Multi-omics identifies VMP1 as a tumor-intrinsic biomarker associated with metastatic progression and immune microenvironment remodeling in HNSCC

Highlights

• •VMP1 is consistently upregulated across metastatic HNSCC contexts.
• •High VMP1 expression is associated with poor prognosis in HNSCC patients.
• •VMP1-high tumors exhibit reduced CD8⁺ T-cell effector features.
• •VMP1 is associated with NF-κB activation and PD-L1 upregulation.
• •VMP1 promotes proliferative, migratory, and invasive phenotypes in HNSCC cells.

Abstract

Head and neck squamous cell carcinoma (HNSCC) frequently exhibits metastatic progression and develops an immunosuppressive microenvironment. However, the tumor cell–intrinsic factors contributing to metastasis-associated immune remodeling remain incompletely understood. Here, we integrated two independent single-cell RNA-seq cohorts (GSE243933 and GSE181919) to compare primary and metastatic HNSCC ecosystems and identified vacuole membrane protein 1 (VMP1) as a malignant cell–intrinsic gene consistently upregulated across metastatic contexts. In TCGA-HNSCC and two GEO cohorts (GSE65858 and GSE117973), high VMP1 expression was associated with poor prognosis and enrichment of invasion-related programs, including epithelial–mesenchymal transition (EMT) and TNFα/NF-κB signaling, and was also associated with decreased CD8⁺ T-cell infiltration and impaired effector features. In an independent HNSCC tissue cohort, immunohistochemistry confirmed elevated VMP1 protein expression in tumors and its association with advanced clinicopathological features and unfavorable survival. Functionally, gain- and loss-of-function experiments in SCC25 and CAL27 cells showed that VMP1 enhances proliferation, migration, and invasion, accompanied by EMT-like marker switching, increased RelA/p65 phosphorylation, and elevated PD-L1 expression. Collectively, VMP1 identifies a metastasis-associated tumor-intrinsic factor associated with invasive progression and reduced CD8⁺ T-cell activity. The concomitant NF-κB activation and PD-L1 induction support a potential VMP1-associated NF-κB/PD-L1 link that may contribute to the association between invasion-related programs and immune modulation. These findings support the potential biomarker relevance of VMP1 and provide a rationale for further mechanistic and clinical investigation in HNSCC.

Graphical abstract

Introduction

Head and neck squamous cell carcinoma (HNSCC) is one of the most common and lethal malignancies worldwide [1]. Despite continuous advances in multimodal treatment strategies, including surgery, radiotherapy, chemotherapy, and immunotherapy, the overall prognosis of patients with advanced or metastatic HNSCC remains poor [[2], [3], [4]]. Tumor metastasis is a major cause of treatment failure and cancer-related mortality and is frequently accompanied by profound alterations in the tumor immune microenvironment, which are considered a key biological basis for limited therapeutic efficacy, particularly in the context of immunotherapy [[5], [6], [7], [8]].

Previous studies have demonstrated that the metastatic process of HNSCC not only reflects enhanced invasive capacity of tumor cells but is also associated with extensive remodeling of the immune microenvironment [9]. However, most existing studies are based on bulk transcriptomic analyses or focus on specific immune cell populations, making it difficult to systematically dissect, at single-cell resolution, the coordinated changes between tumor cell–intrinsic states and immune ecosystem remodeling [10]. In particular, given the marked heterogeneity of immune features across different metastatic sites, whether there exist tumor cell–intrinsic molecules that are stably upregulated across distinct metastatic contexts and potentially contribute to metastasis-associated phenotypes remains insufficiently explored [[11], [12], [13]].

