Nuclear KRT19 links the NF-κB-FSCN1 signaling to gastric cancer metastasis

Abstract

Tumor metastasis is the leading cause of cancer-related mortality, yet the contribution of keratins to this process remains incompletely understood. Here, integrated single-cell and bulk transcriptomic analyses identify keratin 19 (KRT19) within a migration-activated epithelial program as a gene of potential functional relevance. Clinical datasets and tissue microarrays show that KRT19 expression is markedly elevated in gastric cancer and is strongly associated with aggressive pathological features. Functional assays demonstrate that KRT19 depletion impairs cellular migration, invasion, and three-dimensional spheroid infiltration, while while intrasplenic injection of KRT19-silenced cells into male BALB/c nude mice significantly reduces hepatic metastatic colonization, as monitored by serial bioluminescent imaging. Mechanistically, nuclear KRT19 interacts with hnRNPU to facilitate β-TrCP-mediated ubiquitination and degradation of IκBα, thereby sustaining NF-κB activity. Activated NF-κB directly engages the FSCN1 promoter and enhances its transcription, and FSCN1 restoration partially rescues the migratory and metastatic deficits caused by KRT19 knockdown. Together, these findings define a signaling cascade in which an intermediate filament rewires transcriptional programs to control cytoskeletal remodeling and motility, providing mechanistic insight into gastric cancer dissemination and highlighting epithelial structural proteins as underappreciated drivers of metastasis.

Nuclear KRT19 drives gastric cancer metastasis by activating the NF-κB-FSCN1 axis, revealing how an intermediate filament rewires transcriptional and cytoskeletal programs to promote tumor cell motility.

Introduction

Metastasis represents a primary cause of therapeutic failure in cancer patients, involving intricate signaling pathways and complex interplay among diverse cell populations^1,2. A pivotal phase in tumoral metastasis entails the migratory and invasive behaviors of malignant cells³. Through precise regulation of cytoskeletal actin filaments and microtubule dynamic rearrangement kinetics, neoplastic entities breach the basement membrane barrier, accessing vascular and lymphatic conduits as pathological invasion routes, ultimately establishing secondary malignant growths within distant organs⁴. The cytoskeleton serves as the fundamental scaffold for cellular structure and function. Compared to microtubules and actin filaments, our understanding of the third major cytoskeletal component—intermediate filaments—remains relatively limited. The intermediate filaments, as essential constituents of the cytoskeleton, fulfill vital physiological functions by preserving cellular stability while simultaneously orchestrating pivotal roles during cellular morphological transformations⁵. Notably, this morphodynamic plasticity underpins tumor cell migration and invasion. Mounting evidence in recent years indicates that altered expression patterns of specific intermediate filament proteins bear significant correlations with neoplastic cells’ motile aptitude and metastatic potential, presaging their paramount functional nexus within the cytoskeleton-mediated migration-invasion networks^6–8. Nevertheless, precise molecular mechanisms through which intermediate filament proteins orchestrate cellular motility and invasion remain shrouded in tantalizing obscurity. Elucidating their governing principles holds transcendent significance for comprehending the fundamental nature of neoplastic dissemination and uncovering prospective therapeutic targets.

Keratins, as integral members of the intermediate filament protein family, are embedded within the cellular framework, where they interact with a multitude of cellular proteins to activate intricate signaling networks that orchestrate cell migration, invasion, and metastasis^9,10. Conventional wisdom posits that keratins merely function as cytoskeletal scaffolds, maintaining mechanical integrity. However, Hobbs et al. revealed that select members, exemplified by Keratin 17 (K17), translocate into the nucleus via nuclear localization signals to directly modulate cardinal oncological processes. This seminal work postulates a novel “localization-dictated functionality” paradigm within tumor biology¹¹. Subsequent investigations elucidate that nuclear-localized keratins execute pivotal functions transcending their structural roles, crucially modulating gene expression and sculpting epigenetic landscapes^12,13. Keratin 19 (KRT19), a type I intermediate filament protein, has been shown in hepatocellular carcinoma to be closely associated not only with tumor aggressiveness, malignant biological behavior, and prognosis, but also with its own expression being co-regulated by epigenetic mechanisms and crosstalk with cancer-associated fibroblast signaling within the tumor microenvironment, suggesting that elevated KRT19 expression plays a crucial role in the progression of malignant tumors^13–16. Nevertheless, the mechanistic contributions of KRT19 in gastric carcinoma remain largely unexplored, and whether it possesses biological activities extending beyond its scaffold functions persists enigmatic, thus warranting urgent elucidation through rigorous scientific investigation.

In this study, we have elucidated the pivotal role of KRT19 in the malignant progression of gastric cancer and methodically delineated the consequences of its aberrant overexpression on gastric cancer cell motility through comprehensive in vivo and in vitro experimental paradigms. For the first time, we unveiled that nuclear-localized KRT19 interacts with heterogeneous nuclear ribonucleoprotein U (hnRNPU), orchestrating the ubiquitin-mediated degradation of IκBα to subsequently activate the NF-κB signaling cascade within the nuclear compartment. Downstream, RELA, as a pivotal transcription factor, binds to and activates the FSCN1 promoter, thereby enhancing its transcriptional expression. FSCN1, an actin-binding protein, facilitates the bundling of F-actin, promoting the formation of filopodia and markedly augmenting the migratory and invasive capacities of gastric cancer cells. This research not only first unveils the KRT19-mediated ubiquitination regulatory mechanism via nuclear localization, but also establishes the “KRT19-hnRNPU-NF-κB-FSCN1” regulatory axis. It thereby provides an innovative mechanistic paradigm for the non-structural functions of intermediate filament proteins in oncology, while simultaneously offering a theoretical foundation and potential therapeutic targets for anti-metastatic strategies aimed at KRT19-associated pathways.

Results

Non-negative matrix factorization of epithelial cells identifies KRT19 as a signature gene

Enhanced migratory and invasive capabilities of tumor cells represent a pivotal mechanism driving the progression and distant dissemination of gastric cancer¹⁷. However, the epithelial reprogramming processes underpinning this phenotypic shift remain incompletely understood. To systematically delineate the molecular features associated with the activation of motility programs in gastric epithelial cells, we employed a functional module identification strategy at single-cell resolution, aiming to uncover lineage-specific regulatory factors and their functional orientations. Through the harmonized integration of two publicly accessible 10× Genomics single-cell sequencing datasets (GSE183904 36 samples, 10 normal; and GSE201347: 22 samples, 10 normal), this study ultimately compiled 246,852 high-caliber single-cell profiles derived from 38 gastric cancer specimens and 20 adjacent non-malignant tissue samples (Fig. 1A). Utilizing dimensionality reduction and cluster analytical techniques, these cells were segregated into nine distinct cellular subtypes (Fig. 1B, C). To investigate the differential transcriptional states and co-expression patterns between normal gastric epithelium and tumor-derived epithelial cells, a second round of dimensionality reduction and clustering was performed, yielding a total of 36,689 epithelial cells. These cells exhibited pronounced intratumoral heterogeneity based on their tissue of origin (Supplementary Information: Fig. S1A, B). To rigorously characterize the composition of this epithelial population and distinguish malignant cells from their non-malignant counterparts, we inferred large-scale chromosomal copy number variations (CNVs) based on the transcriptomic data (Supplementary Information: Fig. S1C). This analysis classified the epithelial cells into chromosomally unstable (tumor-derived) and genome-stable (non-malignant) groups (Supplementary Information: Fig. S1D, E). Differential expression analysis confirmed that the non-malignant fraction was defined by high expression of gastric lineage markers (e.g., MUC5AC), whereas the tumor-derived fraction exhibited distinct malignancy-associated signatures. To further elucidate co-expression patterns across the entire epithelial landscape derived from multiple specimens and identify cardinal functional modules during malignant progression, we employed non-negative matrix factorization clustering on the integrated gene expression matrix of 36,689 epithelial cells. This analysis revealed eight distinct meta-programs (MP1–8) exhibiting unique functional characteristics and cellular states (Fig. 1D). Each MP was defined by a signature of the top 50 genes with the highest NMF scores (Supplementary data 1), further elucidating co-expression patterns underlying intra-tumor heterogeneity (Fig. 1E). These gene expression signatures delineate predominant cellular states within the epithelium, while quantitative assessment of metaprogram distribution across individual specimens revealed that MP1-expressing cells were consistently prevalent across the majority of samples, co-existing with other prominent states such as MP4 (Supplementary Information: Fig. S1F). To elucidate the biological functions of each MP and delineate their respective states, we conducted functional enrichment analyzes of epithelial cells across the distinct MPs (Fig. 1F). Based on the functional clustering of the top 50 signature genes and pathway enrichment analyzes of the signature genes, we annotated the biological identity of each MP.

Fig. 1. Single-cell transcriptomic landscape of gastric cancer reveals eight meta-programs for epithelial and identifies KRT19 as a signature gene.

A Workflow for identifying KRT19 as a candidate regulator in gastric cancer, integrating single-cell transcriptomics with clinical datasets and functional validation assays. B UMAP of single-cell transcriptomes showing major cell populations in gastric cancer, including epithelial, fibroblast, endothelial, myeloid, mast, plasma, T, B, and cycling cells; C Dot plot of representative marker genes across annotated cell types; D Hierarchical clustering analysis of pairwise similarities between NMF programs derived from epithelial cells across all gastric cancer samples. Similarities between programs were quantified using the Jaccard index based on their signature genes. The heatmap shows the consensus clustering results. The right panel delineates eight distinct epithelial meta-programs (MP1–MP8) emerging from the consensus NMF programs; E UMAP visualization of epithelial cells stratified by the eight MPs; F Heatmap showing the represented GSVA enrichment scores of KEGG, GO, and Hallmark pathways across the eight epithelial MPs, highlighting distinct functional signatures; G Dot plot displaying the expression levels of keratin family members across eight MPs; H Heatmap displaying the mRNA expression levels of keratin family members across digestive system malignancies. Data were derived from The Cancer Genome Atlas (TCGA) cohorts, specifically: Esophageal Carcinoma (TCGA-ESCA), Stomach Adenocarcinoma (TCGA-STAD), Colon Adenocarcinoma (TCGA-COAD), and Rectum Adenocarcinoma (TCGA-READ).

