), caspase-3 (9662S, CST), cleaved-caspase-8 (8592 T, CST).

Annexin V-FITC/PI apoptosis detection assay
HCT116 cells were seeded in 6-well plates at a density of 2 × 105 cells/well and allowed to attach overnight. Cells were then treated for 48 h with N8 (50 μM), Amlexanox (100 μM), or DMSO as a solvent control. After treatment, both floating and adherent cells were collected and washed with PBS. Cell apoptosis was detected using an Annexin V-FITC Apoptosis Detection Kit (Beyotime, Shanghai, China). The cells were then resuspended in 195 μL of Annexin V-FITC binding buffer. Subsequently, 5 μL of Annexin V-FITC and 10 μL of propidium iodide (PI) were added for staining. The samples were then incubated on ice in the dark and analysed immediately by a BD FACSymphony™ A1 flow cytometer (BD Biosciences, CA, USA). The percentage of apoptotic cells was determined by summing the percentages of cells in the lower-right (early apoptosis, Annexin V+/PI−) and upper-right (late apoptosis, Annexin V+/PI+) quadrants.

pCMV-mCherry-GFP-LC3 reporter assay
HCT116 cells were seeded on 35-mm glass-bottomed dishes and transfected with the pCMV-mCherry-GFP-LC3 plasmid (Beyotime, Shanghai, China) using UltraFection 3.0 transfection reagent (GLPBIO, Suzhou, China). After 24 h, the cells were treated with N8 (50 µM) or amlexanox (100 µM) for 48 h. Fluorescence microscopy was used for observation, capturing images with channels for GFP (488 nm, green), mCherry (594 nm, red), and DAPI (405 nm, blue) to stain nuclei. For quantification, at least 30 cells were selected for each group. The number of red and yellow puncta per cell was counted, and the average number of puncta per cell was statistically analysed.

Statistical analysis
All experiments were conducted at least three times and the data were presented as the mean ± standard deviation (SD). One-way analysis of variance (ANOVA) followed by Dunnett’s post-hoc test was used to determine the significance of differences between multiple experimental groups and a single control group. A p values < 0.05 was considered statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Results

Identification of N8 using high-throughput screening
The IKKε kinase domain model was generated by AlphaFold, focusing on the N-terminal domain (residues 1–310). The per-residue confidence score (pLDDT) was used to evaluate model performance. The majority of residues in the IKKε kinase domain showed a high pLDDT score (Figure 1A), indicating that the model is of high quality. To optimise the IKKε kinase domain model for virtual screening, we first performed MD simulations in the presence of C35, a known co-crystal inhibitor of IKKε homologue. The IKKε-C35 complex was subjected to 100-ns simulations under constant pressure and temperature (NPT) using AMBER 20 with CUDA acceleration. Both the IKKε kinase domain and C35 components remained stable within the complex, suggesting a robust model (Figure S1). The RMSD between the predicted IKKε structure and the closest homologous experimental structure, TBK1, is 1.54 Å (Figure 1B). Ramachandran plot analyses of both the AlphaFold-predicted IKKε structure and the MD-optimised representative structure reveal no unfavourable phi-psi dihedral angles (Figure S2). These results confirm the high quality and reliability of the predicted IKKε structure.
A representative structure of the IKKε-C35 complex was obtained for use in further virtual screening experiments. In these subsequent, structure-based virtual screening experiments, a library of 12 million compounds was used (Supplemental Table 1). After a series of drug-likeness filtering steps (Figure 1C), 27 compounds were selected as final hits. Compound N8, which had the lowest docking energy scores and the highest level of drug-likeness was selected for further investigation.
Compound N8 (Figure 2A) formed strong interactions with IKKε. It had a binding energy of −12.87 kcal/mol, while the co-crystal inhibitor C35 and a known IKKε inhibitor, BAY-985, had binding energies of −13.12 kcal/mol and −12.3 kcal/mol, respectively. The 2D protein-ligand interactions analysis revealed that the core fragment of all these three compounds have formed hydrogen bonds with the backbone of Cys89 (C89) and Lys38 (K38), a key residue in the hinge region of the kinase domain (Figure 2B and 2C). The inhibitory effect of N8 on IKKε was initially confirmed through an in vitro kinase activity assay. We synthesised a reported IKKε substrate peptide and utilised the ADP-Glo Kinase Assay—a universal biochemical assay suitable for studying all kinases—to analyse the inhibitory effect of N8 on IKKε kinase activity. Notably, N8 demonstrated an ability to inhibit kinase activity comparable to that of the well-established IKKε inhibitor, amlexanox (Figure 2D).
To evaluate the specificity of N8 for IKKε, we performed molecular docking studies comparing its binding to IKKε and TBK1, given their high degree of homology. Structural alignment of the predicted IKKε model with the experimental TBK1 structure (PDB ID: 5W5V) revealed an RMSD of 1.54 Å (Figure 1B), indicating significant structural similarity. Docking analysis showed that N8 binds to IKKε with a binding energy of −12.87 kcal/mol, compared to −12.26 kcal/mol for TBK1, suggesting a slightly higher affinity for IKKε. This modest difference in binding energy indicates that N8 may preferentially target IKKε but could exhibit cross-reactivity with TBK1. To further elucidate N8’s selectivity profile, future studies will involve comprehensive kinase panel screening to assess its activity against a broader range of kinases, including TBK1 and other IKK family members.