Vacuole membrane protein 1 (VMP1) is a multi-pass transmembrane protein localized to the endoplasmic reticulum, originally identified in studies of acute pancreatitis, and is known to play an essential role in autophagosome formation, cellular stress adaptation, and membrane dynamics [[14], [15], [16]]. Mechanistically, VMP1 has been shown to promote autophagy induction through its interaction with Beclin 1 and to regulate SERCA-dependent ER–isolation membrane contacts required for autophagosome formation [17,18]. Increasing evidence has implicated VMP1 in tumor progression, with elevated expression reported in pancreatic cancer, glioma, colorectal cancer, and hematological malignancies, where it is associated with invasive progression and therapeutic resistance [[19], [20], [21]]. Nevertheless, existing studies have largely focused on the tumor cell–intrinsic biological functions of VMP1, while its role in HNSCC, particularly in the context of metastasis-associated immune microenvironment remodeling, remains poorly characterized.

Based on these considerations, the present study integrates multi-cohort single-cell transcriptomic data, bulk transcriptomic analyses, and clinical validation to systematically characterize the immune microenvironment associated with HNSCC metastasis and to identify tumor cell–intrinsic molecules consistently upregulated across distinct metastatic settings [22]. Furthermore, by combining in vitro functional assays and molecular analyses, we aim to elucidate the role of VMP1 in HNSCC invasive progression and to explore its potential association with the upregulation of the immune checkpoint molecule PD-L1, thereby providing new insights into metastasis-associated immune regulatory mechanisms and potential therapeutic targets.

Materials and methods

Single-cell transcriptomic data and public database acquisition

All single-cell RNA sequencing (scRNA-seq) data analyzed in this study were obtained from the Gene Expression Omnibus (GEO) database [23]. Two independent scRNA-seq datasets were included: GSE243933 and GSE181919 [[24], [25]]. The GSE243933 cohort comprised 41 head and neck squamous cell carcinoma (HNSCC) samples, including primary tumors and distant metastatic lesions [25], whereas the GSE181919 cohort consisted of 23 HNSCC samples encompassing primary tumors and lymph node metastases [24].

To further validate the robustness of the findings, transcriptomic data and corresponding clinical information of HNSCC patients were downloaded from The Cancer Genome Atlas (TCGA-HNSCC https://portal.gdc.cancer.gov/) database and used for gene expression profiling, survival analyses, and immune-related analyses. In addition, two independent GEO datasets, GSE65858 (n = 270) and GSE117973 (n = 77), were included as external validation cohorts [26,27].

Single-cell data processing and cell type annotation

Single-cell transcriptomic data were processed using the Seurat R package (version 5.4.0) following standard workflows, including quality control, normalization, identification of highly variable genes, dimensionality reduction, and unsupervised clustering. Major cell types and immune cell subsets were annotated based on canonical marker genes reported in previous studies and in reference single-cell atlases. These included epithelial/tumor cells (EPCAM, KRTs), T cells (CD3D/E), CD8⁺ T cells (CD8A/B, GZMB), regulatory T cells (FOXP3, IL2RA), macrophages (LYZ, CST3), antigen-presenting macrophages (CXCL9/10), and tumor-associated macrophages (APOC1, CCL18) [25]. The expression patterns of representative marker genes for each cell population are shown in Fig. 1.

Fig. 1.

Single-cell landscape reveals immune remodeling and transcriptional reprogramming in metastatic HNSCC. (A) Heatmap showing canonical marker gene expression across major cell types identified in the GSE243933 cohort (n = 41; 30 primary tumors and 11 metastatic lesions). (B) UMAP visualization of primary (n = 30) and metastatic (n = 11) HNSCC lesions in the GSE243933 cohort, colored by cell type annotation. (C) Bar plots showing the relative proportions of major cell populations in primary and metastatic tumors, including total cells and CD45⁺ immune compartments. Cell proportion comparisons were performed at the sample level. (D) Heatmap displaying marker gene expression across T-cell subclusters. (E) UMAP visualization of annotated T-cell subsets in primary and metastatic lesions. (F) Proportion of T-cell subsets in primary versus metastatic tumors. Cell proportion comparisons were performed at the sample level. (G) Heatmap showing marker gene expression of macrophage subpopulations. (H) UMAP visualization of macrophage subclusters in primary and metastatic lesions. (I) Proportion of macrophage subsets in primary versus metastatic tumors. Cell proportion comparisons were performed at the sample level. (J) Volcano plot showing differentially expressed genes between primary and metastatic lesions. (K) Hallmark gene set enrichment analysis (GSEA) indicating activated and suppressed pathways in metastatic tumors compared with primary tumors. (L) KEGG pathway enrichment analysis showing significantly altered signaling pathways in metastatic lesions. For panels C, F, and I, cell proportions were summarized at the tumor-sample level and are shown descriptively. Differential expression and enrichment analyses in panels J–L were performed as described in the Methods.

