The present study protocol was approved by the Ethics Committee of China Medical University Hospital (approval no. CMUH108-REC1-127). Tumor mass and tumor slice specimens from patients with HNSCC were obtained from the Human Biobank of China Medical University Hospital (Taichung, Taiwan). For proteomic analysis and protein expression validation, three tumor masses from the short-term survival group and three from the long-term survival group were included. Patients with an overall survival of <2 years were classified as the short-term survival group, whereas those with an overall survival of ≥5 years were defined as the long-term survival group. For immunohistochemical analysis, tumor tissue samples from 130 patients enrolled between June 2014 and September 2016 were retrospectively analyzed. The inclusion criteria included i) newly diagnosed squamous cell carcinoma; ii) advanced-stage disease (stage III–IV); and iii) receipt of both radiotherapy and surgical treatment. Specimens were excluded if patients i) had documented HIV infection or ii) declined to provide informed consent for tissue use.

The demographic and clinicopathological characteristics of the HNSCC cohort are summarized in Table I. Pathological tissue specimens were primarily obtained from tumors of the oral cavity, tongue and oropharynx. In addition, one paired set of tumor slices derived from the primary tumor and lymph node metastasis of the same patient was also included in the analysis.

LUZP1 gene expression levels between normal tissue and primary tumor, as well as across different nodal metastasis statuses, were analyzed using the UALCAN web platform based on The Cancer Genome Atlas (TCGA) database (17). Pairwise gene expression correlation analyses were conducted using the Gene Expression Profiling Interaction Analysis (GEPIA) web server, (http://gepia.cancer-pku.cn/), which integrates TCGA and Genotype Tissue Expression data processed through a standardized analytical pipeline (18). The monotonic associations between LUZP1 and NFKB1 or NFKB2 expression levels were evaluated using Spearman's rank correlation coefficient. ChIP-seq bioinformation of cell lines was obtained from the Encyclopedia of DNA Elements (ENCODE) project database (https://www.encodeproject.org/). ENCODE cell lines are well-characterized and publicly available cell models that have undergone extensive validation, including authentication and quality control assessments. Data were accessed through the ENCODE portal and used in accordance with the project's data usage guidelines.

The human pharyngeal squamous cell carcinoma cell line FaDu (cat. no. HTB-43) was obtained from the American Type Culture Collection. The human oral squamous carcinoma cell line OECM-1 (cat. no. SCC180) was purchased from Merck KGaA and the human tongue squamous cell carcinoma cell line SAS (cat. no. JCRB0260) was obtained from Japanese Collection Research Bioresources Cell Bank. FaDu cells were maintained in Dulbecco's Modified Eagle Medium (DMEM; cat. no. 12100046; Gibco; Thermo Fisher Scientific, Inc.); OECM-1 cells were cultured in RPMI-1640 medium (cat. no. 31800022; Gibco; Thermo Fisher Scientific, Inc.) supplemented with 2 mM l-glutamine (cat. no. 5-10K00-H; BioConcept Ltd.), while SAS cells were maintained in DMEM supplemented with 2 mM l-glutamine. Docetaxel-resistant (DTXr) and cisplatin-resistant (CISr) OECM-1 cell lines were established by continuous exposure to DTX (0.2 nM; cat. no. BML-T129; Enzo Life Sciences) and CIS (0.5 µM; cat. no. 232120 Merck KGAa) for 3 months, which gradually enhanced their chemoresistance. All cells were maintained in a humidified incubator with 5% CO₂ and 95% air at 37°C.

