The IKKα-regulated microRNA miR-9-5p mediates lung cancer growth and invasiveness via CDH1/Wnt/β-catenin signalling Authors: Simoni Besta, Eugenia Roupakia, Zoi Kanaki, Giannis Vatsellas, Mark Samuels, Spyros Foutadakis, Maria Tokamani, Alessandro Barbon, Raphael Sandaltzopoulos, Dimitris C. Kanellis, Jiri Bartek, Maria Hatziapostolou, Christos Polytarchou, Dimitris Thanos, Apostolos Klinakis, Evangelos Kolettas Published: 2026 ABSTRACT IKKα is a serine/threonine kinase that acts as a tumour suppressor in non-small cell lung cancer (NSCLC) in an NF-κΒ-independent manner, however, little is known about its downstream signalling pathways. To interrogate the IKKα signalling network in NSCLC, we investigated the impact of IKKα silencing on the miRNA expression profile of NSCLC cells. Nanostring miRNome analysis identified miR-9-5p, a known oncomir, as the top upregulated miRNA upon IKKα depletion. We show that overexpression of miR-9-5p in human lung cancer cells increases cell migration and invasion and promotes epithelial-to-mesenchymal cell transition (EMT) through loss of E-cadherin and activation of the Akt1/β-catenin pathway. In vivo tumour xenograft models show that overexpression of miR-9-5p promotes tumour growth and widespread transcriptomic reprogramming, altering the expression of genes related to EMT, apoptosis and NF-κB signalling. Overall, we show that miR-9-5p acts as an oncogenic effector upon IKKα loss, promoting tumour progression through EMT and Akt-1/GSK-3β/β-catenin pathway activation, leading to increased invasiveness and tumour growth. Introduction Lung cancer is the leading cause of cancer death and the most common cancer type worldwide [1]. Non-small cell lung cancer (NSCLC) is the most common lung cancer type, and is characterised by different histopathological subtypes, with lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) being the major subtypes. Despite recent advances in targeted therapies, survival rates for the disease remain low. Understanding the signalling pathways driving lung cancer is therefore essential for the development of novel therapies. The nuclear factor kappa B (NF-κB) signalling pathways and upstream activating Ser/Thr kinases, IκB Kinase α (IKKα) and IκB Kinase β (IKKβ) play key roles in lung cancer onset, development and progression by regulating the expression of target genes. In cancer, NF-κB transcription factors (TFs) and their upstream signalling components exhibit either tumour-promoting or growth-suppressive activities depending on the cellular context [2–4]. NF-κB TFs are part of a regulatory network, and NF-κB responses are typically self-limiting through the induction of negative feedback loops; however, these auto-regulatory loops often become deregulated in cancer [3, 5], contributing to lung tumour initiation, development and progression [6–12]. IKKα activates non-canonical NF-κB signalling, but also has several important NF-κB-independent functions in signal transduction, gene expression, cell cycle regulation and cell death, impacting normal physiology and tumour development [13–16]. These diverse mechanisms suggest that IKKα may regulate oncogenesis through multiple downstream mechanisms. In the nucleus, IKKα regulates the expression of target genes by phosphorylating histone H3, the transcriptional co-repressor silencing mediator of retinoic acid and thyroid hormone receptor (SMRT) and the coactivator CREB-binding protein (CBP), inducing the expression of both NF-κB-dependent and -independent genes [2, 14, 17]. Additionally, most functional studies show that IKKα is a tumour suppressor in both LUAD [8, 18, 19] and LUSC [4, 20]. As such, IKKα is frequently downregulated in lung cancers compared to normal tissue and undergoes hemizygous deletion in approximately 22% of LUAD cases [18, 20]. IKKα loss promotes Kras-initiated NSCLC development through a redox regulatory pathway. This involves upregulation of NOX2 and downregulation of NRF2, leading to ROS accumulation and acceleration of NSCLC [18]. We previously demonstrated that IKKα silencing causes upregulation of HIF-1α and hypoxia-mediated processes, supporting NSCLC growth during hypoxic conditions in vivo in an NF-κΒ-independent manner [8]. However, beyond this, the downstream pathways of IKKα in cancer remain poorly understood. As microRNAs (miRNAs) are increasingly recognised as key players in cancer [21], we explored how IKKα loss may impact the miRNA expression profile of lung cancer cells. By binding to cis-regulatory mRNA elements usually found in the 3’ untranslated region, miRNAs work at the post-transcriptional level, negatively regulating mRNA translation and stability [21, 22]. Interestingly, the interplay between TFs and miRNAs is a common regulatory mechanism in many biological processes, including in NF-κB signalling where it regulates cell proliferation, survival, stress, inflammation, senescence and tumour progression [5, 23, 24]. IKKα is also negatively regulated by miR-15a, miR-16 and miR-223 which bind to the 3′ UTR of the IKKα mRNA, leading to changes in the expression of non-canonical NF-κB target genes during macrophage differentiation [24, 25]. As IKKα is frequently downregulated in lung cancer and accumulating evidence suggests that this event can promote oncogenesis