Identification of metastasis-associated candidate genes

In both GSE243933 and GSE181919 scRNA-seq datasets, differential gene expression analyses were performed between tumor cells from metastatic lesions and primary tumors using the FindMarkers function implemented in Seurat. Genes significantly upregulated in metastatic tumor cells were identified after multiple testing correction, with thresholds set at adjusted P 〈 0.05 and |avg_log2FC| 〉 0.5.

Subsequently, intersecting analyses were conducted using R to identify candidate genes consistently upregulated across different metastatic contexts. Gene list processing was performed using the dplyr and tidyr packages, and the overlap between datasets was visualized using Venn diagrams.

Survival analysis

Survival analyses were conducted using TCGA-HNSCC and GEO validation cohorts. Patients were stratified into high- and low-expression groups based on the optimal cutoff value for candidate gene expression, determined using the survminer package in R. This cutoff strategy was used specifically for prognostic analyses to identify the expression threshold with the greatest discriminatory ability for survival outcomes. Kaplan–Meier survival curves were generated, and statistical significance between groups was evaluated using the log-rank test.

Bulk RNA-seq differential expression and pathway enrichment analyses

Differential gene expression analysis of TCGA-HNSCC RNA-seq data was performed using the DESeq2 package by stratifying samples into VMP1-high and VMP1-low groups according to the median expression level of VMP1. The median cutoff was used for transcriptomic and pathway-level comparisons to obtain balanced groups, reduce cutoff-driven bias, and improve comparability across cohorts. For GEO microarray datasets, differential analysis was conducted with the limma package under the same median-based grouping strategy. Multiple testing correction was applied to identify significantly differentially expressed genes.

Kyoto Encyclopedia of Genes and Genomes (KEGG https://www.genome.jp/kegg/docs/relnote.html) pathway enrichment analysis was performed using the clusterProfiler package. In addition, Gene Set Enrichment Analysis (GSEA https://www.gsea-msigdb.org/gsea/index.jsp) was conducted using the Hallmark gene sets, with multiple testing correction performed using the Benjamini-Hochberg method.

Bulk RNA immune infiltration analysis

Immune cell infiltration was estimated from bulk transcriptomic data using CIBERSORTx with the LM22 leukocyte gene signature matrix. For each sample, CIBERSORTx was run with 1000 permutations, and only deconvolution results with CIBERSORTx P < 0.05 were retained for downstream comparisons [28]. Differences in immune cell composition between VMP1-high and VMP1-low groups defined by the median VMP1 expression level were compared. Furthermore, single-sample GSEA (ssGSEA) was applied to score functional signatures related to CD8⁺ T cells and macrophages, and correlations with VMP1 expression levels were assessed. Immune infiltration analyses were primarily performed using TCGA-HNSCC RNA-seq data, while GEO microarray datasets were used for correlation validation.