Compound N8 effectively reduces the viability of several cancer cell lines
To further assess the anticancer activity of N8, we used to treat several cancer cell lines, including HCT116 (colorectal cancer), HepG2 (hepatocellular carcinoma), T24 (bladder cancer), MDA-MB231 (breast cancer), A549 (lung cancer), and Hela (cervical cancer). N8 significantly reduced the viability of all six cancer cell lines in a concentration-dependent manner (Figure 3, S3-S5). Moreover, N8 demonstrated lower IC50 values than amlexanox in all tested cell lines and showed comparable potency to the tested BAY-985 in HCT116, HepG2, T24, and MDA-MB-231 cells, indicating that it was more effective at suppressing cancer cell viability (Table 1, Figure S4 and S5).

N8 effectively inhibits the proliferation of cancer cells
A colony formation assay was next performed to evaluate the effects of N8 on cell proliferation. N8 and amlexanox significantly decreased the ability of the HCT116 and HepG2 cell lines to form colonies in a dose-dependent manner. (Figure 4). N8 was more potent and effective at lower concentrations compared to amlexanox, making it a stronger inhibitor of colony formation. Amlexanox, while still effective, requires higher doses to achieve the same level of inhibition as N8. These results confirmed that N8 effectively inhibits the proliferation of cancer cells.

N8 inhibits the migration of cancer cells
We next performed a transwell migration assay to evaluate the effect of N8 on the migration of cancer cells. N8 treatment significantly decreased cancer cell migration relative to the DMSO vehicle control in a concentration-dependent manner (Figure 5). Moreover, N8 has a more pronounced effect on reducing cell migration at 12.5 µM compared to amlexanox, which shows a less significant reduction in migration at both tested concentrations. These results demonstrate that N8 effectively inhibits the migration of cancer cells.

N8 does not induce apoptosis
To investigate the potential mechanism of N8’s anticancer activity, we first analysed its effect on cell apoptosis. In this experiment, the effect of N8 on the expression of apoptotic markers in HCT116 cells was examined. The results from the Annexin V/PI apoptosis detection assay demonstrate that N8 treatment significantly increased the populations of late apoptotic/necrotic cells (Annexin V+/PI+) and cells with compromised membrane integrity (Annexin V−/PI+) without inducing a significant rise in early apoptotic cells (Annexin V+/PI−) (Figure 6A). The Western blot results show that, after 48 h of treatment with increasing concentrations of N8 or amlexanox, the expression of cleaved caspase 3 and cleaved caspase 8 did not show a significant concentration-dependent increase for N8 or amlexanox (Figure 6B and 6C). These findings are consistent with previous reports indicating that N8 does not primarily induce apoptosis in HCT116 cells. Instead, the lack of significant activation of apoptotic markers such as cleaved caspases suggests that N8 may promote cell death through other mechanisms.

N8 affects cancer cell autophagy
Autophagy is a process used by cells to remove damaged organelles and maintain normal cellular function, and it can also influence the proliferation and migration of cells. We investigated the effect of N8 on autophagy processes by detecting the typical autophagy markers LC3B and p62. Our results showed a significant increase in the LC3B-II to LC3B-I ratio following N8 treatment, indicating an increase in autophagosome formation (Figure 7A). Concurrently, p62 levels accumulated in a dose-dependent manner (Figure 7B), suggesting that autophagic flux was blocked. To confirm this, a pCMV-mCherry-GFP-LC3 reporter assay was performed. The assay revealed a significant increase in yellow puncta (autophagosomes) in N8-treated cells, while the number of red puncta (autolysosomes) remained unchanged. This data indicates that while N8 induces autophagosome formation, it also impairs the subsequent fusion of autophagosomes and lysosomes, thereby disrupting autophagic flux (Figure 7C). This finding is consistent with our Western blot results for LC3 and p62. In contrast, cells treated with amlexanox showed no significant increase in either yellow or red puncta, which is consistent with its lower inhibitory effect. These findings confirm that N8 significantly induces autophagy but also impairs the fusion of autophagosomes with lysosomes, disrupting autophagic flux.