independently of NF-κB signalling, in this study we focused on the effects of IKKα depletion on miRNA expression and their mechanisms of action, rather than the pathway as a whole. To further explore the tumour suppressor functions of IKKα in NSCLC, we investigated the miRNA expression profile of IKKα-deficient lung cancer cells, revealing miR-9-5p as an IKKα-regulated, tumour-promoting miRNA. Furthermore, overexpression of miR-9-5p increases cell migration, invasion and tumour growth, accompanied by modulation of EMT markers and activation of Akt/GSK-3β/β-catenin signalling. In addition, we show that miR-9-5p upregulation produces profound changes on the transcriptome of lung cancer cells, altering the expression of genes involved in diverse cellular processes, including apoptosis, EMT and NF-κB signalling [5]. Overall, we show that IKKα loss drives miR-9-5p upregulation, supporting cancer progression through widespread transcriptomic changes and EMT. Results Expression profiling of IKKα-regulated miRNAs Previous studies from our laboratory have shown that IKKα acts as an important NSCLC tumour suppressor by regulating hypoxia-mediated processes, enhancing tumour growth in vivo [8]. To further investigate the impact of IKKα on lung cancer development, we generated A549 ΙΚΚα knockdown (ΙΚΚαKD) and empty vector control cell lines to study the impact of ΙΚΚα-dependent alterations on the miRNA expression profiles (Fig. 1a). Total RNA was isolated from cells, enriched for miRNAs and then subjected to miRNome profiling using digital molecular barcoding-based Nanostring technology. Data were analysed using the nSolver software [26–28]. In IKKαKD cells, 125 of the panel of 800 miRNAs were expressed at levels above the detection threshold (Table S1). In IKKαKD cells, miR-92b-3p and miR-122-5p exhibited the strongest downregulation, while miR-9-5p and miR-376a-3p exhibited the strongest upregulation compared to control cells. A549 IKKβKD cells (Supplementary Fig. 1a) were analysed in parallel and compared to IKKαKD cells (Fig. 1b, and Table S1). The levels of miR-9-5p were reduced in the IKKβKD cells, compared to control cells, suggesting that IKKβ plays a different role to IKKα in the regulation of miR-9-5p, generating an opposite effect. The common experimentally supported mRNA targets of the differentially expressed miRNAs in the A549 IKKαKD cells were identified using miRNet and TarBase (v9.0) [29, 30] (Supplementary Fig. 1b). Pathway analysis showed enrichment in p53 signalling, FoxO signalling, cell cycle, adherens junctions and cell senescence pathways (Supplementary Fig. 1c). As the top hit, miR-9-5p was selected for further investigation as it has been previously implicated in lung cancer malignancy [31–34] and RT-qPCR validated the upregulation of miR-9-5p in A549 IKKαKD cells (Fig. 1c).Fig. 1Nanostring miRNome analysis of A549 IKKα knockdown cell lines.a Western blot for the expression of IKKα in A549 control and IKKαKD cells. β-actin was used as a loading control. b Heatmap of miRNA expression levels in A549 IKKαKD and IKKβKD cell lines. This figure shows selected miRNAs with a fold change >1.5 or <0.75. c RT-qPCR analysis of miR-9-5p expression levels in A549 IKKαKD compared to control cells. Graphs show mean ± SEM, **p < 0.01 (two-tailed unpaired Student’s t test), n = 3 d Western blot for IKKβ in control and IKKβ-overexpressing cells (IKKβCA). β-actin was used as a loading control. e RT-qPCR analysis of miR-9-5p expression levels in A549 control and IKKβCA cells. Graphs show mean ± SEM, ****p < 0.0001 (two-tailed unpaired Student’s t test), n = 3. f RT-qPCR analysis of A549 control and IKKαKD cells treated with either DMSO, 10 ng/ml TNFα, or 2 μM Sulfasalazine. Graphs show mean ± SEM, ns, non-significant (two-way ANOVA), n = 3. g Western blot for p52/p100 in control and p52KD cells. GAPDH was used as a loading control. h RT-qPCR analysis of miR-9-5p levels in control and p52KD cells. Graphs show mean ± SEM, ****p < 0.0001 (two-tailed unpaired Student’s t test) n = 3. i Western blot for IKKα in control, IKKα knockout (IKKαKO) and IKKαKO cells expressing a kinase-dead mutant IKKα (K44M). β-actin was used as a loading control. j RT-qPCR analysis of miR-9-5p expression levels in A549 control, IKKα knockout and IKKα kinase-dead cells. Graphs show mean ± SEM, **p < 0.01, ****p < 0.0001 (two-way ANOVA), n = 3. k Mean expression of miR-9-5p in tumours compared to normal adjacent tissue, based on TCGA data. Analysis was performed by OncoMir [37]. Student’s t test (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). l KEGG pathway analysis of top miR-9-5p targets. Analysis was performed by miRPath v4.0 using experimentally supported target genes from TarBase v8.0 [30, 38]. FDR-adjusted p-values are shown. Nanostring miRNome analysis of A549 IKKα knockdown cell lines. a Western blot for the expression of IKKα in A549 control and IKKαKD cells. β-actin was used as a loading control. b Heatmap of miRNA expression levels in A549 IKKαKD and IKKβKD cell lines. This figure shows selected miRNAs with a fold change >1.5 or <0.75. c RT-qPCR analysis of miR-9-5p expression levels in A549 IKKαKD compared to control cells. Graphs show mean ± SEM, **p < 0.01 (two-tailed unpaired Student’s t test), n = 3 d Western blot for