Single-cell immune microenvironment analysis

For single-cell immune microenvironment analysis, tumor samples were stratified into VMP1-high and VMP1-low groups based on the median sample-level expression of VMP1 in malignant tumor cells. This median-based strategy was used for single-cell immune microenvironment analyses to maintain balanced sample groups and to ensure consistency with the transcriptomic and pathway-level comparisons. Gene set activity scores for CD8⁺ T cell functional signatures and M2 macrophage polarization signatures, derived from previously published studies, were calculated using the UCell algorithm [29]. Scores were summarized according to tumor sample-level VMP1 grouping. Differences in gene set activity between groups were evaluated using the Wilcoxon rank-sum test.

Tissue specimens and immunohistochemistry

A total of 88 HNSCC tumor tissue samples and 32 paired adjacent normal tissue samples were collected for immunohistochemical (IHC) analysis of VMP1 expression. The institutional HNSCC tissue cohort included 35 oropharyngeal squamous cell carcinoma cases, 30 oral cavity squamous cell carcinoma cases, 13 laryngeal squamous cell carcinoma cases, and 10 hypopharyngeal squamous cell carcinoma cases. All tumor specimens were obtained from treatment-naïve patients who underwent primary surgical resection. Patients who had received preoperative chemotherapy, radiotherapy, immunotherapy, targeted therapy, or neoadjuvant treatment were excluded to minimize treatment-related effects on tumor cell states and immune microenvironmental features. p16 status was assessed by immunohistochemistry as part of pathological evaluation when available. In this cohort, p16 status was available for 67 patients, including 30 p16-positive and 37 p16-negative cases. All tissue samples were obtained with approval from the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital, in accordance with the Declaration of Helsinki, and written informed consent was obtained from all patients.

Formalin-fixed, paraffin-embedded tissue sections were deparaffinized and rehydrated, followed by blocking of endogenous peroxidase activity with 0.3% hydrogen peroxide and non-specific binding with goat serum. Sections were incubated with a primary antibody against VMP1 at 4 °C overnight, followed by incubation with the appropriate secondary antibody and visualization using DAB. Hematoxylin was used for counterstaining.

IHC staining was independently evaluated by two pathologists blinded to clinical information. VMP1 expression was assessed using a semi-quantitative scoring system based on staining intensity (0–3) multiplied by the percentage of positive tumor cells (0–4). An IHC score <6 was defined as low expression, whereas a score ≥6 was considered high expression. Associations between VMP1 expression and clinicopathological characteristics were analyzed using the χ² test. Survival analyses were performed using the Kaplan–Meier method and log-rank test. Univariate and multivariate Cox regression analyses were conducted using Cox proportional hazards models to evaluate the prognostic relevance of VMP1 and clinicopathological variables. Variables included in the multivariate Cox regression model were selected based on clinical relevance, data completeness, and avoidance of collinearity.

Cell culture

Human HNSCC cell lines SCC25 and CAL27, as well as HEK293T cells, were obtained from the American Type Culture Collection (ATCC). All cell lines were authenticated by short tandem repeat (STR) profiling and routinely tested for mycoplasma contamination. SCC25 cells were cultured in DMEM/F12 medium, while CAL27 and HEK293T cells were maintained in DMEM, both supplemented with 10% fetal bovine serum, at 37 °C in a humidified atmosphere containing 5% CO₂.

Antibodies and reagents

Primary antibodies used for Western blotting (WB) and immunohistochemistry (IHC) included antibodies from Cell Signaling Technology (Danvers, MA, USA): anti-VMP1 (#12,929; WB 1:1000, IHC 1:200), anti-p65 (#8242; WB 1:1000), and anti-phospho-p65 (p-p65, #3033; WB 1:1000). Antibodies from Proteintech (Rosemont, IL, USA; Wuhan, China) included anti-E-cadherin (60,335-1-Ig; WB 1:1000, IHC 1:200), anti-Vinculin (26,520-1-AP; WB, loading control), and anti-PD-L1/CD274 (28,076-1-AP; WB and IHC). Anti-N-cadherin antibody (ab98952; WB 1:1000) was obtained from Abcam (Cambridge, UK).