HNSCC cell lines (FaDu, OECM-1 and SAS) were treated with pharmacological inhibitors and cytokines to evaluate their effects on LUZP1 expression. The pharmacological inhibitors used in the present study included SB431542 (5 and 10 µM; cat. no. 616464; Merck KGaA), LY294002 (5 and 10 µM; cat. no. HY-10108; MedChemExpress), rapamycin (5 and 10 µM; cat. no. R0395; Merck KGaA), BAY 11-7085 (5 and 10 µM; cat. no. HY-10257; MedChemExpress), and YC-1 (10 and 30 µM; cat. no. sc-202856; Santa Cruz Biotechnology, Inc.). Cytokine treatments included IL-1β (3 ng/ml; cat. no. C01002-100UG; Croyez Bioscience Co., Ltd.) and TNFα (10 ng/ml; cat. no. C01047-100UG; Croyez Bioscience Co., Ltd.). Cells were incubated with the indicated inhibitors or cytokines at 37°C for 24 h unless otherwise specified. For time-course analysis of LUZP1 induction, cells were treated with IL-1β (3 ng/ml) for 0, 1, 4 and 24 h or with TNFα (10 ng/ml) for 0, 6 and 24 h. In addition, TNFα (10 ng/ml) was applied to OECM-1 cells to assess its effects on tumor cell migration. Cell migratory capacity was assessed using a Transwell migration assay, as described in the Transwell migration and invasion assays section.

For the in vitro migration assay, tumor cells (2×10⁴ cells in 200 µl of media) were seeded into the upper chamber of a polyethylene terephthalate membrane Transwell insert (cat. no. 353097; Corning, Inc.) placed on a 24-well plate. For the in vitro invasion assay, tumor cells (4×10⁴ cells in 200 µl of media) were seeded into the upper chamber of Transwell inserts precoated (at 37°C for 24 h) with Matrigel (1 g/l; cat. no. 356234; BD Biosciences). Serum-free media was added to the upper chamber, while medium supplemented with 10% FBS (Corning, Inc.) was added into the lower chamber as a chemoattractant. After incubation at 37°C for 24 h (for migration assay) or 48 h (for invasion assay), the cells that migrated or invaded through the membrane were fixed with 3.7% formalin (10 min; room temperature; cat. no. 15512; Merk KGAa) and stained with 0.1% crystal violet (30 min; room temperature; cat. no. C0775; Merck KGAa). The stained cells were destained with 50% ethanol (30 min; room temperature; cat. no. 459836; Merck KGAa) and 0.1% acetic acid (cat. no. A6283; Merck KGAa), and then the absorbance of the extracted dye was measured spectrophotometrically at 570 nm for quantification and cell images were acquired using a bright-field light microscope.

Cells (3,000 cells per well) were seeded into ultra-low attachment (ULA) 96-well round-bottom plates and cultured for 3 days to allow spheroid formation. After spheroids were established, a 3D invasion assay was performed. Growth factor-reduced Matrigel (cat. no. 354230; Corning, Inc.) was diluted in serum-free DMEM to obtain a 2× working concentration. The Matrigel solution was added to pre-chilled ULA 96-well round-bottom plates containing the spheroids at a final volume of 100 µl per well, resulting in a final Matrigel concentration of 3.75 mg/ml. Plates were then incubated at 37°C for 1 h to allow Matrigel polymerization. 3D tumor spheroid invasion was quantified using a radial outgrowth distance approach. Phase-contrast images were acquired at 0, 24, 48 and 96 h following spheroid embedding using a Nikon TS100 inverted bright-field light microscope. For each spheroid, the invasion front radius (RX) at each time point was determined by measuring the radial distance from the spheroid center to the outer boundary of the contiguous invading cell population. To minimize the influence of isolated cells and image noise, the invasion front was defined as the 95th percentile (p95) of radial distance measurements obtained along multiple angular axes. The baseline spheroid radius at 0 h (R0) was defined using the same p95 criterion. Invasion distance at each time point was calculated as D=RX−R0. Spatial measurements were converted from pixels to micrometers using the scale bar provided in the 0 h images. All image analyses were performed using ImageJ software (version, 1.54 g; National Institutes of Health) and individual spheroids were analyzed independently as biological replicates for statistical analysis.

For the colony formation assay, tumor cells with or without LUZP1 knockdown (1,000 cells per well) were seeded into 6-well plates and cultured for 10 days to allow colony formation. A colony was defined as a cluster containing ≥50 cells. The resulting colonies were fixed with 3.7% formalin (10 min; room temperature) and stained with 0.1% crystal violet (30 min; room temperature). For quantitative analysis, the bound dye was solubilized using 50% ethanol containing 0.1% acetic acid, and the absorbance was measured spectrophotometrically at 570 nm.