The MP1 program is enriched for canonical glandular epithelial markers, including KRT19, MUC5AC, and TFF1, consistent with features of the gastric epithelial lineage. In addition to these lineage-associated markers, MP1 concurrently exhibits elevated expression of genes involved in cytoskeletal organization and cell motility, such as ANXA2 and CAPN8, indicating enhanced dynamic cellular behaviors. We have designated MP1 as the ‘Differentiated Epithelial Program with Motility-Associated Features’. Notably, pathway analysis further revealed enrichment of processes related to substrate adhesion–dependent cell spreading and blood vessel remodeling (Fig. 1F), suggesting that MP1 cells are not functionally quiescent despite their epithelial differentiation features. Instead, this transcriptional profile reflects an epithelial program in which lineage identity is retained while motility- and microenvironment-associated pathways are engaged. The MP2 program exhibits pronounced enrichment of canonical neuroendocrine markers and regulators—including CHGA, INSM1, NEUROD1, PCSK1, and SNAP25—demonstrating definitive neuroendocrine-like differentiation characteristics. The MP3 program demonstrates marked enrichment of cardinal factors critically implicated in cell cycle progression—including BIRC5, MKI67, CDK1, CCNB1, and AURKB—which orchestrate DNA replication and mitotic events. This signature collectively portrays a hyperproliferative cellular state characterized by predominant G2/M-phase occupation. The MP4 program exhibits pronounced enrichment of canonical stress-response genes—including ATF3, DUSP1, HSPA1A, JUNB, and CDKN1A—manifesting the definitive expression signature of the Cellular Stress-Responsive Program. The MP5 program is enriched with key metabolic genes such as PDHA1, IDH2, ATP5PO, and GSTA1, reflecting a highly active metabolic state dominated by oxidative phosphorylation, and is thus defined as the Oxidative Metabolic Program. The MP6 program is enriched for glandular secretory markers such as MUC6, PGC, REG1A, and TFF2, exhibiting a prototypical profile of mature gastric glandular epithelial differentiation, and is thereby designated as the Glandular Secretory Epithelial Program. The MP7 program is enriched for pivotal immune regulatory genes, including CD69, CXCR4, CCL5, RUNX3, and GPR183, exhibiting hallmark features of immune responsiveness, and is thus designated as the Immune-Responsive Epithelial Program. The MP8 program is characterized by prominent expression of neuroendocrine markers such as CHGA, SCG3, SLC18A2, and TPH1, delineating a highly differentiated epithelial phenotype with pronounced secretory and neuroendocrine-like features. Given its broad expression across samples, its core epithelial lineage identity, and the inclusion of genes implicated in cytoskeletal remodeling and cellular motility (such as KRT19, ANXA2, and MYL12B), MP1 likely represents a fundamental epithelial program with latent transformative potential. Owing to its persistent universality, functional duality, and housing of pivotal cytoskeletal regulatory factors, MP1 was designated as the central meta-program for subsequent investigation. Within MP1, several keratin genes exhibit high expression, consistent with a glandular epithelial lineage. Notably, however, an unbiased comparison of the keratin family across metaprograms revealed a clear distinction between pan-epithelial keratins and MP1-selective keratins: KRT8 and KRT18 show broadly elevated expression across multiple MPs, reflecting a general epithelial backbone, whereas KRT19 displays comparatively stronger enrichment and a higher fraction of expressing cells within MP1 than in other MPs (Fig. 1G). Accordingly, KRT19 was emphasized not because KRT8/18 are unimportant, but because KRT19 better captures the program-level specificity of MP1 in our single-cell framework. Further analysis incorporating The Cancer Genome Atlas (TCGA) datasets (ESCA, STAD, COAD, and READ) revealed that KRT19 is among the most highly expressed keratin genes across multiple digestive system malignancies, underscoring its broad utility as a stable epithelial marker (Fig. 1H). Given its dual prominence within the MP1 program and tumor expression profiles, KRT19 was selected as the representative gene of MP1 for in-depth investigation.

Transcriptomic analysis revealed elevated expression of KRT19 in gastric cancer and underscored its potential clinical significance

Initial integrative analysis of pan-cancer data from TCGA combined with GTEx datasets revealed aberrant overexpression of KRT19 in numerous solid tumors (Fig. 2A). Particularly pronounced overexpression was documented across gastrointestinal malignancies—including esophageal cancer, gastric cancer, colon cancer, and rectal cancer—highlighting its consistent dysregulation within this cancer spectrum. Moreover, systematic analysis of multiple GEO datasets (GSE13159, GSE19826, GSE220917, GSE29272, GSE33335, GSE63089, and GSE66229) further confirmed the aberrant upregulation of KRT19 in gastric cancer (Fig. 2B). Concurrently, when contrasted with primary gastric tumor tissues, hepatic metastasis specimens exhibited elevated KRT19 expression levels (Fig. 2C). Furthermore, western blot analysis of 12 matched pairs of primary tumors and liver metastases showed that KRT19 was increased in most metastatic lesions (10/12 pairs), although two pairs (P7M7 and P9M9) showed lower levels in metastases, consistent with inter-patient heterogeneity (Fig. 2D). These findings significantly imply that KRT19 may play a crucial role in malignant advancement of gastric cancer. Presently, the expression profile of KRT19 in gastric cancer and its clinical relevance remain insufficiently characterized. To elucidate this, we performed immunohistochemical analyzes on 242 primary gastric cancer specimens alongside 242 paired adjacent non-neoplastic tissue samples to evaluate KRT19 expression patterns (Fig. 2E, and Supplementary data 2). The majority of tumor tissues exhibited more intense KRT19 immunostaining than their matched normal counterparts. Expression levels were quantified using the Immunoreactive Score (IRS) system. As illustrated in Fig. 2F, IRS evaluations for KRT19 revealed markedly higher values in malignant tissues relative to paired normal specimens. Paired statistical analysis revealed a markedly elevated expression of KRT19 in tumor tissues, reaching a high level of significance (p < 2 × 10⁻¹⁶). To evaluate the prognostic significance of KRT19 in gastric cancer, Kaplan–Meier survival analysis was conducted based on its expression levels. Patients were stratified into high and low expression groups according to the median value of KRT19 expression. As shown in Fig. 2G, patients with high KRT19 expression exhibited significantly reduced overall survival (OS) compared to those in the low-expression group (HR = 0.52, 95% CI: 0.38–0.72, p = 7.08 × 10⁻⁵). Similarly, as depicted in Fig. 2H, high KRT19 expression was also associated with markedly shorter recurrence-free survival (RFS) (HR = 0.65, 95% CI: 0.45–0.95, p = 0.0242). Moreover, univariate logistic regression analysis of clinical data revealed that advanced T stage (T3/4 vs. T1/2) and positive nodal status (N1–3 vs. N0) were significantly associated with an increased risk of high KRT19 expression, with odds ratios of 1.80 (P = 0.034) and 2.47 (P = 0.001), respectively (Fig. 2I). These findings suggest a close correlation between KRT19 expression and both tumor invasion depth and lymph node metastasis. In contrast, no significant associations were observed between KRT19 expression and patient age, sex, or tumor differentiation status (all P > 0.05), suggesting that KRT19 expression in gastric cancer is more closely influenced by tumor biological behavior than by demographic factors. The aforementioned findings indicate that aberrant overexpression of KRT19 serves as a significant predictive biomarker for adverse clinical outcomes in gastric cancer patients.

Fig. 2. KRT19 is upregulated in gastric cancer and associates with malignant features and metastasis.

A Pan-cancer comparison of KRT19 mRNA expression (log₂) between TCGA tumors and GTEx normal tissues. Boxplots show median (center line), interquartile range (box), and points denote individual samples. Two-sided Wilcoxon rank-sum test was used for tumor vs. normal comparison within each cancer type; ^*P < 0.05, **P < 0.01, ^***P < 0.001, ^****P < 0.0001; ns, not significant. B Validation in independent GEO gastric cohorts (GSE13159, GSE19826, GSE220917, GSE29272, GSE63089, GSE33335, GSE66229), demonstrating significantly elevated KRT19 expression in gastric cancers relative to normal tissues; Boxplots are defined as in A; Two-sided Wilcoxon rank-sum test; significance as in A. C Analysis of GSE23786 dataset revealing higher KRT19 levels in liver metastases compared to primary gastric tumors; Boxplots are defined as in A. Two-sided Wilcoxon rank-sum test; **P < 0.01. D Western blot analysis of KRT19 in 12 paired primary gastric cancer (P1- P12) and matched liver metastases (M1- M12) tissues, with GAPDH as loading control; KRT19 was higher in metastases in 10 of 12 pairs, whereas 2 pairs showed lower metastatic levels, reflecting biological heterogeneity; Densitometric quantification (KRT19/GAPDH) across the 12 pairs is provided in Fig. S5A; two-sided paired test; P = 0.016. E Representative immunohistochemistry (IHC) for KRT19 in normal mucosa and gastric tumor; scale bars: 200 μm; F Paired IHC immunoreactive scores (IRS) for KRT19 across 242 matched normal–tumor specimens, showing significant upregulation in tumors. Each line connects paired samples from the same patient. Two-sided paired Wilcoxon test; P < 2 × 10⁻¹⁶. G, H Kaplan–Meier curves for recurrence-free survival (RFS) and overall survival (OS) stratified by KRT19 expression (high vs. low); hazard ratios and log-rank p-values are indicated; I Forest plot of logistic regression analyzing associations between high KRT19 expression and clinicopathological variables; T stage and N stage show significant positive associations (odds ratios with 95% CIs and p-values shown); J Representative immunoblot of KRT19 in gastric epithelial cells (GES-1) and gastric cancer cell lines (AGS, MKN1, HGC27, MKN45, SNU719, FU97); GAPDH, loading control. Densitometric quantification (KRT19/GAPDH) from three independent experiments is provided in Fig. S5B (mean ± SD); one-way ANOVA with multiple-comparisons test; significance as indicated.

KRT19 promotes the migration, invasion, and metastasis of gastric cancer

KRT19 exhibits elevated expression within gastrointestinal malignancies; however, its precise role in gastric cancer remains to be elucidated. Employing Western blot analyzes on the normal gastric mucosal cell line GES1 and six gastric cancer cell lines (Fig. 2J), AGS and MKN1 cell lines were selected as suitable candidates for establishing stable knockdown models, while the HGC27 line was chosen to generate overexpression models; subsequent functional assays were performed accordingly. Results from the CCK-8 assay (Fig. 3A, B) indicated that knockdown of KRT19 did not significantly alter the proliferative capacity of AGS and MKN1 cells. Consistently, quantitative analysis of the colony formation assay revealed no significant difference in the number of colonies between the shCon and shKRT19 groups (Fig. 3C). Furthermore, the EdU incorporation assay demonstrated that the number of EdU-positive proliferating cells remained unchanged upon KRT19 depletion (Fig. 3D). These findings suggest that KRT19 may not be involved in regulating cell cycle progression or apoptosis during the malignant progression of gastric cancer. To investigate the biological functions of KRT19 in cancer motility, a comprehensive suite of functional assays was conducted to elucidate its implications for gastric cancer cell migratory and invasive capacities. Wound healing and transwell migration assays demonstrated that KRT19 knockdown markedly attenuated the migratory dynamics of AGS and MKN1 cellular lineages, whereas its ectopic overexpression in HGC27 cells significantly enhanced the motility of gastric cancer cells (Fig. 3E, F). Transwell invasion assays further corroborated that silencing KRT19 significantly diminished the invasive capacity of AGS and MKN1 cells, whereas its overexpression markedly enhanced the invasiveness of gastric cancer cells (Fig. 3E). To further validate the invasive potential of gastric cancer cells, 3D spheroid invasion assays were performed using AGS and MKN1 cell lines. Bright-field and GFP fluorescence imaging revealed that spheroids formed by shKRT19 cells were markedly smaller and more compact compared to those in the shCon group. In contrast, shCon cells exhibited irregular, loosely organized structures with prominent invasion into the surrounding matrix. Conversely, ectopic KRT19 expression in HGC27 cells produced the opposite phenotype. KRT19-overexpressing spheroids displayed increased invasive outgrowth characterized by a larger invasion area and more pronounced radial protrusions extending into the matrix, whereas vector-control spheroids remained relatively compact with limited peripheral dissemination (Fig. 3G). These findings indicate that KRT19 knockdown significantly impairs the invasive and spreading capabilities of gastric cancer cells within a three-dimensional extracellular matrix (ECM) environment. Collectively, these findings thus demonstrate that aberrant KRT19 overexpression potentiates gastric cancer cell migration and invasion in vitro. Moreover, proliferation assay results robustly ruled out the possibility that KRT19’s pro-invasive and pro-migratory effects stem from promoting cellular expansion. To further validate the impact of KRT19 on the migratory and invasive behavior of gastric cancer cells, a hepatic metastasis model was established via intrasplenic injection in nude mice. MKN1 cells transduced with either shCon or shKRT19 constructs were injected, and metastatic progression in the liver was monitored through in vivo bioluminescence imaging on days 5, 10, 15, 20, 25, and 30 post injection. The results revealed that the shCon group exhibited early and progressively intensifying metastatic signals, indicating a heightened metastatic capacity; in contrast, the shKRT19 group displayed markedly weaker signals with no evident trend of progression (Fig. 3H). Autopsy of hepatic specimens at the experimental endpoint revealed that the shCon cohort exhibited larger and more discernible metastatic lesions, whereas the shKRT19 cohort presented a marked reduction in both the number and volume of metastatic nodules (Fig. 3H), demonstrating that KRT19 knockdown significantly impairs the hepatic metastatic potential of gastric cancer cells. Notably, one mouse in the shCon group died prior to the Day 30 endpoint due to tumor-associated burden, resulting in missing Day-30 imaging and histology for that individual. Collectively, these findings provide compelling evidence for the pivotal role of KRT19 in the metastatic progression of gastric cancer.