IKKβ in control and IKKβ-overexpressing cells (IKKβCA). β-actin was used as a loading control. e RT-qPCR analysis of miR-9-5p expression levels in A549 control and IKKβCA cells. Graphs show mean ± SEM, ****p < 0.0001 (two-tailed unpaired Student’s t test), n = 3. f RT-qPCR analysis of A549 control and IKKαKD cells treated with either DMSO, 10 ng/ml TNFα, or 2 μM Sulfasalazine. Graphs show mean ± SEM, ns, non-significant (two-way ANOVA), n = 3. g Western blot for p52/p100 in control and p52KD cells. GAPDH was used as a loading control. h RT-qPCR analysis of miR-9-5p levels in control and p52KD cells. Graphs show mean ± SEM, ****p < 0.0001 (two-tailed unpaired Student’s t test) n = 3. i Western blot for IKKα in control, IKKα knockout (IKKαKO) and IKKαKO cells expressing a kinase-dead mutant IKKα (K44M). β-actin was used as a loading control. j RT-qPCR analysis of miR-9-5p expression levels in A549 control, IKKα knockout and IKKα kinase-dead cells. Graphs show mean ± SEM, **p < 0.01, ****p < 0.0001 (two-way ANOVA), n = 3. k Mean expression of miR-9-5p in tumours compared to normal adjacent tissue, based on TCGA data. Analysis was performed by OncoMir [37]. Student’s t test (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). l KEGG pathway analysis of top miR-9-5p targets. Analysis was performed by miRPath v4.0 using experimentally supported target genes from TarBase v8.0 [30, 38]. FDR-adjusted p-values are shown. To further explore the interplay between NF-κB factors and miR-9-5p expression levels, we generated A549 cells overexpressing a constitutively active form of IKKβ (A549 IKKβCA) (Fig. 1d). This resulted in an increase in miR-9-5p expression compared to control cells, as expected based on the miRNome profiling (Fig. 1e). In addition, A549 control and IKKαKD cells were treated with either TNFα, an inducer of the canonical NF-κB pathway, or sulfasalazine, an inhibitor of IKK/NF-κB signalling [35, 36] (Supplementary Fig. 1d). Neither treatment altered the levels of miR-9-5p in either control or IKKαKD cells, suggesting that, in contrast to chronic canonical NF-κB activation by overexpression of IKKβCA, acute activation or inhibition of the pathway is not sufficient to affect miR-9-5p levels in this system (Fig. 1f). As IKKα is the key regulator of non-canonical NF-κB signalling, we also assessed whether p52 contributes to miR-9-5p regulation. Silencing of p100/p52 in A549 cells caused a reduction in miR-9-5p levels, indicating that p52 positively regulates miR-9-5p expression (Fig. 1g, h). This opposite effect to the increase observed upon IKKα silencing suggests that these factors may regulate miR-9-5p through distinct mechanisms, and that the effect is NF-κB-independent. Finally, we examined whether the kinase activity of IKKα is required for the regulation of miR-9-5p. Using a CRISPR-generated A549 IKKαKO cell line, we re-expressed a kinase-dead mutant of IKKα (IKKαK44M) and measured the levels of miR-9-5p (Fig. 1i, j). Re-expression of IKKαK44M partially rescued the expression of miR-9-5p, suggesting that the regulation of miR-9-5p is at least partly dependent by the kinase activity, although incomplete rescue indicates that additional mechanisms may contribute to IKKα-mediated regulation. Together, these results indicate that IKKα negatively regulates miR-9-5p, whereas other NF-κB signalling components can also impact miR-9-5p levels in lung cancer cells. Next, the OncoMiR database [37] was used to assess the expression of miR-9-5p in different forms of cancer using the Cancer Genome Atlas (TCGA) datasets (Fig. 1k). Interestingly, LUAD and LUSC showed the highest increase in miR-9-5p expression in tumour samples compared to normal tissue among the different cancer types analysed, suggesting a key role in lung cancer development. To explore the biological processes associated with miR-9-5p experimentally supported targets, KEGG pathway analysis was performed using miRPath [38], highlighting adherens junction, focal adhesion and transcriptional misregulation in cancer as among the top enriched terms (Fig. 1l). Taken together, these findings suggest that miR-9-5p upregulation is functionally linked to several oncogenic processes and may have an important role in lung cancer progression. Upregulation of miR-9-5p promotes invasiveness through the regulation of EMT To further explore the role of miR-9-5p in lung cancer, we established a miR-9-5p-overexpressing clonal A549 cell line using Twl-PGKgfp-H1 (control) and Twl-PGKgfp-H1-miR-9-5p (miR-9-5p) lentivectors (Fig. 2a) (Supplementary Fig. 2a) [39]. Our pathway analysis of the top experimentally supported target genes of miR-9-5p showed an enrichment in genes involved in focal adhesion and adherens junctions (Fig. 1f), suggesting alterations in EMT-related processes. To assess this, cell migration and invasion assays were performed, showing that overexpression of miR-9-5p increased migration in scratch assays (Fig. 2b, c) and promoted cell invasion in transwell-based invasion assays (Fig. 2d, e). Immunoblotting showed downregulation of the epithelial cell marker E-cadherin and upregulation of the mesenchymal cell markers N-cadherin and vimentin upon miR-9-5p overexpression (Fig. 2f). E-cadherin is typically bound to β-catenin near the cell membrane, where