The effects of chemotherapeutic agents, DTX and CIS, on cell viability were assessed using the methylthiazol MTT method. Tumor cells were seeded in 24-well plates at a density of 2×10⁴ cells per well and treated with DTX (0, 1, 5 and 10 nM) for 48 h or CIS (0, 5, 10 and 20 µM) for 24 h at 37°C. Following treatment, 200 µl of MTT solution [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (cat. no. M6494; Invitrogen; Thermo Fisher Scientific, Inc.); 1 g/l in PBS] was added to each well and incubated at 37°C for 4 h. The resulting formazan crystals were solubilized in DMSO (cat. no. C6164, Merck KGAa), and absorbance was measured at 570 nm using an enzyme-linked immunosorbent assay reader (DYNEX^® Technologies). Cell viability (%) was calculated as (T/U) × 100, where T represents the absorbance of treated cells and U represents absorbance of untreated cells.

Total protein lysates from homogenized tumor tissues or cultured cells were extracted using RIPA lysis and Extraction buffer (cat. no. 89901; Thermo Fisher Scientific, Inc.) supplemented with a proteinase inhibitor cocktail (cat. no. 11836170001; Roche Diagnostics GmbH). Protein concentrations were quantified using the Protein Assay kit (cat. no. 5000006EDU; Bio-Rad Laboratories, Inc.) by measuring absorbance at 595 nm. Equal amounts of protein (20 µg) were resolved by 9.5% SDS-PAGE gels and transferred onto PVDF membranes (cat. no. 88518; Thermo Scientific, Inc.) at 400 mA for 3 h. Membranes were blocked with 5% skimmed milk in TBS-0.05% Tween (TBST) for 1 h at room temperature, and incubated with primary antibodies at 4°C for 16–24 h. The primary antibodies used were as follows: Anti-LUZP1 (1:1,000; cat. no. 17483-1-AP; Proteintech Group, Inc.), anti-E-cadherin (2,000; cat. no. 3195; Cell Signaling Technology, Inc.), anti-vimentin (2,000; cat. no. 3932; Cell Signaling Technology Inc.) and anti-β-actin (10,000; cat. no. 4970; Cell Signaling Technology, Inc.). After three washes with TBST (15 min each), membranes were incubated with horseradish peroxidase-conjugated secondary antibody, peroxidase AffiniPure Goat Anti-Rabbit IgG (H+L) (10,000; cat. no. 111-035-144; Jackson ImmunoResearch Laboratories, Inc.) at room temperature for 1 h. Immunoreactive bands were visualized using an enhanced chemiluminescence substrate (Western Lighting Plus ECL; cat. no. 0RT2655, PerkinElmer, Inc.) and captured with an ImageQuant LAS 4000 luminescence image system (Cytiva). Original western blot images are provided in Table SI.