Discussion
Kinase-structure-based virtual screening is a widely utilised and effective approach in the development of kinase inhibitors38. In this study, we virtually screened IKKε kinase structures using various compound libraries (especially novel compound libraries, kinase compound libraries, and tumour compound libraries) to identify N8 as a novel inhibitor of IKKε kinase activity. The IKKε-inhibiting activity of N8 was confirmed in the in vitro kinase activity assay, demonstrating its potent action on IKKε kinase activity. Additionally, N8 effectively reduced the viability and proliferation of a range of cancer cell lines, including those derived from colorectal, liver, bladder, breast, lung, and cervical cancers. Notably, N8 exhibited greater inhibition of proliferation and migration in colorectal cancer cells compared to amlexanox, a clinically approved IKKε inhibitor used for treating aphthous ulcers and asthma.
IKKε plays a crucial role in promoting growth, metastasis and drug resistance of various cancers16,39, thereby making it an attractive target for cancer therapy. Pharmacological inhibition of IKKε by inhibitors has been demonstrated to effectively suppress proliferation, migration, and invasion in several cancer cells40–42. IKKε is involved in regulating multiple molecular pathways associated with cancer cell proliferation. Recent evidence suggests that IKKε modulates cancer cell proliferation through the regulation of autophagy43. The underlying mechanisms by which IKKε inhibitors influence autophagy are complex, involving a network of pathways and molecular interactions. For example, the IKKε inhibitor amlexanox has been reported to suppress autophagy without affecting cell cycle progression or inducing apoptosis in melanoma cells44. Furthermore, amlexanox modulates autophagic flux by influencing the expression of key proteins such as LC3 and the autophagy adaptor protein p62, as evidenced by an increased LC3-II/LC3-I ratio and a decrease in p62 levels in ovarian cancer cells41.
It is noteworthy that the role of amlexanox in autophagy is context-dependent, exhibiting both suppressive and inductive effects across different studies. This discrepancy may stem from cell-type-specific signalling backgrounds and the multi-target nature of amlexanox, where off-target effects could exert opposing influences on autophagy. In contrast, our compound N8 demonstrates a consistent and potent phenotype of autophagic inhibition in our models. Specifically, N8 induced cell death in HCT116 colorectal cancer cells by blocking autophagic flux, as evidenced by the accumulation of LC3-II and p62, rather than through apoptosis. This suggests a potentially novel and more reliable mechanism for N8 in exerting its anticancer effects via IKKε inhibition in colorectal cancer. However, while our data strongly suggest that N8-induced cytotoxicity results from inhibited autophagic flux, definitive proof of causality requires direct rescue experiments. To unequivocally validate this mechanism, future work will employ co-treatment strategies with established autophagy modulators. This will involve treating cells with N8 alongside autophagy inducers such as rapamycin. A significant attenuation of cell death upon co-treatment with rapamycin would provide powerful evidence for causality. Conversely, we would expect synergistic cytotoxicity when N8 is combined with late-stage autophagy inhibitors like chloroquine or 3-MA. Validating this mechanistic causality in vitro will be paramount before advancing to in vivo efficacy models.
Previous research has established that IKKε activates mTOR signalling by inhibiting its negative regulator, TSC1. This activation, in turn, suppresses autophagy through ULK1 phosphorylation45. Furthermore, IKKε deficiency has been shown to reduce mTOR signalling and induce autophagy in cardiomyocytes46. While our data indicate that N8 significantly induces autophagy, it also markedly disrupts autophagic flux in HCT116 cells, suggesting a complex mechanism. The precise upstream pathways involved, however, require further investigation. Given the complexity of the mechanisms outlined above, it is crucial to advance this research into in vivo models. Currently, the translational potential of N8 is constrained by the lack of in vivo data, such as pharmacokinetic profile, systemic toxicity, and antitumor efficacy in animal models. As demonstrated in studies of other IKKε inhibitors like amlexanox, robust efficacy in xenograft models is a crucial step in validating target engagement and therapeutic potential. Future investigations will prioritise establishing patient-derived xenograft (PDX) or cell-derived xenograft (CDX) models of colorectal cancer to evaluate the tumour-suppressive effects of N8 in vivo. Furthermore, comprehensive pharmacokinetic and toxicological studies are essential to determine a safe and effective dosing regimen, bridging our promising in vitro results to potential clinical applications. The molecular weight of N8 is 416.51, with a predicted logP of 4.48, making it a versatile scaffold for optimising IKKε inhibition. Its key features include: (1) aromatic coumarin and phenyl rings, offering rigid platforms for substitutions to adjust electronic properties, hydrophobicity, and π-π interactions with IKKε; (2) an amide linkage, connecting pharmacophores with subtle modification points for conformational flexibility and hydrogen bonding; (3) a piperidine/decahydroquinoline moiety with a basic tertiary amine and stereocenters, enabling steric fit and stereoisomerism exploration; and (4) a flexible ethylamine linker, allowing conformational adjustments and spatial optimisation in the IKKε active site. These attributes provide multiple handles for systematic chemical derivatization, supporting comprehensive SAR studies to develop more potent, selective, and bioavailable IKKε inhibitors. Future efforts will involve synthesising N8 analogs to further enhance its therapeutic potential against IKKε-overexpressing cancers. It is important to note that the current study primarily focuses on IKKε, and the kinase selectivity profile of N8, particularly against the highly homologous TBK1, has not been fully established. While molecular docking suggested a potential preference for IKKε, comprehensive kinase panel profiling in future studies will be essential to validate its selectivity and define its potential as a selective IKKε inhibitor.
In conclusion, our results indicate that N8 is a promising IKKε inhibitor with significant anticancer potential, and its effects warrant further exploration in animal models and clinical studies to assess its applicability as a therapeutic agent.

Supplementary Material

Supplementary materials.docx