it forms cadherin cell junctions [40–42]. Overexpression of miR-9-5p resulted in increased levels of active (non-phospho-Ser45) and total β-catenin (Fig. 2g), indicating enhanced β-catenin stability and suggesting activation of the canonical Wnt signalling pathway.Fig. 2Overexpression of miR-9-5p promotes cell invasion and migration and alterations in EMT markers.a RT-qPCR analysis of A549 stable cell lines overexpressing miR-9-5p. Graphs show mean ± SEM. ***p < 0.001 (two-tailed unpaired Student’s t test), n = 3. b Cell migration scratch assay of A549 control and miR-9-5p-overexpressing cell lines. Representative images are shown, taken at 0 h and 24 h after the scratch. c Quantification of the scratch assays. Graphs show mean ± SEM, *p < 0.05 (two-tailed unpaired Student’s t test), n = 3 biological repeats. d Matrigel invasion assay showing increase in invasiveness with miR-9-5p overexpression. Representative images are shown. e Quantification of invasion assays, **p < 0.01 (two-tailed unpaired Student’s t test), n = 3 independent experiments. f Western blot of EMT markers E-cadherin, N-cadherin and vimentin in A549 control and miR-9-5p overexpressing cell lines. β-actin was used as a loading control. g Western blot of proteins related to Wnt signalling in A549 control and miR-9-5p overexpressing cell lines. GAPDH was used as a loading control. Overexpression of miR-9-5p promotes cell invasion and migration and alterations in EMT markers. a RT-qPCR analysis of A549 stable cell lines overexpressing miR-9-5p. Graphs show mean ± SEM. ***p < 0.001 (two-tailed unpaired Student’s t test), n = 3. b Cell migration scratch assay of A549 control and miR-9-5p-overexpressing cell lines. Representative images are shown, taken at 0 h and 24 h after the scratch. c Quantification of the scratch assays. Graphs show mean ± SEM, *p < 0.05 (two-tailed unpaired Student’s t test), n = 3 biological repeats. d Matrigel invasion assay showing increase in invasiveness with miR-9-5p overexpression. Representative images are shown. e Quantification of invasion assays, **p < 0.01 (two-tailed unpaired Student’s t test), n = 3 independent experiments. f Western blot of EMT markers E-cadherin, N-cadherin and vimentin in A549 control and miR-9-5p overexpressing cell lines. β-actin was used as a loading control. g Western blot of proteins related to Wnt signalling in A549 control and miR-9-5p overexpressing cell lines. GAPDH was used as a loading control. Furthermore, we found that levels of inactive phosphorylated (Ser9) glycogen synthase kinase-3β (GSK-3β) were increased with miR-9-5p overexpression, while total GSK-3β remained unchanged, suggesting decreased inhibition of β-catenin. As Akt1 is the primary kinase phosphorylating GSK-3β, we examined total and phosphorylated Akt1 (Ser473) and observed significantly elevated active Akt1 levels in miR-9-5p-overexpressing cells (Fig. 2g). Overall, these results indicate that miR-9-5p overexpression leads to increased migration and invasiveness, EMT induction and increased activation of Akt/β-catenin signalling. Overexpression of miR-9-5p promotes tumour formation in vivo To investigate the role of miR-9-5p in lung tumourigenesis in vivo, we generated tumour xenograft models using A549 and H1299 cells overexpressing miR-9-5p (Supplementary Fig. 2b, c) in immunocompromised NOD Scid Gamma (NSG) mice. Cells were injected bilaterally in the flanks and allowed to grow for 4-6 weeks. The growth of the miR-9-5p overexpressing tumours was significantly faster than the controls (Fig. 3a, b), which is also reflected in the increased tumour weight at the study endpoint (Fig. 3c, d) (Supplementary Fig. 3a). Persistence of miR-9-5p overexpression in the tumour xenografts was also confirmed by RT-qPCR analysis of the endpoint tumours (Supplementary Fig. 3b, c).Fig. 3Upregulation of miR-9-5p is oncogenic in vivo.Tumour growth over time with (a) A549 or (b) H1299 cells, n = 9 for each condition. Results are presented as mean ± SEM *p < 0.05, ***p < 0.001, ****p < 0.0001) (two-way ANOVA with Bonferroni’s multiple comparison test), n = 9 for each condition. Tumour weight of tumour xenografts of (c) A549 and (d) H1299 cells. Results are presented as mean ± SEM (two-tailed unpaired Student’s t test, ***p < 0.001, n = 9 for each condition. Volcano plot of differentially expressed genes in (e) A549 and (f) H1299 miR-9-5p tumours, determined by DESeq2 analysis. Significant hits (n = 4 biological samples for each condition, p-adjusted < 0.05) are depicted as red for upregulated and blue for downregulated genes. Pathway enrichment analysis of downregulated genes in (g) A549 and (h) H1299 miR-9-5p tumours using MSigDB Hallmark gene sets. FDR-adjusted p-values are shown. Upregulation of miR-9-5p is oncogenic in vivo. Tumour growth over time with (a) A549 or (b) H1299 cells, n = 9 for each condition. Results are presented as mean ± SEM *p < 0.05, ***p < 0.001, ****p < 0.0001) (two-way ANOVA with Bonferroni’s multiple comparison test), n = 9 for each condition. Tumour weight of tumour xenografts of (c) A549 and (d) H1299 cells. Results are presented as mean ± SEM (two-tailed unpaired Student’s t test, ***p < 0.001, n = 9 for each condition. Volcano plot of differentially expressed genes in (e) A549 and (f) H1299 miR-9-5p