ChIP-qPCR was performed to assess NF-κB p65 (RelA) occupancy at selected genomic loci in OECM-1 cells using the SimpleChIP Enzymatic Chromatin IP Kit (Magnetic Beads) (cat. no. 9003; Cell Signaling Technology, Inc.). OECM-1 cells were cultured under standard conditions, stimulated with TNF-α (10 ng/ml; cat. no. C01047-100UG; Croyez Bioscience Co., Ltd.) for 24 h. Cells were cross-linked with 1% formaldehyde at room temperature for 10 min and quenched with glycine. Nuclei were isolated, and chromatin was enzymatically digested with micrococcal nuclease at 37°C for 25 min, followed by sonication (40 kHz on ice) using eight cycles of 20 s ON and 30 s OFF to generate chromatin fragments ranging from 150 to 900 bp, as confirmed by 1% agarose gel electrophoresis. A 2% aliquot of sheared chromatin was processed in parallel as an input control. For immunoprecipitation, 10 µg of chromatin was incubated overnight at 4°C with 5 µl of ChIP-grade anti-NF-κB p65 antibody (cat. no. 8242; Cell Signaling Technology, Inc.). Immunoprecipitation with an anti-histone H3 antibody (cat. no. 4620; Cell Signaling Technology, Inc.) was included as a positive control. Immune complexes were captured using ChIP-grade Protein G magnetic beads (cat. no. 9006; Cell Signaling Technology, Inc.) and subjected to sequential low-salt and high-salt washes. Bound chromatin was eluted and reverse cross-linked at 65°C for 2 h in the presence of proteinase K, followed by DNA purification using spin columns and elution in DNA elution buffer. qPCR was performed using a QuantStudio 5 Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific, Inc.) with KAPA SYBR FAST qPCR Master Mix (cat. no. KM4106; Roche Diagnostics). Each reaction contained 1 µl of ChIP or input DNA in a total volume of 10 µl. Technical triplicates were averaged and enrichment was calculated as a percentage of input using ΔCq normalization (19) with adjustment for the 2% input dilution. Information of qPCR primers is provided in Table SII.

Lentiviral plasmids pCMV-ΔR8.91 and pMD.G, along with specific short hairpin RNA (shRNA) constructs in the pLKO.1 vector, were obtained from the National RNAi Core Facility, Academia Sinica. A GFP-targeting shRNA (cat. no. 30323; pLKO.1 GFP shRNA) was used as a control and was purchased from Addgene, Inc. The shRNA sequences used in this study are listed in Table SIII. For lentivirus production, HEK293T cells were transiently co-transfected with the shRNA expression plasmid and the packaging plasmids pCMV-ΔR8.91 and pMD.G at a plasmid ratio of 4 µg:3.6 µg:0.4 µg (shRNA:pCMV-ΔR8.91:pMD.G). Plasmid DNA was mixed with Lipofectamine^® 2000 (Invitrogen; Thermo Fisher Scientific, Inc.) and incubated at RT for 20 min to allow complex formation, followed by transfection, according to the manufacturer's instructions. A total of 48 h after transfection at 37°C, the viral supernatants were collected, filtered through a 0.22 µm membrane filter, and used for subsequent infection. HNSCC cells were infected with shRNA-containing viral supernatant in the presence of polybrene (8 µg/ml). After 24 h of incubation at 37°C, the medium was replaced with fresh complete medium containing puromycin (2 µg/ml) for 2 weeks for selection. Cells were maintained under puromycin selection and harvested for downstream analyses as indicated in each experiment.

Tumor cells cultured on cover glasses or in 6-well plates were rinsed three times with PBS and fixed in 4% formaldehyde (cat. no. 1.04003; Merck KGAa) at room temperature for 7 min. Following fixation, cells were washed three times with PBS (5 min each). Permeabilization was performed with 0.1% Triton X-100 (cat. no. T8532; Merck KGAa) in PBS for 7 min, at room temperature, followed by three additional PBS washes. Non-specific binding was blocked with 1% BSA in PBS at room temperature for 1 h or at 4°C overnight. Tumor cells were then incubated with LUZP1 antibodies (1:100 dilution in 1% BSA/PBS; cat. no. 17483-1-AP, Proteintech Group Inc.) at room temperature for 1 h, washed and subsequently incubated with fluorophore-conjugated secondary antibodies (1:400 dilution in 1% BSA/PBS; cat. no. A-11008, Thermo Fisher Scientific, Inc.) at room temperature in the dark for 1 h. F-actin was visualized by staining with fluorescent phalloidin (1:200 dilution in 1% PBS; cat. no. R415; Invitrogen; Thermo Fisher Scientific, Inc.) at room temperature in the dark for 30 min. After three final PBS washes, cover glasses were mounted onto slides using DAPI-containing mounting medium (cat. no. 20004; AAT Bioquest, Inc.), sealed with nail polish and stored at 4°C until imaging. Fluorescent images were acquired using a fluorescence microscope (Nikon Corporation).