Construction of stable VMP1 knockdown and overexpression cell lines

To generate stable VMP1 knockdown cell lines, short hairpin RNA (shRNA) sequences targeting VMP1 were cloned into lentiviral vectors. For overexpression, the full-length coding sequence of VMP1 was cloned into a lentiviral expression vector. Recombinant lentiviruses were packaged in HEK293T cells by co-transfection with packaging plasmids. Viral supernatants were collected at 48 h and 72 h post-transfection, filtered through 0.45 μm membranes, and used to infect HNSCC cells in the presence of Polybrene (10 μg/mL). Stable cell lines were selected using appropriate antibiotics, and VMP1 expression levels were validated by quantitative PCR and Western blotting. Two shRNA sequences were used: shVMP1#1 (sense: 5′-GGCAUCGUCAAAGCAUGUTT-3′;antisense:5′-ACAAUGCUUGUGACGAUGCCTT-3′) and shVMP1#2(sense: 5′-GGCAGAAUAUUGUCCUGUGTT-3′;antisense:5′-CACAGGACAAUAUUCUCUGCCTT-3′). shVMP1#1 demonstrated more stable knockdown efficiency and was therefore used in subsequent functional assays.

Western blot analysis

Cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors on ice for 30 min. After centrifugation at 12,000 rpm for 15 min, supernatants were collected and protein concentrations were determined using the BCA assay. Equal amounts of protein (15–20 μg) were separated by SDS-PAGE and transferred onto PVDF membranes. Membranes were blocked with 5% non-fat milk for 1 h at room temperature and incubated with primary antibodies at 4 °C overnight, followed by HRP-conjugated secondary antibodies. Protein bands were visualized using enhanced chemiluminescence (ECL) and quantified using ImageJ software. All experiments were independently repeated at least three times.

RNA extraction and quantitative real-time PCR

Total RNA was extracted using TRIzol reagent according to the manufacturer’s instructions, and RNA concentration and purity were assessed using a NanoDrop spectrophotometer. Complementary DNA (cDNA) was synthesized using a reverse transcription kit. Quantitative real-time PCR was performed using a SYBR Green system on an ABI 7500 Real-Time PCR instrument. GAPDH served as an internal control, and relative gene expression levels were calculated using the 2⁻ΔΔCt method. All experiments were performed in triplicate. Primer sequences were as follows: VMP1 forward, TTGGAACAGGGCTGCACACCTT; reverse, TCAGGATAGGGTGGTTCGGGAA. GAPDH forward, TGCACCACCAACTGCTTAGC; reverse, GCATGGACTGTGGTCATGAG.

CCK-8 cell proliferation assay

HNSCC cells were seeded into 96-well plates at a density of 1.0 × 10³ cells per well. At the indicated time points, CCK-8 reagent (APExBIO, Shanghai, China) was added to each well and incubated at 37 °C for 2 h. Absorbance at 450 nm was measured using a microplate reader, and cell growth curves were generated to evaluate cell proliferation.

Colony formation assay

SCC25 or CAL27 cells were seeded at low density (500 cells/well) in 6-well plates and cultured for 10–14 days. Colonies were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. Colonies containing ≥50 cells were counted as valid colonies and subjected to statistical analysis.

EdU incorporation assay

DNA synthesis activity was assessed using an EdU incorporation assay (UE Biosciences, Shanghai, China). HNSCC cells in the logarithmic growth phase were seeded into 12-well plates at a density of approximately 8.0 × 10⁴ cells per well. After attachment, cells were incubated with EdU at a final concentration of 50 μM for 2 h. Cells were then fixed, permeabilized, and subjected to fluorescent labeling according to the manufacturer’s instructions, with all staining procedures performed protected from light. Nuclei were counterstained with DAPI. Images were acquired using an APEXVIEW APX100 digital imaging system (Evident, formerly Olympus), and EdU-positive cells were quantified from three randomly selected microscopic fields per well. All experiments were independently repeated three times.