tumours, determined by DESeq2 analysis. Significant hits (n = 4 biological samples for each condition, p-adjusted < 0.05) are depicted as red for upregulated and blue for downregulated genes. Pathway enrichment analysis of downregulated genes in (g) A549 and (h) H1299 miR-9-5p tumours using MSigDB Hallmark gene sets. FDR-adjusted p-values are shown. Next, in order to identify miR-9-5p-dependent changes in gene expression in the tumour samples, mRNA sequencing was performed (Tables S2, S3), highlighting several differentially expressed genes (DEGs) (Fig. 3e, f) (Supplementary Fig. 4a-d). Pathway enrichment analysis of significantly downregulated genes using the MSigDB Hallmark database identified hypoxia, glycolysis, cholesterol homeostasis, p53 signalling and apoptosis in both cell lines. In addition, IFN-α/-γ and TNFα/NF-κB-mediated inflammatory response pathways were enriched in A549 cells, while in H1299 cells, altered pathways included EMT and Hedgehog signalling, an important cell-to-cell communication pathway [43] (Fig. 3g, h). Importantly, IFN-α/-γ and NF-κB signalling are linked to EMT [44, 45]. These results demonstrate that miR-9-5p overexpression accelerates lung tumour growth in vivo and drives gene expression changes affecting several key oncogenic pathways. Analysis of downregulated genes in each cell line showed that miR-9-5p overexpression resulted in significant gene downregulation in both xenograft models, with only a small subset of targets shared between A549 and H1299. Among the 32 common downregulated genes of both cell lines, 18 genes were predicted miR-9-5p targets according to TarBase v.9.0 (Fig. 4a), including the ΕΜT-implicated genes CLDN1, LTBP1 and IGFBP3, which were further validated by RT-qPCR (Fig. 4b, c).Fig. 4Upregulation of miR-9-5p results in widespread transcriptomics changes in lung cancer cells.a Venn diagram depicting the common downregulated genes in A549 and H1299 tumours. Out of 32 common genes, 18 genes are miR-9-5p predicted targets, highlighted in bold. b, c RT-qPCR analysis of CLDN1, LTBP1 and IGFBP3 expression in A549 and H1299 miR-9-5p samples. GAPDH was used as an internal control. Graphs show mean ± SEM, **p < 0.01, ***p < 0.001, ****p < 0.0001 (two-tailed unpaired Student’s t test), n = 3. d, e Significantly enriched EMT related TFs identified by transcription factor motif enrichment analysis of differentially expressed genes with ISMARA in (d) A549 and (e) H1299 miR-9-5p tumours. f Schematic representation of the mechanism through which miR-9-5p promotes oncogenesis in lung cancer cells. Upregulation of miR-9-5p results in widespread transcriptomics changes in lung cancer cells. a Venn diagram depicting the common downregulated genes in A549 and H1299 tumours. Out of 32 common genes, 18 genes are miR-9-5p predicted targets, highlighted in bold. b, c RT-qPCR analysis of CLDN1, LTBP1 and IGFBP3 expression in A549 and H1299 miR-9-5p samples. GAPDH was used as an internal control. Graphs show mean ± SEM, **p < 0.01, ***p < 0.001, ****p < 0.0001 (two-tailed unpaired Student’s t test), n = 3. d, e Significantly enriched EMT related TFs identified by transcription factor motif enrichment analysis of differentially expressed genes with ISMARA in (d) A549 and (e) H1299 miR-9-5p tumours. f Schematic representation of the mechanism through which miR-9-5p promotes oncogenesis in lung cancer cells. Transcription factor motif enrichment analysis on the differentially expressed genes was carried out by Integrated Motif Activity Response Analysis (ISMARA) (Supplementary Fig. 4e, f) [46], showing enrichment in EMT-related TFs, such as ZEB1, TFAP4, HES1, REL and STAT3 in A549 miR-9-5p overexpressing cells (Fig. 4d), and ZEB1, SNAI2 and ZNF281 in H1299 cells (Fig. 4e). Importantly these EMT TFs repress E-cadherin transcription and are linked to NF-κB signalling [44, 45], and HES1 is a known target of miR-9-5p involved in neurogenesis [47]. Notably, the miR-9-5p seed motif was highly enriched in the A549 miR-9-5p overexpressing cells, validating the regulatory effect of miR-9-5p on its target genes in this model (Supplementary Fig. 4e). Overall, our study identifies miR-9-5p as a critical mediator of NSCLC progression, downstream of IKKα loss. Upregulation of miR-9-5p increased migration, invasion and EMT, stabilised β-catenin and activated canonical Wnt signalling through Akt-dependent inhibition of GSK-3β. In vivo, overexpression of miR-9-5p accelerated tumour growth and caused extensive transcriptomic reprogramming. Collectively, these findings establish miR-9-5p as an oncogenic driver in NSCLC and highlight it as a potential therapeutic target to limit tumour progression (Fig. 4f). Discussion IKKα has several NF-κΒ-dependent and independent roles, acting as a tumour suppressor in lung cancer [4, 8, 18–20]. However, the precise role of IKKα in lung cancer is still underexplored and further research is required to elucidate the molecular mechanisms of the tumour suppressive action of IKKα. In this study, we identified several differentially expressed miRNAs following IKKα silencing. Furthermore, most IKKαKD hits did not correlate with IKKβKD hits, suggesting IKK complex-independent regulation (Fig. 1b). Pathway enrichment analysis showed the altered miRNA-targeted genes implicated