Fig. 3. KRT19 promotes motility and metastatic potential of gastric cancer cells without altering proliferative capacity.

A, B Cell growth curves (CCK-8) of AGS and MKN1 cells with stable KRT19 knockdown (shKRT19) versus control (shCon); Data are mean ± SD from 3 independent experiments; group differences across time were analyzed by two-way ANOVA (time × group), with significance indicated as shown (ns). C Representative colony-formation images of AGS and MKN1 cells with shCon or shKRT19. Colony numbers were quantified (Fig. S6A) and showed no significant difference (ns; two-sided unpaired t-test within each cell line). D Representative EdU incorporation assay (Hoechst, blue; EdU, red) in AGS and MKN1 cells with shCon or shKRT19. EdU-positive cells were quantified (Fig. S6B) and showed no significant difference (ns; two-sided unpaired t-test); scale bars: 100 μm. E Representative transwell assays: KRT19 knockdown reduces migration and invasion in AGS and MKN1 cells, whereas KRT19 overexpression in HGC27 enhances both; Quantification of migrated/invaded cell numbers is shown in Fig. S6C (mean ± SD, n = 3 independent experiments). Statistics: two-sided unpaired t-test (significance as indicated); scale bars: 100 μm. F Representative Wound-healing assays (0 h and 24 h): diminished gap closure upon KRT19 depletion in AGS/MKN1 and accelerated closure upon KRT19 overexpression in HGC27; Quantification of migrated/invaded cell numbers is shown in Fig. S6D (mean ± SD, n = 3 independent experiments). Statistics: two-sided unpaired t-test (significance as indicated). scale bars: 100 μm; G 3D spheroid invasion in matrix: shKRT19 spheroids display reduced invasive outgrowth (bright-field and GFP); Red circles to the “spheroid core” Invasion index [(total area-core area)/core area]is shown in Fig. S6E (mean ± SD, n = 3 independent experiments). Statistics: two-sided unpaired t-test (significance as indicated); scale bars: 100 μm. H In vivo liver-metastasis model with intrasplenic injection and longitudinal bioluminescent imaging at days 5, 10, 15, 20, 25, and 30. The pseudocolor scale indicates radiance intensity (IVIS output), with “High” indicating stronger signal and “Low” indicating weaker signal. Compared with the control group, shKRT19 cells exhibited consistently lower hepatic metastatic signals, with markedly reduced metastatic burden evident at day 30. Representative gross liver specimens and H&E staining further corroborate diminished metastases in the shKRT19 group. In the shCon group, one mouse died prior to the Day 30 endpoint due to severe metastatic burden. Bioluminescent radiance (log10) was analyzed using a linear mixed-effects model for overall group and time×group effects, and per-time-point differences were assessed as indicated in Fig. S6F.

KRT19 serves as a positive regulatory factor of the NF-κB signaling pathway and FSCN1 expression in gastric cancer

To further investigate the underlying mechanisms by which KRT19 depletion attenuates the invasive and migratory capacities of gastric cancer cells, RNA-seq analysis was performed on shKRT19-AGS and shCon-AGS cell lines. Among the most significantly downregulated genes in the shKRT19-AGS cells was FSCN1 (Fig. 4A, and Supplementary data 3). FSCN1 encodes Fascin, one of the most extensively investigated actin-bundling proteins, documented to be overexpressed in diverse tumor tissues¹⁸. This protein facilitates the cross-linking of actin microfilaments into rigid, compact parallel bundles, exerting a critical function in maintaining cellular architecture. Fascin protein’s actin-bundling function activates cytoskeletal remodeling; this regulatory mechanism proves indispensable during the formation and extension of filopodia and lamellipodia—primary actin-based membrane protrusions¹⁹. Filopodia constitute integral components in cellular migration, ECM adhesion, and intercellular communication. Studies have shown that these slender protrusions are essential for the anchoring of metastatic tumor cells within secondary tissues or organs^20,21. Immunohistochemical analysis of five tissue microarrays comprising 242 pairs clinical samples revealed that FSCN1 is highly expressed in gastric cancer tissues compared to normal counterparts (Fig. 4B), with a significant positive correlation observed between KRT19 and FSCN1 expression levels (Supplementary Information: Fig. S2A–C). Western blot analysis confirmed a marked downregulation of FSCN1 in gastric cancer cells with stable KRT19 knockdown (Fig. 4C). To elucidate the functional enrichment profiles of differentially expressed genes (DEGs) following KRT19 knockdown, we conducted Gene Ontology analysis and constructed the annular clustering diagram depicted in Fig. 4D. The inner ring depicts logFC values of genes, exhibiting a chromatic gradation from cerulean (diminished expression) to amber (elevated expression); the peripheral ring delineates enriched GO terms associated with each gene, with distinct hues representing diverse functional categories—including “actin filament organization,” “epithelial cell migration,” and “regulation of actin cytoskeleton organization,” among others. The findings reveal significant enrichment of DEGs in pivotal processes, including cytoskeletal reorganization, cellular motility, and stress fiber formation, thereby substantially substantiating that KRT19 likely orchestrates gastric cancer cell cytoskeletal remodeling—potentially through regulation of key mediators like FSCN1—consequently modulating cellular migratory propensities. Employing immunofluorescence assays across both AGS and MKN1 cell lineages, we observed that following KRT19 knockdown, cells underwent substantial morphological alterations—transitioning from their inherent rhomboid-like configurations to circular forms devoid of protrusions, conversely, ectopic KRT19 expression in HGC27 cells promoted a more elongated, protrusion-rich morphology compared with vector controls, consistent with an enhanced migratory phenotype (Fig. 4E, and Supplementary Information: Fig. S2D, F). Immunofluorescence assays revealed that within shCon group cells, F-actin filaments bundled by FSCN1 localize peripherally to sustain cellular architecture and filopodia, whereas upon KRT19 knockdown, diminished FSCN1 expression precipitated pronounced F-actin disarray, exhibiting aberrant accumulations at intercellular junctions; concomitantly, the cells underwent significant cytoskeletal remodeling reflecting these morphological changes, KRT19 overexpression in HGC27 cells increased FSCN1 signal and restored organized peripheral F-actin bundling, accompanied by more prominent filopodial protrusions compared with vector controls (Fig. 4F, and Supplementary Information: Fig. S2E, G). Live-cell imaging analysis demonstrated that KRT19 knockdown substantially suppressed filopodia density and length in both AGS and MKN1 cells compared to shCon controls (Fig. 4G); conversely, ectopic KRT19 expression in HGC27 cells elicited the opposite phenotype, markedly enhancing filopodia formation (Supplementary Information: Fig. S2H). Through a prolonged 30-min imaging observations, shCon cells exhibited enhanced motility with their filopodia displaying greater extension capability and robust retractive capacity (Supplementary video 1, the statistical analysis of filopodia dynamics were in Fig. S11E), conversely, KRT19 overexpression in HGC27 cells further increased filopodial dynamic turnover, with significantly accelerated extension and retraction rates compared with vector controls (Supplementary video 2, the statistical analysis of filopodia dynamics were in Fig. S11E). Collectively, these findings propose that KRT19 potentially modulates FSCN1 expression to orchestrate cytoskeletal remodeling, thereby facilitating enhanced migratory capacity in gastric carcinoma cells.

Fig. 4. KRT19 supports NF-κB pathway activation and FSCN1 expression in gastric cancer cells.

A RNA-seq volcano plot (shKRT19-AGS vs shCon-AGS) highlighting FSCN1 among the most significantly down-regulated genes; B Representative IHC images of FSCN1 in paired human gastric cancer and adjacent non-malignant tissues on TMAs (n = 242 pairs). Whole-slide scans were evaluated and scored across the entire tissue cores by two independent blinded pathologists; representative high-magnification fields were selected after scoring to reflect the consensus staining intensity and typical localization pattern while avoiding technical artifacts. FSCN1 shows predominantly cytoplasmic staining in malignant epithelial cells, whereas adjacent glands exhibit weak/negative staining; scale bars: 200  μm; C Immunoblotting in AGS/MKN1 cells with stable KRT19 knockdown and in HGC27 cells with KRT19 overexpression. KRT19 depletion reduces phospho-p65 (Ser536) and FSCN1 without altering total p65 levels, indicating reduced pathway activation, whereas KRT19 overexpression increases the phosphorylation of p65 and FSCN1 expression; GAPDH, loading control; Densitometric quantification (normalized to GAPDH) from n = 3 independent experiments are shown in Fig. S7A (mean ± SD); statistics were performed using two-sided unpaired t-test (shCon vs shKRT19; vector vs KRT19), with significance as indicated. D Functional enrichment of differentially expressed genes upon KRT19 knockdown visualized as an annular/circular map: the inner ring encodes gene log2 fold change and the outer ring lists enriched GO terms (e.g., actin-filament organization, epithelial cell migration, regulation of actin cytoskeleton); E Confocal microscopy showing KRT19 immunofluorescence (red) and DAPI nuclear staining (blue) in shCon-AGS and shKRT19-AGS cells, with attenuated KRT19 signal and morphological changes upon knockdown; scale bars: 20 μm; Images are representative of n = 3 independent experiments. F Confocal microscopy showing FSCN1 (yellow) and F-actin (phalloidin; red). In control cells, FSCN1 bundles F-actin into organized peripheral filaments supporting filopodia, whereas KRT19 knockdown leads to reduced FSCN1 expression, pronounced F-actin disarray, and abnormal accumulations at intercellular junctions; scale bars: 20 μm; Images are representative of n = 3 independent experiments. G Live-cell imaging revealing a shift from spindle-like morphology with prominent filopodia (shCon) to rounded cells with diminished filopodial extension/retraction (shKRT19) in AGS and MKN1; Filopodia length and density were quantified (mean ± SD) and show significant reductions upon KRT19 knockdown (Fig. S7B); statistics: two-sided unpaired t-test, ^****P < 0.0001. Scale bars, 20 μm; Images are representative of n = 3 independent experiments. H Gene-set enrichment analyses (GSEA) of EMT and TNFα/NF-κB hallmark pathways. Patients were stratified into KRT19-high (top 10%) and KRT19-low (bottom 10%) cohorts based on expression levels. Note that the enrichment peaks appear at the right side of the rank list, indicating that these pathways are significantly activated in the KRT19-high population (labeled on the right). bottom heatmap summarizes normalized enrichment scores; I Western blots of EMT markers showing that KRT19 knockdown decreases N-cadherin, Vimentin, α-SMA, MMP2/9 and increases E-cadherin, with reciprocal changes upon KRT19 overexpression in HGC27. Densitometric quantification (normalized to GAPDH) from n = 3 independent experiments are shown in Fig. S7C (mean ± SD); statistics were performed using two-sided unpaired t-test, with significance as indicated.