Formalin-fixed, paraffin-embedded clinical tissue specimens were sectioned at 4-µm thickness and baked at 60–65°C for 1 h, followed by a 10 min cooling step. Slides were deparaffinized twice in xylene (cat. no. 4312, Muto Pure Chemicals Co. Ltd.) 10 min each and rehydrated through a graded ethanol series (cat. no. 459836, Merck KGAa; 100, 95, 75 and 50%; 5 min each), then rinsed in double distilled H₂O for 5 min. Heat-induced epitope retrieval was performed in either Tris buffer (pH 9.0; cat. no. 920p-06, Merck KGAa) or citrate buffer (pH 6.0; cat. no. CBB500; ScyTek Laboratories, Inc.) under pressure at 95–100°C for 15 min, followed by cooling at room temperature for 30 min. Endogenous peroxidase activity was quenched with 3% hydrogen peroxide (cat. no. 31642, Thermo Fisher Scientific Inc.) for 10 min and sections were washed three times with PBS (10 min each). After blocking with 5% normal goat serum (cat. no. 005-000-121; Jackson ImmunoResearch Laboratories, Inc.) for 1 h at room temperature, slides were incubated overnight at 4°C with anti-LUZP1 primary antibody (1:15 dilution; cat. no. 17483-1-AP, Proteintech Group Inc.). Following three PBS washes, sections were incubated with biotinylated goat anti-rabbit IgG (1:400; cat. no. BA-1000; Vector Laboratories, Inc.) at room temperature for 1 h, and signal amplification was performed using the ABC-HRP kit (cat. no. K-6100; Vector Laboratories, Inc.). Visualization was achieved with Gill's hematoxylin (cat. no. 3801520, Leica Biosystems) for nuclear counterstaining. Finally, slides were dehydrated through graded ethanol (cat. no. 459836; Merck KGAa; 75, 95 and 100%; 3 min each), cleared in xylene (cat. no. 4312, Muto Pure Chemicals Co. Ltd.) 10 min twice and mounted with coverslips. Immunohistochemical images were acquired using a Zeiss Axio Imager Z2 upright fluorescence microscope (Duke) and quantitative analysis was conducted using Tissue FAXS cell analysis software (version, 7.1; Tissuegnostics). The immunohistochemical signals were scored as 0, 1, 2 and 3, and a score ≥2 indicated positive detection.

Total protein extracts (20 µg per sample) were separated on a 9.5% SDS-PAGE gel and divided into four gel fractions for in-gel tryptic digestion as previously described (20). Peptide mixtures were analyzed using an Orbitrap Exploris 480 mass spectrometer (Thermo Scientific, Inc.) coupled with an Ultimate 3000 RSLC nano-liquid chromatography system (Thermo Scientific, Inc.) equipped with a nanospray ionization source (Thermo Scientific, Inc.). Mass spectrometric analyses were performed in positive ion mode under data-dependent acquisition settings. Full mass spectrometry survey scans were acquired in the m/z range of 375-1,500 with an automatic gain control (AGC) target of 4×10⁵ and a resolution of 120,000 at m/z 200. The twenty most abundant multiple-charged precursor ions were sequentially fragmented by collision-induced dissociation and the tandem mass was analyzed in the Orbitrap at a resolution of 30,000. Protein identification and label-free quantification were conducted using Proteome Discovery software (version, 2.4; Thermo Scientific, Inc.), with the identification threshold set at P<0.05. Label-free quantification was conducted using Proteome Discoverer version 2.4 (Thermo Fisher Scientific, Inc.) incorporating the Minora feature detection algorithm. Quantitative analysis was based on the peak abundances of precursor ions corresponding to the identified peptides. Proteins exhibiting significant upregulation in the experimental group were subjected to identify significantly enriched biological pathways by Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis in the DAVID website (https://davidbioinformatics.nih.gov/).