in the cell cycle, p53 signalling and cellular senescence signalling, which are all highly dysregulated in lung cancer [48] (Supplementary Fig. 1c). The most highly upregulated miRNA in IKKαKD cells was the oncogenic, NF-κB-regulated miRNA, miR-9-5p [5, 21]. Expression of miR-9-5p is induced by pro-inflammatory stimulation in human monocytes [24, 49]. Furthermore, miR-9-5p targets NFKB1, which encodes the p105/p50 precursor subunit and renders lung cancer cells sensitive to ionising radiation [49], suggesting that miR-9-5p is part of a negative feedback loop mechanism regulating the inflammatory responses [5, 24]. We next assessed whether IKKα regulates miR-9-5p expression through canonical or non-canonical NF-κB signalling. Constitutive activation of IKKβ increased miR-9-5p levels in A549 cells, while inflammatory stimulation with TNFα or treatment with the IKK inhibitor sulfasalazine did not alter miR-9-5p levels (Fig. 1d–f). These findings suggest that acute activation or inhibition of the canonical pathway do not affect miR-9-5p expression; however, sustained activation through IKKβ can promote miR-9-5p expression. This observation is consistent with cancer-associated NF-κB signalling, where a sustained activation, rather than transient pathway activation, drives a pro-tumourigenic transcriptional program [50]. As IKKα is a key regulator of the non-canonical NF-κB pathway, we assessed miR-9-5p expression following p52 depletion. Knockdown of p52 reduced miR-9-5p levels, suggesting that p52 positively regulates miR-9-5p expression (Fig. 1g, h). This contrasts with the increase in miR-9-5p observed upon IKKα depletion, indicating that IKKα is unlikely to regulate miR-9-5p through p52 and may act through a distinct mechanism. To further explore the mechanism of IKKα-mediated regulation, we re-expressed a kinase-dead IKKα mutant (K44M) in IKKα knockout cells. This resulted in partial rescue of miR-9-5p levels, suggesting that the kinase activity may contribute to the regulation of miR-9-5p (Fig. 1i, j). However, the incomplete rescue indicates that additional mechanisms may be involved. Consistent with this, IKKα has previously been reported to mediate transcriptional regulation of target genes through kinase-independent roles [51]. Collectively, these findings indicate that IKKα negatively regulates miR-9-5p in lung cancer cells, while other NF-κB pathway components, including IKKβ and p52, promote its expression in this context. In lung cancer, miR-9-5p is often overexpressed and associates with increased metastasis [31, 52], however the complete mechanism is yet to be elucidated, therefore, we decided to further study the effects of miR-9-5p in NSCLC development. We found that LUAD displayed the highest upregulation of miR-9-5p in TCGA cancer samples compared to normal tissue among the different forms of cancer (Fig. 1k). Pathway enrichment analysis of the top experimentally supported target genes of miR-9-5p pointed to terms such as adherens junction and focal adhesion pathways (Fig. 1l), and experimentally, miR-9-5p upregulation significantly increased cell migration and invasion, linking miR-9-5p to an enhanced EMT phenotype in NSCLC cells (Fig. 3b, e). These results agree with previously published studies, where miR-9-5p promotes cell invasion and EMT in diverse cancer types [21, 53–57]. In many cancers, including lung cancer, loss of E-cadherin is a critical marker of EMT [58]. E-cadherin is a miR-9-5p target in breast cancer [53], colon cancer [21, 54], cervical cancer [59] and osteosarcoma [56], Here, we show that E-cadherin protein levels decrease in A549 miR-9-5p overexpressing cells (Fig. 2b, e). While E-cadherin levels decreased, N-cadherin and vimentin protein levels were elevated (Fig. 2f). The switch from E-cadherin to N-cadherin expression in epithelial cells is highly associated with EMT and with an invasive phenotype [45, 60]. Furthermore, vimentin upregulation promotes cell migration by supporting cytoskeletal reorganisation [61] and is associated with poor prognosis and metastasis in several cancers, including lung cancer [62]. In addition, miR-9-5p overexpressing cells showed higher levels of active Akt1 and β-catenin, while GSK-3β, a negative regulator of β-catenin [63], was phosphorylated and inactivated, indicating altogether upregulation of Akt1/GSK-3β/β-catenin pathway (Fig. 2g). Wnt signalling regulates cell fate determination and organogenesis during embryonic development, and is often activated in cancer, where it promotes proliferation and migration. β-catenin is the main downstream effector of the pathway, and when activated, it translocates to the nucleus where it induces expression of EMT-related genes [64]. Therefore, these results suggest that upregulation of miR-9-5p promotes enhanced EMT characteristics possibly through the Wnt pathway. Our in vivo models showed accelerated tumour growth with miR-9-5p overexpression in both cell types, underscoring its potential as a key promoter of NSCLC malignancy. Other studies have demonstrated the tumour-promoting role of miR-9-5p in vivo in different cancer types, such as mouse glioma [65], Hodgkin’s lymphoma [66], and breast cancer [53]. In breast cancer, miR-9-5p promotes growth, angiogenesis and metastasis by targeting CDH1. E-cadherin