Our experimental findings demonstrate that silencing KRT19 in gastric cancer cell lines leads to a significant downregulation of FSCN1 expression. Conversely, overexpression of KRT19 in HGC27 cells, which naturally exhibit low KRT19 levels, results in a corresponding upregulation of FSCN1. To further elucidate the molecular mechanisms through which KRT19 regulates cellular motility, Gene Set Enrichment Analysis (GSEA) demonstrated profound suppression of epithelial-mesenchymal transition (EMT) and NF-κB signaling pathways across low- KRT19 cohorts in TCGA, GEO, and our RNA-seq datasets relative to high-expression counterparts (Fig. 4H). Notably, in our RNA-seq-based GSEA, these pathways ranked among the top two most enriched signaling cascades. To substantiate these findings, Western blot analyzes revealed that KRT19 knockdown profoundly suppresses EMT while concomitantly diminishing expression levels of key motility markers—including Vimentin (VIM) and α-Smooth Muscle Actin (α-SMA). Conversely, KRT19 overexpression augments the progression of EMT (Fig. 4I). These findings further underscore the pivotal role of KRT19 in modulating cellular motility and the invasive and migratory capacities of gastric cancer cells. Building upon GSEA analysis revealed a pronounced suppression of the NF-κB signaling pathway in samples with low KRT19 expression. To determine whether KRT19 modulates NF-κB signal transduction—given that phosphorylation of p65 at serine 536 is a well-established marker of NF-κB activation—we assessed the phosphorylation status of p65, a key transcription factor within the pathway. Our experimental data demonstrate that KRT19 depletion in gastric cancer cell lines markedly diminishes p65 phosphorylation (Ser536) without affecting total p65 protein levels, indicating a suppression of NF-κB pathway activation. Conversely, KRT19 augmentation conversely elevates phospho-p65 levels (Fig. 4C). In the TCGA cohort, Pearson correlation analysis further confirmed a strong positive association between KRT19 expression and NF-κB family targets, as well as their downstream effectors (Supplementary Information: Fig. S2I). These findings underscore KRT19 as a positive regulator of NF-κB signaling and an upstream activator of FSCN1 expression, playing a pivotal role in orchestrating cytoskeletal remodeling, preserving cellular morphology, and enhancing cell motility.

As an upstream transcription factor of FSCN1, NF-κB p65 (RELA) is capable of directly binding to the promoter region of the FSCN1 gene

Our RNA sequencing data establish that FSCN1 is transcriptionally responsive to KRT19 knockdown, while Western blot analysis further demonstrated significant downregulation of FSCN1 protein expression in human gastric cancer cell lines upon KRT19 suppression. These findings suggest that the downregulation of FSCN1 is mediated through transcriptional mechanisms. Subsequent investigative efforts focus on identifying upstream transcription factors modulated by KRT19 perturbation that functionally govern the induction of FSCN1 mRNA expression. We first analyzed tumor samples from the TCGA-STAD dataset, identifying DEGs based on FSCN1 mRNA expression levels (high vs. low expression groups). Utilizing these DEGs, we conducted GSEA analysis (http://software.broadinstitute.org/gsea/msigdb/collections.jsp#C3) to explore potential transcription factors, and further employed the online tool PROMO for transcription factor prediction. GSEA identified 74 putative transcription factors critically involved in FSCN1 regulation, while PROMO predicted 48 candidates (Supplementary data 4); intersecting these datasets yielded seven key transcription factors, including RELA (p65) and NF-κB (Fig. 5A, B). To further validate the bioinformatic findings, we investigated whether NF-κB transcriptionally modulates FSCN1 expression within gastric cancer cell. Indeed, activation of NF-κB by tumor necrosis factor-α (TNF-α) markedly upregulated FSCN1 protein expression in HGC27 cells, accompanied by increased phosphorylation of p65 at Ser536 (a widely used readout of NF-κB activation), whereas total p65 levels remained largely unchanged (Fig. 5C), moreover, KRT19 overexpression further potentiated the TNF-α–induced FSCN1 upregulation in HGC27 cells (Supplementary Information: Fig. S3A). Conversely, treatment with the NF-κB inhibitor BAY significantly suppressed FSCN1 expression in AGS and MKN1 cells, together with reduced p65 phosphorylation but without obvious changes in total p65 abundance (Fig. 5D). To elucidate whether P65 directly binds to the FSCN1 promoter, we conducted CUT&Tag assays in gastric cancer cells. The heatmap and average signal profile (Fig. 5E) revealed pronounced enrichment of P65 at transcription start site (TSS) regions genome-wide, in stark contrast to the negligible signal observed in IgG controls. Notably, a conspicuous peak of P65 enrichment was localized proximate to the FSCN1 gene TSS (Fig. 5F), demonstrating direct interaction with the promoter region. These findings corroborate the postulate that FSCN1 functions as a direct transcriptional target of P65 in gastric cancer cells. To further validate the regulatory effect of P65 on FSCN1 promoter activity, we constructed a luciferase reporter plasmid containing the FSCN1 promoter and 5′UTR region, and co-transfected it with a RELA overexpression vector into 293T cells. Dual-luciferase assay results revealed that RELA overexpression significantly enhanced FSCN1 promoter-driven luciferase activity compared to the control group, indicating that P65 functions as a positive transcriptional activator of FSCN1 (Fig. 5H). To further ascertain whether this regulatory effect is dependent on specific binding sites, we utilized the JASPAR database (http://jaspar.genereg.net/) to identify two putative NF-κB binding motifs within the FSCN1 promoter region, which were subsequently subjected to site-directed mutagenesis (Fig. 5G). The results demonstrated that RELA-induced luciferase activity was significantly diminished in the context of the mutated promoters compared to the wild-type construct, indicating that these predicted sites serve as bona fide binding elements essential for transcriptional activation (Fig. 5H). These novel findings confirm that NF-κB directly binds to the promoter region of FSCN1. Taken together, the current body of evidence indicates that inhibition of NF-κB signaling is essential for the FSCN1 downregulation observed upon KRT19 depletion. KRT19 promotes gastric cancer metastasis, at least in part, through the NF-κB–FSCN1 regulatory axis.

Fig. 5. NF-κB p65 (RELA) directly regulates FSCN1 transcription.

A, B Venn diagram integrating GSEA-predicted transcription factors with PROMO database, identifying RELA as a candidate regulator of FSCN1; C Western blot in HGC27 cells treated with TNFα (100 ng/mL) for 0, 3, and 6 h, showing increased FSCN1 and NF-κB activation (p-p65 Ser536) and total p65 protein levels unchanged. GAPDH, loading control. Densitometric quantification normalized to GAPDH is shown in Fig. S8A (mean ± SD, n = 3 independent experiments). Statistics were performed using One-way ANOVA followed by Dunnett’s multiple-comparisons test versus 0 h. significance is indicated (^*P < 0.05, ^**P < 0.01, ^****P < 0.0001; ns, not significant). D Western blot in AGS and MKN1 cells treated with BAY 11-7082 (5 ng/mL) for 0, 3, and 6 h, showing suppression of FSCN1 and reduced NF-κB activation (p-p65 Ser536) and total p65 protein levels unchanged. GAPDH, loading control. Densitometric quantification normalized to GAPDH is shown in Fig. S8B (mean ± SD, n = 3 independent experiments). Statistics: One-way ANOVA followed by Dunnett’s multiple-comparisons test versus 0 h; significance is indicated as shown. E CUT&Tag profiling confirming p65 binding enrichment around transcription start sites (TSS). F Genome browser tracks showing p65 occupancy at the FSCN1 promoter region compared to IgG control. G Schematic representation of the FSCN1 promoter region showing two predicted NF-κB binding motifs identified by JASPAR, with site-directed mutagenesis applied to disrupt these elements. H Dual-luciferase reporter assays in 293T cells demonstrating that RELA overexpression significantly enhances FSCN1 promoter-driven activity in the wild-type construct, whereas this effect is markedly reduced in the mutant promoters, confirming the functional importance of the predicted NF-κB binding sites.