All data are presented as the means ± SD. Statistical significance between two independent groups was assessed using a two-tailed unpaired Student's t-test, whereas comparisons among multiple groups were performed using one-way analysis of variance (ANOVA) followed by Tukey's post hoc test. One-to-many comparisons were conducted using one-way ANOVA followed by Dunnett's post hoc test. A factorial two-way ANOVA was performed to assess the effects of two independent variables, followed by Tukey's honestly significant difference post hoc test for multiple comparisons among groups. Correlations between gene expression levels were assessed using Spearman's rank correlation coefficient. Overall survival (OS) and progression-free survival (PFS) were analyzed using the Kaplan-Meier method, and differences between survival curves were evaluated by the log-rank test. All statistical analyses were conducted using SPSS Statistics software (version, 22.0; IBM, Corp.). P<0.05 was considered to indicate a statistically significant difference.

To identify protein signatures potentially associated with poor prognosis, a comparative proteomic analysis of tumor tissues from patients with HNSCC was performed with long- and short-term survival. Patients with an overall survival of <3 years were classified as the short-term survival group, whereas those with an overall survival of ≥5 years were defined as the long-term survival group. Among the 4,246 proteins identified, 187 exhibited a ≥4-fold increase in abundance in short-term survival compared with long-term survival tissues (Table SIV). KEGG pathway enrichment analysis revealed that these upregulated proteins were primarily associated with pathways related to fluid shear stress, drug metabolism, glutathione metabolism and chemical carcinogenesis (Fig. 1A). Fluid shear stress is associated with the metastatic dissemination of tumor cells within the circulatory system (21), whereas drug metabolism is associated with increased therapeutic resistance (22). These findings suggest that tumor metastasis and treatment resistance may represent key determinants underlying poor prognosis and short-term survival in patients with HNSCC. Differential protein expression between short- and long-term survival tissues was visualized using a proteomic volcano plot (Fig. 1B), where the x-axis represents the log₂ fold change in protein abundance and the y-axis represents the -log₁₀ p-value, indicating statistical significance. LUZP1 emerged as one of the most upregulated proteins in tumor tissues of patients with HNSCC with short-term survival (Fig. 1B; Table SIV). The elevated expression of LUZP1 was further confirmed in additional tumor specimens by western blot analysis (Fig. 1C). The LUZP1 mRNA expression level of HNSCC tissues was significantly higher when compared with that in the normal tissues via the UALCAN database (Fig. 1D), which provides an in-depth analysis of gene expression data from TCGA (17). Additionally, LUZP1 gene knockdown by shRNA significantly impaired the colony formation of OECM-1 cells (Fig. 1E), suggesting that LUZP1 could enhance the capacity for unlimited proliferation of HNSCC cells (23).

Since LUZP1 may contribute to HNSCC metastasis, its effects on the metastatic properties of HNSCC cells was examined. A one-way ANOVA analysis of the Transwell assays revealed a significant difference among Control, shLUZP1#1 and shLUZP1#2 groups (migration: F(2,6)=142.71, P=9.0×10⁻⁶; invasion: F(2,6)=193.09, P=4.0×10⁻⁶). Post hoc comparisons using Dunnett's test demonstrated that both shLUZP1#1 and shLUZP1#2 were significantly different from the Control group (shLUZP1#1 vs. Control, P=1.14×10⁻⁴; shLUZP1#2 vs. Control, P=1.82×10⁻⁴), indicating that knockdown markedly reduced the migration and invasion abilities of HNSCC cells (Fig. 2A and B). Transwell assays demonstrated that LUZP1 knockdown markedly attenuated cell migration and invasion (Fig. 2A and B). 3D tumor spheroid invasion assay was conducted to validate the role of LUZP1 in regulating tumor metastatic abilities. A one-way ANOVA revealed a significant difference among Control, shLUZP1#1 and shLUZP1#2 groups (F(2,6)=74.81, P=5.7×10⁻⁵). Dunnett's post hoc test demonstrated that both shLUZP1#1 and shLUZP1#2 were significantly reduced compared with the Control group (shLUZP1#1 vs. Control, P=2.6×10⁻⁴ and shLUZP1#2 vs. Control, P=1.1×10⁻⁴), indicating that LUZP1 knockdown significantly attenuated the invasive capacity of tumor spheroids (Fig. 2C), consistent with the previous two-dimensional migration and invasion assays (Fig. 2A and B). In addition, LUZP1 knockdown reduced the formation of invadopodia-like protrusive structures (Fig. 2D), visualized by phalloidin staining of the key cytoskeletal component F-actin (24). Furthermore, LUZP1 knockdown increased the expression of the epithelial marker E-cadherin while decreasing the expression of the mesenchymal marker vimentin (Fig. 2E), suggesting a role for LUZP1 in promoting EMT. Moreover, analysis of paired primary tumors and metastatic sites from the same patients with HNSCC revealed markedly increased LUZP1 expression in metastatic lesions compared with their matched primary tumors (Fig. 2F). This observation further supports the notion that LUZP1 enhances the metastatic potential of HNSCC cells.