downregulation by miR-9-5p also activates β-catenin signalling, upregulating VEGF-A expression and increasing angiogenesis, while overexpression of miR-9-5p in non-metastatic breast tumour cells enables them to form lung micrometastases in mice [53]. This current study provides in vivo evidence that elevated miR-9-5p expression both enhances tumour growth and orchestrates extensive transcriptomic changes consistent with a more aggressive phenotype. Through RNA sequencing of the resulting tumour xenografts, we showed that the upregulation miR-9-5p led to the dysregulation of genes involved in EMT and NF-κB TF signalling [3, 5, 44, 45, 67]. The limited overlap of downregulated genes between the two cell lines likely reflects their distinct genetic and molecular backgrounds, which influence baseline gene expression, pathway activation, and miRNA target availability. Mechanistically, our study supports a model where miR-9-5p upregulation contributes to EMT and tumour progression partially by stabilising β-catenin through targeting CDH1, and by Akt-mediated inhibition of GSK-3β, inducing canonical Wnt signalling. This is further supported by the TF motif enrichment analysis of the DEGs, which showed activation of the TFs ZEB1, SNAI2, REL, STAT3, and TFAP4, which are well-established regulators of EMT and tumour plasticity [44, 68]. Importantly, ZEB1 and Slug directly inhibit CDH1 transcription. These crucial EMT regulators are direct or indirect CDH1 transcriptional repressors and downstream NF-κB targets and induce EMT, migration and invasion [69, 70]. Furthermore, the miR-9-5p seed motif itself was highly enriched within the transcriptomic landscape of A549 xenografts, suggesting a direct regulatory role of this miRNA on a substantial fraction of the dysregulated genes. Importantly, several of these genes, including CLDN1, LTBP1, and IGFBP3, have established roles in maintaining epithelial integrity and suppressing metastasis, suggesting their downregulation may be critical for the invasive phenotype conferred by miR-9-5p. Claudin-1 (CLDN1) is a tight junction protein that suppresses metastasis in LUAD, while low levels of claudin-1 are associated with poorer prognosis [71, 72]. Latent TGF-β-binding protein 1 (LTBP1) is a large extracellular matrix protein and regulator of TGF-β bioavailability [73]. In esophageal squamous cell carcinoma, LTBP1 promotes EMT via TGF-β signalling and high levels correlate with lymph node metastasis [74], however, studies in other cancer types show conflicting roles [75]. Insulin‑like growth factor binding protein 3 (IGFBP3) acts as the main carrier of insulin-like growth factor (IGF) in blood circulation. In lung cancer, IGFBP3 inhibits MMP-1 release, reducing invasiveness, while increased plasma IGFBP3 correlates with favourable diagnosis [76]. Therefore, the pro-invasive phenotype of miR-9-5p overexpressing cells could be, in part, due to the regulation of these target genes, as well as the induction of EMT transcription factors. Conclusion This study aimed to explore the miRNAs regulated by IKKα to better understand its downstream signalling and tumour suppressor activity in NSCLC. Our results demonstrate that IKKα negatively regulates the expression of miR-9-5p, a tumour promoting miRNA in lung cancer. We showed that upregulation of miR-9-5p promotes EMT by downregulating E-cadherin and by activating the Akt1/GSK-3β/β-catenin pathway. Phenotypically, overexpression of miR-9-5p increased cell migration and invasion in vitro and tumour growth in vivo. Transcriptomics analysis of the tumour xenografts revealed widespread changes in gene expression, favouring tumour growth and EMT. Overall, these data identify a novel IKKα-regulated miRNA and demonstrate how miR-9-5p contributes to EMT and transcriptomic changes in lung cancer. Methods Cell culture A549, NCI-H1299, Phoenix and HEK293T cells were obtained from American Type Culture Collection (ATCC). Cells were recently authenticated and regularly tested for contamination. A549 and H1299 cells were cultured in low glucose DMEM (Sigma #D6046-500ML) and Phoenix and HEK293T cells were cultured in high glucose DMEM (Sigma #D6429-500), all supplemented with 10% Fetal Bovine Serum (FBS) (Biowest #S181B-500), 1% Penicillin-Streptomycin (Biowest #L0022-100) and 2 mM L-Glutamine (Biowest #X0550-100). Cells were maintained in a humidified incubator, with 5% CO2 at 37 °C. Establishment of stable cell lines To establish IKKαKD, IKKβKD and p52KD cell lines, the pSUPER.retro vector was used containing shIKKα, shIKKβ or shp52, and non-targeting shRNA-containing pSUPER.retro plasmid was used to generate control cells [77–80]. To establish the A549 IKKβCΑ (constitutively active) cell line, pCLXSN-IKKβCA carrying a constitutively active T-loop mutant form of human IKKβ was used [81]. pCLXSN-ires-GFP was used as a control. Cells were selected with 1 μg/ml puromycin for 1 week. To generate miR-9-5p overexpressing cell lines, the lentiviral vectors Twl-PGKgfp-H1 and Twl-PGKgfp-H1-miR-9-5p were used [39]. To generate the A549 IKKαKO cell line, the lenti-vector pLentiCRISPR v2 containing IKKα sgRNA was used, and the empty vector was used as a control. The vectors were kindly gifted by Dr Ana O’Loghlen [82]. Following lentiviral transduction, the cells