KRT19 stimulates NF-κB signaling through its interaction with hnRNPU

To elucidate the molecular mechanisms by which KRT19 modulates NF-κB signaling, we evaluated alterations in NF-κB pathway-associated genes following KRT19 knockdown in gastric cancer cell AGS and MKN1 through qRT‌-PCR and western blot. The experimental outcomes demonstrate that within AGS and MKN1 cell lines, following KRT19 knockdown, IκBα expression becomes significantly augmented at the protein level, while no discernible differences manifest transcriptionally (Fig. 6A, B, and Supplementary Information: Fig. S3B). Critically, diminished KRT19 expression exerted no discernible effect upon IKK phosphorylation status in either AGS or MKN1 cells. These observations collectively indicate that KRT19 likely governs NF-κB signaling principally through orchestrating IκBα post-translational modifications rather than influencing IKK activation. Given that KRT19 knockdown led to a reduction in IκBα protein abundance without altering its mRNA levels or upstream IKK phosphorylation, we speculated that KRT19 might modulate IκBα protein stability through a post-translational mechanism independent of canonical kinase signaling. As many post-translational modifications rely on protein–protein interactions to recruit modifying enzymes or scaffold complexes, we subsequently performed co-immunoprecipitation followed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) to identify potential KRT19-interacting partners (Fig. 6C), while highly sensitive silver staining revealed expected baseline background in the IgG control, additional bands were enriched in the KRT19 immunoprecipitate. Among the identified proteins, hnRNPU was consistently identified within KRT19-immunoprecipitated specimens. As evidenced by the extracted ion chromatogram (XIC), numerous unique peptides originating from hnRNPU—including the signature peptide SSGPTSLFAVTVAPPGAR—manifested robust signal intensity and distinct elution profiles (Fig. 6D). The corresponding MS/MS fragmentation spectrum for this peptide revealed an uninterrupted y-ion cascade (y₁–y₁₄), definitively confirming its identification (Fig. 6D). These findings substantiate the physical interaction between hnRNPU and KRT19 under the examined cellular conditions. It is noteworthy that hnRNPU functions as a scaffold protein facilitating NF-κB pathway activation through modulating IκBα stability; By establishing domain-dependent interactions, it bridges IκBα to its cognate E3 ubiquitin ligase β-TrCP; this molecular linkage is essential for β-TrCP-mediated ubiquitination and subsequent proteasomal degradation of IκBα²². Given that hnRNPU-mediated NF-κB signaling constitutes a nuclear event, coupled with the established interaction between KRT19 and hnRNPU alongside KRT19’s recognized role in NF-κB pathway activation, we postulate that this mechanism likely hinges upon a pivotal nexus involving hnRNPU. Subsequent experimentation was therefore designed to elucidate the subcellular distribution of KRT19, thereby validating the hypothesis that KRT19 executes its signaling function via this interaction within the nuclear environment. Recent evidence reveals that the intermediate filament protein KRT19 exhibits biological functionality within the nucleus¹³. Nuclear and Cytoplasmic Protein Extraction Assay performed in gastric cancer cells, employing concurrent deployment of Western blot analysis (evaluating total, cytoplasmic, and nuclear fractions) and immunofluorescence (Fig. 6E, F), demonstrated conclusively the nuclear localization of KRT19. Subsequently, we executed nuclear and cytoplasmic protein extraction Assay followed by co-immunoprecipitation assays, which definitively established the nuclear compartment interaction between the intermediate filament protein KRT19 and hnRNPU as well as with its known cooperators, including β-TrCP, IKKβ (Fig. 6G). Consequently, we formally hypothesize that nuclear KRT19 may participate in hnRNPU-mediated activation of NF-kB signaling. To substantiate this conjecture, we initially conducted immunofluorescence staining to analyze the nuclear distribution of pivotal NF-kB family members—P65, P50, and RELB—alongside their inhibitory protein IkBα in AGS cell. The findings revealed detectable nuclear localization of subsets of P65, P50, RELB, and IkBα within these cellular systems (Fig. 6H). Subsequently, nuclear proteins were extracted from HGC27 cells treated with the proteasome inhibitor MG132 to assess the ubiquitination status of IκBα. Because MG132 prevents proteasomal degradation, enhanced ubiquitination is detected as an accumulation of poly-ubiquitinated proteins. The results demonstrated that overexpression of KRT19 markedly increased the detectable intensity of the ubiquitin smear (Lane 3) compared to the vector control (Lane 1), indicating enhanced ubiquitin-mediated targeting of IκBα (Fig. 6I), an effect that was reversed upon hnRNPU knockdown (Lane 5), indicating that KRT19 augments hnRNPU-driven ubiquitination of IκBα. Furthermore, treatment with the nuclear export inhibitor Leptomycin B modestly increased IκBα ubiquitination (Lane 4) compared to the untreated group (Fig. 6I). Nuclear proteins were extracted from AGS and MKN1 cells to evaluate the ubiquitination level of IκBα. The results revealed that knockdown of KRT19 suppressed the ubiquitin-mediated degradation of IκBα, while overexpression of hnRNPU partially restored this effect (Fig. 6J, and Supplementary Information: Fig. S3C). Moreover, treatment with the nuclear export inhibitor Leptomycin B modestly enhanced IκBα ubiquitination (Fig. 6J, and Supplementary Information: Fig. S3C). In HEK-293T cells transfected with Flag-KRT19, Myc-IκBα, and HA-Ub plasmids, with proteasomal degradation inhibited by MG132 pretreatment, cells were harvested for co-immunoprecipitation analysis. The experimental outcomes demonstrated that introducing the Flag-KRT19 plasmid markedly increased Myc-antibody-associated ubiquitin levels (Supplementary Information: Fig. S3D). Moreover, employing co-immunoprecipitation methodology with hnRNPU as an internal reference, we discerned that KRT19 knockdown substantially diminished the interaction between hnRNPU and β-TrCP (Supplementary Information: Fig. S3E). Collectively, these findings illustrate that nuclear KRT19 bolsters the interaction between hnRNPU and β-TrCP, thereby promoting both IkBα ubiquitination-mediated degradation and functional activation of the NF-κB signal transduction pathway.

Fig. 6. Nuclear KRT19 engages hnRNPU to promote β-TrCP–dependent ubiquitination of IκBα and sustain NF-κB signaling.

A qRT-PCR in AGS cells showing that KRT19 knockdown does not markedly alter mRNA levels of representative NF-κB pathway components; Data are mean ± SD from n = 3 independent experiments; statistics: two-sided unpaired t-test for shCon vs shKRT19 for each gene (ns as indicated). B Protein immunoblotting of AGS and MKN1 cells following KRT19 knockdown to examine the expression and phosphorylation status of key NF-κB pathway components; GAPDH, loading control. Densitometric quantification (protein/GAPDH) is shown in Fig. S9A (mean ± SD, n = 3 independent experiments). Statistics: two-sided unpaired t-test (shCon vs shKRT19) for each target; significance as indicated. C Silver-stained gel of KRT19 immunoprecipitates (Input/IgG/IP) used for mass-spectrometry identification; D LC–MS/MS identification of hnRNPU from KRT19 IPs; representative XIC and MS/MS spectrum (e.g., peptide SSGPTSLFAVTVAPPGAR) confirm the KRT19–hnRNPU association; E Nuclear–cytoplasmic fractionation protein immunoblotting (Western blot) showing that KRT19 partially localizes in the nucleus of AGS and MKN1 cells (T: Total cell lysate; C: Cytoplasmic fraction; N: Nuclear fraction; Lamin A/C, nuclear marker; GAPDH, cytosolic marker). Nuclear KRT19 levels are reduced following shKRT19 knockdown. Representative of n = 3 independent fractionations. F Confocal microscopy showing increased nuclear accumulation of KRT19 after leptomycin-B (nuclear export inhibitor) compared with DMSO control; scale bars: 20 μm; Images are representative of n = 3 independent experiments. G Immunoblot analysis of endogenous hnRNPU, β-TrCP, and IKKβ proteins co-immunoprecipitated with KRT19 from nuclear extracts of AGS cells; Representative immunoblots from n = 3 independent experiments. H Confocal microscopy confirming nuclear localization of NF-κB family members (p50, p65) and IκBα in AGS cells; scale bars: 20 μm; Representative of n = 3 independent experiments. I, J Nuclear ubiquitination assays of IκBα in HGC27 and AGS cells under different experimental conditions (IP: IκBα; IB: HA); Quantification of normalized IκBα ubiquitination levels is provided in Fig. S9B, C (mean ± SD, n = 3 independent experiments) with the six treatment conditions. Statistics: one-way ANOVA with multiple-comparisons testing across the six conditions; significance is indicated.

FSCN1 functionally compensates for KRT19 deficiency to rescue cell migration

Building upon the preceding findings—namely, that nuclear KRT19 promotes IκBα ubiquitination and degradation via interaction with hnRNPU, thereby activating NF-κB signaling and transcriptionally upregulating FSCN1—we postulate that FSCN1 likely serves as a principal effector downstream of this pathway. We further speculate that restoration of FSCN1 expression may compensate for KRT19 deficiency and potentially reverse the impaired migratory capacity of gastric cancer cells. To elucidate the functional positioning of FSCN1 within the KRT19 regulatory axis and its pivotal role in modulating cellular motility, we established an FSCN1 overexpression model in gastric cancer cell lines with stable KRT19 knockdown. This functional “rescue” experiment was designed to determine the indispensability of FSCN1 in KRT19-mediated regulation of cell migration, thereby reinforcing the mechanistic framework through a comprehensive compensatory approach. Wound healing assays revealed that, compared to the shKRT19+vector group, overexpression of FSCN1 markedly accelerated the migratory closure of AGS and MKN1 cells (Fig. 7A). This observation was further corroborated by Transwell assays, wherein FSCN1 overexpression significantly enhanced the migratory and invasive capabilities of shKRT19 cells, indicating a pronounced compensatory effect of FSCN1 in counteracting the motility suppression induced by KRT19 deficiency (Fig. 7B). In the in vivo metastasis model, bioluminescence imaging was performed at six consecutive time points (days 5, 10, 15, 20, 25, and 30) to dynamically monitor hepatic colonization following splenic injection of gastric cancer cells in nude mice (Fig. 7C). In the shCon+vector group, hepatic fluorescence signals markedly increased from day 10 onward and continued to accumulate over time, indicating robust metastatic capacity (Fig. 7D). In contrast, the shKRT19+vector group exhibited only faint signals throughout the observation period, suggesting that KRT19 knockdown significantly impairs the cells’ in vivo metastatic potential (Fig. 7E). Notably, in the shKRT19 + FSCN1 group, hepatic fluorescence began to intensify from day 15 and progressively increased thereafter, indicating that FSCN1 overexpression partially restores the metastatic capability lost due to KRT19 depletion, thereby reinforcing its pivotal downstream role in this regulatory axis (Fig. 7F). It is worth noting that one mouse in the shKRT19 + FSCN1 group died prior to the day 30 endpoint due to tumor-associated burden, preventing acquisition of its final gross organ and histological images. To evaluate whether the proposed KRT19–hnRNPU–NF-κB–FSCN1–F-actin cascade is reflected in our transcriptomic data, we systematically mapped single-cell expression of axis components across all NMF-defined meta-programs, including KRT19, HNRNPU, BTRC (β-TrCP), NFKBIA (IκBα), NFKB1, RELA, FSCN1, and representative actin/cytoskeletal genes, together with an NF-κB target-gene activity score. MP1 showed the clearest convergence of high KRT19 and FSCN1 with an actin/cytoskeletal remodeling signature (Supplementary Information: Fig. S4A, B), whereas several intermediate regulators displayed limited transcript-level differences across programs, consistent with predominant post-transcriptional/protein-level control (Supplementary Information: Fig. S4B). Notably, although the highest global NF-κB score occurred in the stress-responsive MP4, MP1 exhibited intermediate-to-high NF-κB output (Supplementary Information: Fig. S4C). And importantly, the coupling of NF-κB output with FSCN1 expression, supporting a motility-oriented NF-κB–FSCN1 state in MP1. Collectively, these findings establish that KRT19 potentiates cellular motility and metastasis through orchestrating FSCN1 expression. Functioning as the primary executor within this signaling cascade, FSCN1 reconstitution effectively countervails motility deficits resulting from KRT19 depletion. This phenomenon ultimately validates the functional continuum of the “KRT19–hnRNPU–NF-κB–FSCN1–F-actin” regulatory axis in driving gastric cancer metastasis (Fig. 7G).

Fig. 7. FSCN1 functionally compensates for KRT19 loss in vitro and in vivo.