KEGG pathway analysis of upregulated proteins identified in tumor tissues from short-term survival patients with HNSCC revealed significant enrichment of pathways associated with increased therapeutic resistance (Fig. 1A). This prompted the investigation of whether elevated LUZP1 expression contributes to enhanced chemoresistance in HNSCC. To this end, DTXr and CISr OECM-1 cell lines were established by continuous exposure to DTX or CIS, two chemotherapeutic agents commonly used in HNSCC treatment (25). Immunofluorescence analysis demonstrated that LUZP1 expression was markedly elevated in both OECM-1/DTXr and OECM-1/CISr cells compared with their parental counterparts (Fig. 3A). Furthermore, MTT assays revealed that LUZP1 knockdown increased the sensitivity of OECM-1 and SAS cells to DTX and CIS (Fig. 3B and C), indicating that LUZP1 may play a functional role in mediating chemoresistance in HNSCC.

Based on the aforementioned findings, the signaling pathways associated with metastasis and chemoresistance that may regulate LUZP1 expression were investigated using pathway-specific inhibitors. A total of five inhibitors were employed in the present study, including SB431542 for the TGF-β/Smad pathway, LY294002 for PI3K/AKT/mTOR signaling, Rapamycin for mTOR, BAY 11-7085 for NF-κB signaling and YC-1 for the HIF-1 pathway. Semi-quantitative western blot analysis demonstrated that BAY 11-7085 exerted the most pronounced inhibitory effect on LUZP1 expression across all three HNSCC cells (FaDu, OECM-1 and SAS; Fig. 4A), and repeated treatments with BAY 11-7085 consistently reproduced this inhibitory effect on LUZP1 protein levels (Fig. 4B). These results suggest that LUZP1 expression may be regulated by NF-κB signaling activation. Consistently, in silico analysis using the GEPIA database revealed a significant positive correlation between LUZP1 expression and that of NF-κB p105 (NFKB1) and NF-κB p100 (NFKB2) subunits (Fig. 4C). Based on the aforementioned findings that LUZP1 knockdown increased the sensitivity of HNSCC cells to DTX and CIS (Fig. 3B and C), validation of the contribution of NF-κB-LUZP1 signaling to chemoresistance was next sought. To this end, NF-κB p65 was silenced using gene knockdown by shRNA. Western blot analysis showed that NF-κB knockdown resulted in a marked reduction of LUZP1 expression (Fig. 4D). Consistently, MTT assays demonstrated that NF-κB knockdown significantly enhanced the sensitivity of both OECM-1 and SAS cells to DTX and CIS, leading to a pronounced decrease in drug IC₅₀ values (Fig. 4E and F). These results support that NF-κB-LUZP1 signaling may play a key role in mediating chemoresistance in HNSCC cells. To further validate the regulation of LUZP1 by NF-κB signaling, HNSCC cells were stimulated with the cytokines IL-1β and TNF-α, both known activators of NF-κB signaling (26). Western blot analysis showed that IL-1β and TNF-α both induced LUZP1 expression in a time-dependent manner (Fig. 4G and H). Moreover, to examine the functional consequences of NF-κB-mediated regulation of LUZP1, cell migration following LUZP1 knockdown and/or TNF-α stimulation was assessed. Using a two-way ANOVA to assess the effects of shLUZP1 and TNFα treatment, significant main effects of shLUZP1 (F(1,8)=68.87, P=3.35×10⁻⁵) and TNFα (F(1,8)=216.30, P=4.49×10⁻⁷) were observed, whereas the interaction term did not reach significant (F(1,8)=0.026, P=0.877) (Fig. 4I). Tukey's multiple comparisons post hoc test confirmed significant differences among all four groups (adjusted P≤0.01). The results indicated that TNF-α treatment enhanced the migratory ability of parental HNSCC cells and effectively rescued the migration defect caused by LUZP1 knockdown (Fig. 4I).