were selected with puromycin and single clone colonies were generated. For the establishment of kinase-dead IKKα cell lines (IKKαK44M), the retro-vector pBIN-IKKαK44M-HA, expressing a mouse IKKα with a lysine 44 to methionine mutation, was used in A549 IKKαKO cells and neomycin was used for selection [83]. Retroviral particles were produced by Phoenix cells as previously described [8, 79]. Lentiviral particles were produced by HEK293T cells. Cell treatments For canonical NF-κB pathway activation, cells were treated with 10 ng/ml TNF-α (MedChemExpress #HY-P7058) for 2 h for western blot analysis and 24 h for RT-qPCR. For the inhibition of the canonical NF-κB pathway, cells were treated with 2 μM sulfasalazine (MedChemExpress #HY-14655) for 2 h for western blot analysis and 24 h for RT-qPCR. Nanostring microRNA analysis NanoString nCounter technology was used to examine miRNA expression [26]. 100 ng of total RNA for each sample was used, and RNA quality was assessed using a Bioanalyser. Full details are included in the supplementary information. Real time quantitative polymerase chain reaction (RT-qPCR) Cells were lysed using the QIAzol Lysis Reagent (Qiagen #1023537) following the manufacturer’s instructions. Total RNA was extracted using the miRNeasy Mini Kit (Qiagen #217004). For cDNA synthesis, reverse transcription reaction was carried out using the miRCURY LNA RT Kit (Qiagen #339340) according to manufacturer’s instructions. RT-qPCR reactions were performed using the miRCURY LNA SYBR Green PCR Kit (Qiagen #339346). Fold change was calculated using the 2-ΔΔCt method. Cell lysis and western blotting Cell lysates were prepared using RIPA buffer for 60 min at 4 °C. Cell lysates were then centrifuged for 15 min at 13,000 × g at 4 °C and the pellets were discarded. 40 μg of protein samples were analysed by SDS-PAGE followed by immunoblotting. Immunoblots were developed with an ECL detection kit (BioRad #170–5061) using the Imager Molecular Imager® Chemi Doc™ XRS (BioRad). Wound healing assay 2 × 105 cells were seeded per well of a 24-well plate. After 24 h, a scratch was created using a sterile pipette tip. The cells were then washed in PBS and allowed to grow in serum-free media for 24 h. Images were taken at the beginning and end of the growth period and ImageJ was used to analyse wound closure. Matrigel invasion assay Matrigel (Corning #356230; 80 μl, 1:5 in serum-free media) was added to transwell inserts (24 well, 8 μm pore size, Sarstedt #83.3932.800) and allowed to solidify at 37 °C for 2 h. 4 × 104 cells were seeded in 300 μl serum-free media into the inserts, with 1 ml of media containing 20% FBS in the wells below. After 48 h, the cells were washed with PBS, fixed in 4% PFA for 15 min, then stained with 1% crystal violet in 20% ethanol for 20 min. Inserts were washed thoroughly and 10 representative images were captured per insert using an EVOS microscope. Infiltrating cells were quantified with ImageJ, based on mean intensity. Xenograft mouse models 4 × 106 cells in 200 μl of PBS were subcutaneously inoculated into male immunocompromised NSG (NOD-SCID-IL2Rgamma) mice aged 5 weeks. Each animal was inoculated bilaterally, with modified cells (left side for control Twl-PGKgfp-H1 A549 and H1299 and right side for TwI-miR-9-5p A549 and H1299 cells, n = 9 per cell line), therefore, randomisation to experimental groups was not applicable. Mice were sacrificed after 4 and 6 weeks for H1299 and A549, respectively after cell inoculation. The isolated tumours were weighed and immediately snap-frozen in liquid nitrogen, then stored at −80 °C for RNA isolation. RNA sequencing For total RNA isolation from tumours grown from control and miR-9-5p A549 and H1299 cells, mortar and pestle disruption with liquid nitrogen was used to grind tissue into fine powder. For each sample, 50 mg of tissue was lysed with 700 μl QIAzol lysis reagent. Total RNA was extracted as previously described (n = 4 for each condition). RNA-Seq analysis was performed at the Greek Genome Centre in BRFAA. Differential gene expression analysis was performed using DESeq2 [84–87], followed by pathway enrichment analysis using ShinyGO [88]. Full details are provided in the supplementary information. Statistical analysis and experimental design GraphPad Prism 10 software was used for statistical data analysis. A two-tailed Student’s t-test or analysis of variance (ANOVA) was performed as appropriate, and *p values *<0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 were considered significant. Data are presented as mean ± the standard error of the mean (SEM). Individual data points are shown where possible. Data distribution was assumed to be normal and variance was similar between tested groups. Sample sizes were selected based on previous similar studies. Investigators were not blinded while conducting animal experiments, and no animals were excluded from analysis. Statistical tests and p-values are included in the figure legends. Supplementary information Supplementary Figures Supplementary Information Supplementary Table 1 Supplementary Table 2 Supplementary Table 3 Original uncropped western blots Supplementary Figures Supplementary Information Supplementary Table 1 Supplementary Table 2 Supplementary Table 3 Original uncropped western blots