A Wound-healing assays in AGS and MKN1 cells expressing shCon+vector, shKRT19+vector, or shKRT19 + FSCN1; representative phase-contrast images at 0 and 24 h; scale bars: 100 μm; Relative migration rate was quantified (mean ± SD, n = 3 independent experiments; dots indicate independent experiments) and is provided in Fig. S10A. Statistics: one-way ANOVA followed by multiple-comparisons testing among the three groups within each cell line; significance is indicated (ns, ^**P < 0.01, ^***P < 0.001). B Transwell migration and Matrigel invasion assays under the same conditions; representative crystal-violet–stained fields; scale bars: 100 μm; Migrated/invaded cell numbers were quantified (mean ± SD, n = 3 independent experiments; dots indicate independent experiments) and are provided in Fig. S10B. Statistics: one-way ANOVA with multiple-comparisons testing among the three groups within each cell line; significance is indicated (ns, ^***P < 0.001, ^****P < 0.0001). C Schematic of the in vivo workflow: intrasplenic injection of luciferase-labeled cells followed by longitudinal bioluminescent imaging on days 5, 10, 15, 20, 25, and 30; livers were collected at day 30 for gross examination and H&E staining; Mice were randomly divided into groups (Data represent n = 12 biologically independent animals per group at all time points, except for the shKRT19 + FSCN1 group at Day 30 (n = 11) due to tumor-associated mortality). D–F Serial bioluminescent images for each group across the indicated time points, with representative day-30 liver specimens and matching H&E sections. Relative to shCon+vector, shKRT19+vector shows lower hepatic signal; re-expression of FSCN1 in KRT19-depleted cells increases hepatic signal compared with shKRT19+vector; Note: One mouse in the shKRT19 + FSCN1 group died prior to the day 30 endpoint due to tumor-associated burden, and therefore its corresponding day-30 gross liver and H&E images are unavailable. Quantification of hepatic bioluminescence total flux (log10) across time was analyzed using a linear mixed-effects model and are provided in Fig. S10C (overall group effect P < 0.001; time×group interaction P < 0.001). G Working model. Nuclear KRT19 binds hnRNPU to facilitate β-TrCP–dependent ubiquitination of IκBα, sustaining NF-κB activity; RELA/p65 directly transactivates FSCN1, whose actin-bundling function promotes F-actin organization, cell motility, and metastatic dissemination.

Disscussion

Conventional paradigms posit that keratins primarily serve as cytoskeletal constituents responsible for maintaining cellular mechanical integrity^23,24. However, our investigation delineates and validates a novel signaling cascade through which keratins govern migratory capacity: Nuclear-localized KRT19 forms interactions with hnRNPU to potentiate β-TrCP-mediated IκBα ubiquitination and proteolysis, thereby sustaining constitutive NF-κB pathway activation and consequently augmenting transcriptional expression of downstream effector FSCN1. As a pivotal actin-bundling protein, FSCN1 serves as a potent driver of F-actin assembly, exerting a central role in cytoskeletal remodeling and the orchestration of cellular motility. Previous investigations into cellular motility have predominantly focused on the regulatory mechanisms of the microfilament and microtubule systems²⁵. In contrast, the present study unveils, for the first time, a molecular mechanism by which keratin, acting not merely as a structural scaffold but as a mediator of signal transduction, augments migratory potential. This work innovatively elucidates a novel function of KRT19 in gastric cancer, wherein it transcends its conventional structural role to drive malignant progression through nuclear signaling networks.

Within oncological contexts, the persistent activation of the NF-κB signaling axis not only underpins neoplastic proliferation and survival but also potentiates angiogenesis, distant metastasis, and therapeutic resistance^26,27. Conversely, under physiological conditions, the functions of NF-κB remain indispensable for sustaining anti-tumor immunity²⁸. Throughout metastatic progression, NF-κB functions as a pivotal orchestrator, propelling the cascade of metastasis through multifaceted mechanisms including induction of EMT, modulation of cellular adhesion and angiogenesis, facilitation of ECM degradation, and microenvironmental reprogramming^29–31. This study unveils a previously unrecognized mechanistic insight: nuclear-enriched KRT19 augments the interaction between hnRNPU and the β-TrCP complex, thereby initiating and amplifying the NF-κB signaling cascade. Our findings reveal that KRT19 is not merely a classical marker of epithelial differentiation but a dynamic driver of cellular motility, exerting a pivotal influence on the malignant progression of gastric cancer. Although most studies posit that NF-κB family members are predominantly localized within the cytoplasm^32,33, our immunofluorescence analyzes reveal that both NF-κB and its principal inhibitor, IκBα, are also distributed within the nuclei of gastric cancer cells, thereby providing theoretical support for IκBα-mediated activation of nuclear NF-κB signaling. Employing nucleocytoplasmic fractionation assays, we established that KRT19 exhibits definitive nuclear localization properties; furthermore, application of nuclear export inhibitors elicited pronounced intranuclear accumulation of KRT19 alongside concomitant escalation in IκBα ubiquitination levels—thereby cementing its direct functional role in nuclear NF-κB pathway activation. Notably, recent studies in pancreatic ductal adenocarcinoma have revealed that the head domain of KRT19 can functionally substitute for a signal peptide, enabling its entry into the endoplasmic reticulum and subsequent Golgi-dependent secretion, thereby forming a CXCL12–KRT19 “encasement” that suppresses T-cell infiltration and facilitates immune evasion. This discovery not only redefines the secretory pathway and functional paradigm of intermediate filaments but also provides a pivotal reference for the unconventional roles of KRT19 across diverse malignancies^34,35.

Heterogeneous nuclear ribonucleoprotein U (hnRNPU), a ubiquitously expressed RNA-binding protein, has been empirically demonstrated to interact with myriad mRNA species, mediating multifaceted biological processes including chromatin architecture, pre-mRNA processing, and transcriptional regulation³⁶. Current evidence demonstrates that hnRNPU exerts pivotal regulatory functions across diverse pathophysiological contexts, encompassing neurodevelopmental disorders, neurodegeneration, carcinogenesis, and muscular dystrophy^36–38; emerging evidence further suggests its modulatory involvement in canonical NF-κB signaling cascades³⁹. Through co-immunoprecipitation coupled with mass spectrometric analysis, we unequivocally confirmed the interaction between KRT19 and hnRNPU within the nuclear compartment, with KRT19 expression markedly enhancing the association between hnRNPU and the ubiquitin ligase complex. Furthermore, hnRNPU knockdown and overexpression experiments revealed a positive correlation between hnRNPU abundance and nuclear IκBα ubiquitination levels, substantiating the likelihood that KRT19 modulates NF-κB pathway activity through hnRNPU. Notably, this functional interplay between KRT19 and hnRNPU remains conspicuously unreported in extant literature, likely attributable to the predominant research emphasis on the hnRNPU–β-TrCP complex within pulmonary inflammatory environments or fibroblastic models; its investigation in gastrointestinal malignancies exhibiting elevated KRT19 expression has thus far remained unexplored. Given the aberrantly elevated expression of KRT19 across multiple malignancies of the digestive system, including gastric, esophageal, and colorectal cancers, we posit that KRT19 may similarly engage a nuclear localization–dependent mechanism in these tumors to potentiate signaling networks and the transcription of downstream effector molecules, thereby augmenting cellular motility and unleashing metastatic potential. This discovery offers a novel perspective on the role of intermediate filament proteins in orchestrating oncogenic signaling networks within the nuclear compartment.

Filopodia can be regarded as the “sensory apparatus” of the cell, adept at detecting and integrating extracellular cues while actively probing and extending toward the surrounding microenvironment, thereby playing a pivotal role in cellular motility and migration⁴⁰. FSCN1, the principal actin-bundling protein within filopodia, orchestrates the assembly of 10–30 parallel actin filaments into tightly packed bundles, giving rise to filopodial structures approximately 60–200 nm in diameter⁴¹. Initially, leveraging laser confocal microscopy and live-cell imaging, we observed that silencing KRT19 markedly impaired the ability of gastric cancer cells to maintain their spindle-shaped morphology, concomitantly reducing the number of filopodia and disrupting their dynamic extension–retraction cycles. These findings furnish the most direct morphological evidence for the role of KRT19 in orchestrating cellular motility. Secondly, through integrative analysis of Cut&Tag sequencing data and dual-luciferase reporter assays, we conclusively established that P65—the core NF-κB signaling transcription factor—directly binds to the promoter region of the FSCN1 gene, thereby enhancing its transcriptional activation. Finally, employing a time-series design within a nude mouse liver metastasis model, we dynamically evaluated the impact of KRT19 on the metastatic process, thereby further validating, at the organismal level, the conclusions drawn from our in vitro experiments. In summary, this study delineates a comprehensive mechanistic evidence chain: Nuclear-localized KRT19 activates the NF-κB signaling pathway, upregulating FSCN1 transcription, thereby augmenting the motility of gastric cancer cells and driving their migration, invasion, and distant metastasis.

While we note the unavailability of tumor grade data for stratification in the metastatic cohort (Fig. 2C), the direct pro-metastatic effects observed in our functional models support a causal driver role for KRT19 in metastasis, independent of differentiation-associated correlation. Our clinical observation that KRT19 correlates with advanced metastatic stages but not with histological dedifferentiation challenges the traditional binary view of EMT. Historically, it was postulated that cancer cells must undergo a complete EMT, shedding all epithelial features to initiate metastasis. However, contemporary consensus, as highlighted by Thiery et al. and Nieto et al., posits that the metastatic cascade is more efficiently driven by “Cellular Plasticity”—the ability of cells to dynamically toggle between phenotypic states^42–44. Specifically, we propose that KRT19-high gastric cancer cells reside in a “Partial EMT” (or Hybrid E/M) state. As described by Nieto et al., this hybrid state is not merely a transient transition but a stable, highly aggressive functional phenotype⁴³. By retaining epithelial markers, these cells maintain survival signals and the ability to migrate collectively as clusters—a trait known to confer higher metastatic potential and drug resistance compared to fully mesenchymal single cells⁴⁵. Consequently, the sustained high expression of KRT19 represents a manifestation of this epithelial-mesenchymal plasticity. This plastic state allows cancer cells to successfully navigate the invasion-metastasis cascade (resulting in high T/N stages) via NF-κB–FSCN1–mediated motility, without necessitating the complete loss of epithelial identity (thus showing no linear correlation with tumor grade). The observation that MP4 shows the maximal global NF-κB score is expected, as stress programs are enriched for canonical NF-κB–responsive inflammatory/immediate-early outputs. In contrast, our model does not require the strongest global NF-κB induction; rather, it posits a context-specific co-option in MP1, where NF-κB output is preferentially aligned with FSCN1 and cytoskeletal remodeling. Because key NF-κB nodes (e.g., IκB turnover and RELA nuclear translocation/binding) are regulated mainly at the protein/localization level, scRNA-seq is best suited to capture the transcriptional inputs/outputs (KRT19 and FSCN1/actin remodeling) that we observe in MP1, while the mechanistic linkage is established by our orthogonal functional and biochemical assays.