To examine whether LUZP1 is transcriptionally regulated by NF-κB. An in silico analysis was first carried out to evaluate potential NF-κB-mediated regulation of the LUZP1 locus. Genomic snapshots of the human genome (GRCh38/hg38 assembly) spanning the LUZP1 region (Chr 1:23,082,260-23,191,800) were obtained from the UCSC Genome Browser (27). The resulting genomic visualization (Fig. 4J) integrates multiple regulatory features, including RefSeq-curated exon annotations (28) of LUZP1, NF-κB RelA ChIP-seq binding signals in FaDu cells from the ReMap database (29), and layered H3K27ac ChIP-seq profiles from several ENCODE cell lines (30), which mark active regulatory regions and are associated with transcriptional activation. Additionally, GeneHancer regulatory element annotations (31) were included, with red boxes denoting promoter regions and gray boxes indicating putative enhancers. Together, this analysis revealed potential NF-κB binding sites within the promoter and enhancer regions of LUZP1, suggesting a regulatory relationship between NF-κB and LUZP1 expression. Based on these sequence features, 10 sets of ChIP-qPCR primers were designed targeting candidate NF-κB-enriched regions (indicated by blue vertical bars in Fig. 4J) for experimental validation. ChIP-qPCR analysis demonstrated that TNF-α stimulation significantly enhanced NF-κB binding at the LUZP1 promoter (Fig. 4K), indicating that NF-κB occupancy at the LUZP1 promoter is inducible and context dependent. Collectively, these results demonstrate that the oncogenic roles of LUZP1 could be mediated through NF-κB signaling activation.

To assess the impact of LUZP1 on clinical outcomes, IHC assays were conducted in tissue specimens to determine the correlation between LUZP1 expression levels and OS and PFS in patients with HNSCC. A total of 130 pathologic specimens were collected from patients with HNSCC, with demographic information shown in Table I. Representative IHC staining images are shown in Fig. 5A. IHC signals were semi-quantitatively scored as 0, 1, 2 or 3, and a score ≥2 was defined as positive (LUZP1⁺), as described in the Materials and methods section. The Kaplan-Meier plot analysis revealed that LUZP1⁺ was significantly associated with shorter OS and PFS in the present study cohort (Fig. 5B and C).

To determine whether LUZP1 expression independently predicts patient outcomes, multivariate Cox proportional hazards regression analyses were conducted with adjustment for age, T stage, nodal status, lymphovascular invasion (LVI), tumor differentiation, perineural invasion (PNI) and surgical margin status. High LUZP1 expression emerged as an independent prognostic factor for both OS and PFS (Tables II and III). Specifically, patients with high LUZP1 expression exhibited a significantly increased risk of mortality (OS; HR, 1.789–6.006; P<0.001) and disease progression or recurrence (PFS; HR, 2.202–9.924; P<0.001), independent of established clinicopathological variables. By contrast, tumor differentiation, LVI, PNI and surgical margin status did not retain independent prognostic significance for either OS or PFS in the multivariate models. In multivariate Cox regression analyses (Tables II and III), advanced nodal disease (N3) remained significantly associated with worse survival outcomes. Specifically, N3 was independently associated with reduced overall survival compared with N0 (Table II, P=0.029; Table III, P=0.004). Importantly, the prognostic impact of LUZP1 expression persisted after adjustment for nodal status and tumor stage, indicating that LUZP1 provides prognostic information beyond conventional staging parameters.