Wound healing assay

Cell migration was evaluated using a wound healing assay. HNSCC cells were grown to 90–100% confluence in 6-well plates, and linear scratches were generated using sterile 200 μL pipette tips. After gentle washing with PBS to remove detached cells, serum-free medium was added. Images were captured at 0 h and 24 h (or at indicated time points) at the same locations. Migration was quantified by measuring wound width changes using ImageJ software.

Transwell migration and invasion assays

Transwell assays were performed to assess cell migration and invasion. For migration assays, Transwell inserts without Matrigel coating were used, whereas invasion assays were conducted using inserts pre-coated with Matrigel to mimic the extracellular matrix barrier. HNSCC cells (8 × 10⁴) suspended in serum-free medium were seeded into the upper chambers, while medium containing 10% or 20% fetal bovine serum was added to the lower chambers as a chemoattractant. After 24 h incubation at 37 °C, non-migrated cells were removed, and migrated or invaded cells on the lower surface were fixed, stained with crystal violet, and counted in randomly selected fields.

Statistical analysis

All in vitro experiments were independently repeated at least three times, and data are presented as mean ± standard deviation (SD). Differences between two groups were assessed using two-tailed Student’s t-tests, while comparisons among multiple groups were performed using one-way analysis of variance (ANOVA) with appropriate post hoc tests. Categorical variables were analyzed using the χ² test as appropriate. Differences in IHC scores between adjacent normal tissues and HNSCC tumor tissues were evaluated using the Mann–Whitney U test. Associations between IHC scores and clinicopathological characteristics were evaluated using the χ² test. Survival curves were generated using the Kaplan–Meier method and compared using the log-rank test. Univariate and multivariate survival analyses were conducted using Cox proportional hazards regression models. Hazard ratios (HRs), 95% confidence intervals (CIs), and P values were reported.

For single-cell transcriptomic analyses, differences in cell proportions and functional scores between groups were assessed using the Wilcoxon rank-sum test. All statistical analyses were performed using GraphPad Prism (version 10.0) and R software (version 4.5.2). A two-sided P value < 0.05 was considered statistically significant. In figures, *, **, ***, and **** denote P < 0.05, P < 0.01, P < 0.001, and P < 0.0001, respectively.

Results

Single-cell atlas reveals immune microenvironment remodeling in metastatic HNSCC

Using scRNA-seq data from the GSE243933 cohort (n = 41; 30 primary tumors and 11 metastatic lesions), unsupervised clustering identified malignant cells, epithelial cells, CAFs, endothelial cells, and multiple immune populations with canonical markers (Fig. 1A). UMAP visualization showed a broadly conserved cellular framework between primary and metastatic samples (Fig. 1B). However, metastatic lesions exhibited increased malignant cell and macrophage proportions and reduced T-cell abundance, including within the CD45⁺ compartment (Fig. 1C). T-cell subclustering revealed an increased proportion of Tregs with reduced CD8⁺ cytotoxic effector T cells in metastases (Fig. 1D,F). Macrophages segregated into CXCL9/10⁺ antigen-presenting, FCN1⁺IL1B⁺ inflammatory, S100A8/9⁺ monocyte-like inflammatory, and APOC1⁺CCL18⁺ TAM subsets (Fig. 1G). Metastatic lesions were enriched for APOC1⁺CCL18⁺ TAMs and inflammatory subsets, and depleted for CXCL9/10⁺ antigen-presenting macrophages (Fig. 1I). Differential expression and enrichment analyses further distinguished metastases from primaries (Fig. 1J). Metastatic lesions showed enrichment of stress- and adhesion-associated programs, along with downregulation of MYC-, translation-, and proteostasis-related pathways (Fig. 1K,L). Collectively, metastatic lesions displayed an immune shift characterized by fewer effector T cells and more TAMs, together with altered stress, adhesion, and proteostasis programs.