The role of KRT19 in metastasis appears paradoxical, likely governed by tissue specificity and subcellular localization. While studies in lung cancers indicate that downregulation of cytoplasmic KRT19 destabilizes cell junctions to initiate EMT⁴⁶, its upregulation is consistently linked to aggression in digestive malignancies¹⁵. In a previous study⁴⁷, KRT19 was shown to exert diametrically opposite functional roles in colorectal cancer and breast cancer. In colorectal cancer, KRT19 acts as an oncogenic factor by interacting with β-catenin and promoting its nuclear translocation, thereby activating the Wnt signaling pathway and enhancing tumor cell proliferation and stemness. In contrast, in breast cancer, KRT19 recruits RAC1 to form a KRT19/β-catenin/RAC1 complex, which subsequently induces NUMB transcription and suppresses Notch signaling, ultimately conferring a tumor-suppressive effect. These divergent mechanisms underscore that the functional output of KRT19 is highly context-dependent and critically shaped by the availability of specific intracellular interaction partners, such as RAC1. We propose that these discrepancies reflect distinct subcellular functions. Unlike its structural role in the cytoplasm, nuclear KRT19 functions as a transcriptional regulator. Consistent with recent findings that nuclear KRT19 represses differentiation in hepatocellular carcinoma¹³, our data demonstrate that nuclear KRT19 actively drives gastric cancer metastasis via the NF-κB axis. Thus, while loss of cytoskeletal KRT19 may facilitate mechanical detachment, its nuclear translocation represents a gain-of-function event critical for metastatic signaling.

In summary, we identify a previously unrecognized nuclear role of KRT19 in gastric cancer metastasis. Nuclear KRT19 interacts with hnRNPU to facilitate β-TrCP–mediated IκBα degradation, sustaining NF-κB activation and transcriptional upregulation of FSCN1. As a pivotal actin-bundling protein, FSCN1 drives filopodia assembly and cytoskeletal remodeling, thereby enhancing cell motility, invasion, and metastatic capacity. These findings redefine KRT19 beyond its structural role and establish a mechanistic link between nuclear keratin signaling and metastatic competence, highlighting a potential therapeutic target for KRT19-high gastric cancers.

Methods

Human tissue samples and tissue microarrays

De-identified human tissue samples were collected from January 2014 to December 2017 at Northern Jiangsu People’s Hospital (Jiangsu, China). All experimental protocols involving human tissues were approved by the Ethics Committee of Northern Jiangsu People’s Hospital, and written informed consent was obtained from all participants. All ethical regulations relevant to human research participants were followed. This study was conducted in accordance with the Declaration of Helsinki. Five tissue microarrays (TMAs) were constructed, total containing 242 pairs of gastric tumor tissues and matched adjacent non-tumor tissues. All cases were accompanied by comprehensive clinicopathological data and follow-up information.

Cell culture and shRNA-mediated knockdown and KRT19 overexpression

Human gastric cancer cell lines AGS, MKN1, and HGC27 were obtained from Procell Life Science & Technology Co., Ltd (Wuhan, China). The cell line authentication was performed by Procell Life Science & Technology Co., Ltd (Wuhan, China). AGS cells were cultured in Ham’s F-12 medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin, while MKN1 and HGC27 cells were maintained in RPMI-1640 medium with the same supplements. All cells were cultured in a humidified incubator at 37 °C with 5% CO₂, and routinely tested for Mycoplasma contamination on a monthly basis. Lentiviral vectors expressing short hairpin RNAs (shRNAs) targeting KRT19, as well as corresponding control vectors, were purchased from GeneChem Co., Ltd (Shanghai, China). AGS and MKN1 cells were transduced with either shKRT19 or negative control lentivirus. The shRNA sequence targeting KRT19 is listed in Supplementary data 5. Following transduction, cells were selected with puromycin (2 μg/mL) for 2 weeks to generate stable knockdown lines. KRT19 silencing efficiency was confirmed by Western blotting. Comprehensive details of the reagent kits employed in this study are presented in Supplementary data 6. To generate KRT19-overexpressing cell models, the full-length human KRT19 coding sequence was synthesized and cloned into a lentiviral expression vector. Corresponding empty vectors were utilized as parallel negative controls. The lentiviral particles were introduced into HGC27 cells utilizing polybrene strictly following the manufacturer’s recommended protocols. Following transduction phase, the cells were subjected to selection using puromycin (2 μg/mL) for approximately 2 weeks to establish stable overexpressing cell lines.

Single-cell RNA-seq analysis

Gastric cancer (GC) and matched normal single-cell datasets were processed in Seurat (v5.1.0) under R (v4.2.1). Quality control removed cells with <500 or >10,000 detected genes, mitochondrial transcript fraction >20%, or erythrocyte gene fraction >1%. For 3CA downstream analysis, samples with <100 epithelial cells were excluded. Data were normalized, highly variable genes identified (FindVariableFeatures), and dimensionality reduction/clustering performed (ScaleData, RunPCA, FindNeighbors, FindClusters, RunUMAP). Batch effects were mitigated using IntegrateLayers and HarmonyIntegration. Cell types were annotated by canonical markers (FindAllMarkers). To capture intratumoral heterogeneity, we applied the 3CA pipeline with non-negative matrix factorization (NMF) per sample (ranks K = 4–9), each program was defined by the top 50 genes from NMF coefficients. Robust programs were identified by overlap criteria across ranks and across samples (≥70% within-sample cross-rank overlap; ≥20% cross-sample overlap; ≤20% overlap with other programs in the same tumor). Programs were clustered by Jaccard similarity and iteratively consolidated to derive epithelial metaprograms, each finalized at 50 signature genes. For functional annotation, GSVA (v1.52.3) was used to assess pathway enrichment against MSigDB signatures (GO: BP, REACTOME, KEGG); pathways with adjusted P < 0.05 were considered significant. To reconstruct the genomic heterogeneity of gastric carcinoma, we employed infercnv (v1.22.0) to estimate somatic CNVs, utilizing epithelial cells from matched paracancerous tissues as the normal reference. The inferred CNV profiles were further stratified using K-means clustering (k = 4) and visualized via ComplexHeatmap. This unsupervised clustering allowed us to rigorously distinguish malignant cells, characterized by significant chromosomal amplifications or deletions, from genome-stable non-malignant epithelial subpopulations.

Gene expression data analysis

For bulk transcriptomic profiling, KRT19 expression in GC was analyzed using three integrated sources: (1) TCGA cohort accessed through NCBI’s GDC portal (https://portal.gdc.cancer.gov/), (2) normal gastric tissue profiles from the GTEx database (https://gtexportal.org/home/), and (3) eight Gene Expression Omnibus (GEO) datasets including seven paired tumor/adjacent tissue studies (GSE13159, GSE19826, GSE220917, GSE29272, GSE33335, GSE63089, GSE66229) and one metastasis cohort containing primary tumors with matched liver metastases (GSE237876). Batch effect correction across heterogeneous GEO datasets was implemented using the ‘limma’ package (version 3.58.0) in R statistical environment (v4.3.3) to ensure analytical comparability.

Gene set enrichment analysis

To explore signaling pathways associated with KRT19 expression in gastric cancer, patients were stratified based on KRT19 expression levels, with the top 10% considered as high-expression and the bottom 10% as low-expression groups. Differential gene expression between KRT19-high and KRT19-low cohorts was profiled using the ‘limma’ package (version 3.58.0). Subsequent pathway enrichment analysis employed ‘clusterProfiler’ (v4.10.0) to compute Normalized Enrichment Scores (NES), which quantify the robustness of gene set enrichment by normalizations accounting for gene set size and dependencies. Reference gene sets – including Hallmark collections, Gene Ontology biological processes, and transcription factor targets – were sourced from the Molecular Signatures Database (MSigDB) via the GSEA portal (https://www.gsea-msigdb.org). Gene sets demonstrating |NES| > 1.5 with false discovery rate <0.25 were deemed statistically significant, following GSEA’s recommended significance thresholds.

Immunofluorescence staining

Cell samples were fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, and blocked with 5% bovine serum albumin (BSA) to prevent nonspecific binding. Primary antibodies were incubated overnight at 4 °C. Following three washes with phosphate-buffered saline (PBS), cells in confocal dishes were exposed to Alexa Fluor 488/555/594-conjugated secondary antibodies (1:200 dilution; Beyotime, A0453, A0423; Abcam, AB175473) for 1 h in the dark. Cytoskeletal staining was performed using phalloidin (Actin-Tracker Red-Rhodamine, Beyotime, C2207S) for F-actin visualization, and nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Solarbio Life Sciences, C0060; Beijing, China). Confocal images were acquired sequentially using a laser-scanning confocal microscope (LSM 800; Carl Zeiss, Oberkochen, Germany) with a 63× oil immersion objective to eliminate channel crosstalk.

Immunohistochemistry staining

Tissue sections embedded in paraffin were deparaffinized with Histo-Clear® (National Diagnostics, Cat# HS-202), rehydrated through graded ethanol series, and subjected to antigen retrieval in citrate buffer (pH 6.0). Endogenous peroxidase activity was quenched with 3% hydrogen peroxide, followed by blocking with 5% BSA. Primary antibodies (see Supplementary data 7) were applied overnight at 4 °C. After three PBS washes, sections were incubated with horseradish peroxidase (HRP)-conjugated streptavidin (ZSGB-BIO, PV-9000) for 20 min at room temperature. Nuclei were counterstained with Mayer’s hematoxylin (MaiXin Biotechnology, Fuzhou, China). Digital imaging of tissue microarrays was performed using an Olympus FSX100 microscope (Tokyo, Japan) equipped with a UC90 20×/0.75 NA objective. IHC scoring was conducted by two independent pathologists blinded to clinical data. Each TMA core was evaluated across the entire tissue area using a semiquantitative bivariate assessment system: (A) Staining Intensity 0: None; 1+: Weak (pale yellow); 2+: Moderate (yellow-brown); 3+: Strong (brown) (B) Positive Cell Percentage 0: ≤5%; 1+: 6–25%; 2+: 26–50%; 3+: ≥51% The total score was calculated as: (Intensity score) × (Percentage score) (Range: 0–9). Samples with scores ≥1 were classified as positive expression cases. Representative fields were selected after scoring to reflect typical staining intensity and localization, avoiding edge effects, folds, necrosis, and other technical artifacts. Tumor cores contained gastric adenocarcinoma epithelium, whereas adjacent cores comprised non-malignant gastric mucosal glands.

Cell proliferation assay

Colony Formation Assay: Cells (500–1000/well) were seeded into 6-well plates and cultured for 10–14 days in complete medium. Colonies were fixed with methanol, stained with 0.5% crystal violet (Sigma-Aldrich, C3886), and counted using an automated colony counter (GelCount™, Oxford Optronix). Clusters containing >50 cells were considered valid. Data represent mean ± SEM from three independent experiments. CCK-8 Viability Assay: Cell viability was measured using CCK-8 reagent (Solarbio, CA1210). Briefly, 2 × 10³ cells/well were plated in 96-well plates. At 0, 24, 48, and 72 h, 10% CCK-8 solution was added and incubated for 2 h at 37 °C. Absorbance was measured at 450 nm (reference 650 nm) using a SpectraMax M5 reader. Results were normalized to 0 h and shown as fold-change. EdU Incorporation Assay: Cells grown on chamber slides were incubated with 10 μM EdU for 2 h, then fixed and permeabilized. EdU incorporation was detected using the EdU Cell Proliferation Kit with Alexa Fluor 594 (Beyotime, C0078L), and nuclei were counterstained with DAPI. Images were acquired with a Zeiss LSM 900 confocal microscope (40× objective), and EdU+ cell percentages were quantified from ≥5 fields using ImageJ.

Wound healing and transwell assays

Wound Healing Assay: Gastric cancer cells were grown to approximately 85% confluence in 6-well plates, and linear wounds were created using sterile 200 μL pipette tips. After washing with PBS to remove debris, wound closure was recorded at 0 and 24 h using phase-contrast microscopy (Olympus IX83). The wound width was measured at three random sites per wound, and closure was calculated as a percentage of the initial gap.

Transwell Migration and Invasion Assays: Transfected cells (5 × 10⁴) suspended in serum-free medium were seeded into the upper compartments of Transwell chambers equipped with 8 μm pore-size polycarbonate membrane inserts (Corning, Cat# 3422) fitted into 24-well plates. For migration assays, uncoated inserts were used; for invasion assays, inserts were pre-coated with diluted Matrigel™ (1:8 in serum-free medium; Vazyme, GL201). The lower chambers were filled with medium containing 25% FBS as a chemoattractant. After incubation (24 h for migration, 48 h for invasion), non-migrated cells were removed from the upper membrane surface. Cells that had migrated or invaded to the lower surface were fixed in 4% paraformaldehyde, stained with 0.5% crystal violet, and for each Transwell insert, images were acquired from five non-overlapping fields using a predefined sampling scheme (one central field and four quadrant fields), while excluding the membrane edge to avoid nonspecific cell aggregation. Migrated/invaded cells were counted in these fields and averaged to obtain a per-insert value. Experiments were repeated in three independent experiments.

Live cell imaging

Live-cell imaging was performed using a laser scanning confocal microscope (LSM880; Carl Zeiss) equipped with an environmental chamber maintained at 37 °C and 5% CO₂. Cells were observed using a 63× Plan-Apochromat oil immersion objective (NA 1.43; Carl Zeiss). Time-lapse imaging was conducted to monitor dynamic changes in cell morphology and filopodia formation. Image acquisition and quantitative analysis were carried out using ZEN Blue software (Carl Zeiss). Data were collected from three independent experiments.

3D spheroid invasion assay with GFP imaging

Gastric cancer cells stably expressing GFP were seeded into ultra-low attachment 96-well spheroid formation plates and incubated at 37 °C with 5% CO₂ for 72 h to allow spheroid formation. For invasion assessment, each well was overlaid with an invasion matrix containing type I collagen (Invitrogen), 10% FBS, and 10% sodium bicarbonate. Plates were incubated at 37 °C for 1 h to promote gelation. At 48 h post embedding, spheroid morphology and invasive outgrowth were monitored using both brightfield and GFP fluorescence imaging. Images were acquired with the JuLI™ live-cell imaging system (NanoEnTek), and representative spheroids were selected from each condition. Data were derived from three independent experiments.

Co-immunoprecipitation and mass spectrometry analysis

Co-immunoprecipitation (Co-IP) was performed to identify potential protein interaction partners. Briefly, cells were lysed in IP lysis buffer on ice for 30 min. Lysates were cleared by centrifugation at 14,000 × g for 15 min at 4 °C. Supernatants were incubated overnight at 4 °C with 5–10 μg of specific primary antibody or control IgG, followed by incubation with protein A/G magnetic beads (Vazyme, PB201) for 2 h at 4 °C with gentle rotation. After extensive washing with lysis buffer, immunocomplexes were eluted by boiling in SDS loading buffer or subjected directly to on-bead digestion. Samples for mass spectrometry analysis were sent to BGI Group (Shenzhen, China) for liquid chromatography–tandem mass spectrometry (LC-MS/MS) identification.

Dual-luciferase reporter assay

293T cells were seeded in 48-well plates (1 × 10⁵ cells/well) and transfected at 70–80% confluence using LipoCR reagent (Beyotime, China). Cells were co-transfected with firefly luciferase reporter plasmids containing either FSCN1 promoter deletion (ΔPro), wild-type promoter (Pro-WT), or point mutations at predicted binding sites (mut1, mut2), together with control or RELA expression plasmids. After 72 h, cells were lysed, and luciferase activities were measured using the Dual-Luciferase Reporter Assay Kit (Beyotime, RG088S) following the manufacturer’s instructions. Relative luciferase activity was calculated as the ratio of firefly to Renilla luminescence. All experiments were performed in triplicate.

CUT&Tag assay

CUT&Tag was performed using the High-Sensitivity CUT&Tag Assay Kit (Vazyme, TD904) following the manufacturer’s instructions. Briefly, 1 × 10⁶ AGS cells were washed twice with 1.5 mL Wash Buffer and crosslinked using TF Enhancer to stabilize transcription factor-DNA interactions. Cells were then incubated with activated concanavalin A (ConA) magnetic beads at room temperature for 15 min. Following bead attachment, cells were sequentially incubated with anti-p65 primary antibody (4 °C overnight) and appropriate secondary antibody (room temperature, 1 h). After washing, pAG-Tn5 transposase was added and incubated for 1 h to facilitate targeted integration. The tagmentation reaction was activated by adding the tagmentation buffer and incubating for 1 h, followed by termination with stop buffer at 55 °C for 10 min. Crosslinks were reversed by incubating at 80 °C for 2 h. DNA fragments were purified using Extract Beads Pro, and libraries were amplified by 16 cycles of PCR. Final libraries were sequenced on the Illumina NovaSeq platform by Nanjing Jiangbei Sequencing Center.

Nuclear and cytoplasmic protein extraction

Nuclear and cytoplasmic proteins were extracted using the Nuclear and Cytoplasmic Protein Extraction Kit (Beyotime, P0027, Shanghai, China) according to the manufacturer’s instructions. Briefly, cells were harvested by trypsinization and centrifuged to collect the cell pellet. To extract cytoplasmic proteins, the pellet was resuspended in pre-chilled Cytoplasmic Extraction Reagent A (Buffer A: PMSF = 100:1) followed by Extraction Reagent B, and incubated on ice. After centrifugation, the supernatant (cytoplasmic fraction) was collected.

The remaining nuclear pellet was resuspended in pre-chilled Nuclear Extraction Reagent containing 1% PMSF, followed by vortexing at intervals and incubation on ice for 30 min. Nuclear extracts were collected by centrifugation. All protein fractions were stored at −80 °C until further analysis.

Western blot and silver staining

Protein concentrations were determined using a BCA protein assay kit (Vazyme, Nanjing, China). Equal amounts of protein samples were separated by SDS-PAGE and transferred onto PVDF membranes (Merck, Singapore). Membranes were blocked with 5% bovine serum albumin (BSA; Solarbio, Beijing, China) for 1 h at room temperature, followed by overnight incubation at 4 °C with specific primary antibodies. After three washes with TBST buffer (10 min ×3), membranes were incubated with HRP-conjugated anti-rabbit or anti-mouse secondary antibodies for 2 h at room temperature, and washed again. Protein bands were visualized using an enhanced chemiluminescence detection kit (Biosharp, Hefei, China) and imaged with a chemiluminescence imaging system. For silver staining, proteins were separated by SDS-PAGE and processed using a Rapid Silver Staining Kit (Beyotime, P0017S). After electrophoresis, gels were sequentially subjected to fixation, washing, sensitization, and staining steps, followed by color development according to the kit protocol. All solutions were prepared using Milli-Q-grade ultrapure water.

Reverse transcription quantitative PCR (RT-qPCR)

Total RNA was extracted from cell lysates using TRIzol® reagent (Invitrogen, Thermo Fisher Scientific) following the manufacturer’s protocol. RNA concentration and purity were assessed by spectrophotometry. Complementary DNA (cDNA) was synthesized using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) with oligo(dT) primers. Gene-specific primers were obtained from PrimerBank (https://pga.mgh.harvard.edu/primerbank/). Quantitative PCR was performed using a SYBR Green I qPCR Kit (Aibotek, Wuhan, China; Cat# RK21203) on a StepOnePlus™ Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific) according to the manufacturer’s instructions. Gene expression levels were normalized to GAPDH, and relative mRNA expression was calculated using the 2^−ΔΔCq method. All reactions were run in triplicate.

In vivo tumor metastasis assay

Four-week-old male BALB/c nude mice were obtained from the Animal Experiment Center of Yangzhou University and housed under specific pathogen-free (SPF) conditions. All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Yangzhou University (Approval No. 202303827). With ethical approval granted on January 1, 2024, we have complied with all relevant ethical regulations for animal use. A total of 60 mice were used to establish a liver metastasis model via intrasplenic injection and were randomly divided into five groups (n = 12 per group): shCon, shKRT19, shCon + Vector, shCon + shKRT19, and shKRT19 + FSCN1 overexpression. All injected cells were pretreated with lentiviral vectors carrying the corresponding gene modifications. Bioluminescent imaging was performed every 5 days, for a total of six time points, to dynamically assess tumor metastasis using an in vivo imaging system (IVIS). At the final time point, mice were sacrificed, and liver metastatic nodules were harvested. Tissues were embedded in OCT, sectioned, and subjected to hematoxylin and eosin (H&E) staining for histopathological evaluation.

Statistics and reproducibility

All quantitative data are presented as mean ± SD from at least three independent experiments. All statistical analyzes were performed in R (RStudio, v4.2.1). Comparisons between two groups were conducted using two-tailed Student’s t-test, and comparisons among three or more groups were performed using one-way ANOVA. Overall survival (OS) and RFS were analyzed using the Kaplan–Meier method with log-rank tests. A two-sided p < 0.05 was considered statistically significant.

Patient consent statement and ethics approval statement

This study was approved by the Ethics Committee of Northern Jiangsu People’s Hospital, and written informed consent was obtained from all participants. All animal procedures were approved by the Institutional Animal Care and Use Committee of Yangzhou University (Approval No. 202303827).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: Jiajie Zhou, Mengli Zi, Zijie Xu; Data production, analysis and investigation: Jiajie Zhou, Yifan Cheng, Ruiqi Li, Shuai Zhao, Jie Wang, Yayan Fu, Longhe Sun, Zhen Tian, Chenkai Zhang; Writing, Review and Edit: Jiajie Zhou, Dengyang Fang, Yiqiang Zhou, Ben Li, Qiannan Sun; Supervision: Jiajie Zhou, Mengli Zi, Zijie Xu, Yifan Cheng; Funding acquisition: Jun Ren, Daorong Wang. The author(s) read and approved the final manuscript.

Peer review

Peer review information

Communications Biology thanks Yuhang Wang, Yosuke Tanaka and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor Ophelia Bu. A peer review file is available.

Data availability

All uncropped and unedited blot images are included in Supplementary Information (Figs. S12–S28). The raw sequencing data generated in this study have been deposited in the China National Center for Bioinformation/ National Genomics Data Center (CNCB-NGDC), under BioProject accession PRJCA046939 and BioSample accessions SAM150054. The data are accessible via the following persistent URL: GSA-Human: https://ngdc.cncb.ac.cn/search/specific?db=hra&q=HRA013470; All numerical source data for graphs are available on Zenodo⁴⁸. 10.5281/zenodo.18535282.

Code availability

All analysis scripts used in this study are available on Zenodo⁴⁸. 10.5281/zenodo.18535282.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-10154-7.