Distinct immune ecological features across metastatic sites and identification of metastasis-associated consistently upregulated genes

To compare immune ecological features across metastatic sites, we profiled primary tumors and lymph node metastases using the GSE181919 cohort. Unsupervised clustering identified malignant cells, T cells, B/plasma cells, dendritic cells, macrophages, fibroblasts, endothelial cells and mast cells with canonical marker expression (Fig. 2A). UMAP visualization showed a broadly conserved cellular framework between primary tumors and lymph node metastases (Fig. 2B). However, lymph node metastases showed increased T-cell abundance and reduced malignant cell and macrophage proportions, including within the CD45⁺ compartment (Fig. 2C,D). T-cell subclustering showed more CD8⁺ cytotoxic effector T cells and fewer Tregs in lymph node metastases (Fig. 2E). Macrophage composition shifted toward increased CXCL9/10⁺ antigen-presenting macrophages and decreased APOC1⁺CCL18⁺ TAMs (Fig. 2F). Differential expression and pathway analyses further distinguished lymph node metastases from primary tumors (Fig. 2G). Lymph node metastases showed enrichment of immune-related programs and cell cycle/proliferation signatures, including E2F targets and G2M checkpoint, together with activation of antigen processing and presentation pathways and downregulation of ECM-receptor interaction (Fig. 2H,I). To identify metastasis-associated genes consistently upregulated across distinct metastatic contexts, we intersected genes upregulated in metastases versus primaries between GSE243933 (distant metastases) and GSE181919 (lymph node metastases), yielding five shared candidates (KRT15, CALML5, CDKN2A, LSP1 and VMP1; adjusted P < 0.05, |avg_log2FC| > 0.5) (Fig. 2J), which showed broader and/or stronger expression in metastatic samples at single-cell resolution (Fig. 2K,L). Collectively, lymph node metastases exhibited an immune ecological shift with increased T-cell representation and enhanced immune-related programs, while cross-cohort intersection identified a small set of consistently upregulated metastatic candidate genes.

Fig. 2.

Immune microenvironmental features of primary tumors and lymph node metastases and identification of commonly upregulated candidate genes in the independent single-cell cohort GSE181919. (A) UMAP visualization shows the overall distribution of major cell types, including malignant tumor cells, T cells, B/plasma cells, dendritic cells, macrophages, fibroblasts, endothelial cells, and mast cells in the GSE181919 cohort (n = 23; 19 primary tumors and 4 lymph node metastases). (B) UMAP plots of primary tumors (n = 19) and lymph node metastases (n = 4) in the GSE181919 cohort. (C) Overall cellular composition proportions in primary tumors and lymph node metastases. Cell proportion comparisons were performed at the sample level. (D) Proportional changes of CD45⁺ immune cell subsets between primary and lymph node metastatic lesions. Cell proportion comparisons were performed at the sample level. (E) Distribution of T-cell subsets in primary tumors and lymph node metastases. Cell proportion comparisons were performed at the sample level. (F) Distribution of macrophage subsets in primary tumors and lymph node metastases. Cell proportion comparisons were performed at the sample level. (G) Volcano plot showing differentially expressed genes between primary and lymph node metastatic lesions. (H) Hallmark gene set enrichment analysis indicating significantly activated or suppressed pathways in lymph node metastases. (I) KEGG pathway enrichment analysis highlighting significantly activated or suppressed signaling pathways in lymph node metastases. (J) Intersection analysis of upregulated genes in metastatic lesions relative to primary tumors across the GSE243933 and GSE181919 single-cell cohorts. Differentially expressed genes were defined as those with adjusted P value (p_val_adj) < 0.05 and average log2 fold change (avg_log2FC) > 0.5. (K) Single-cell expression patterns of the identified candidate genes in primary and lymph node metastatic lesions within GSE181919. (L) Comparative expression patterns of these candidate genes in malignant tumor cells from primary and metastatic lesions in the GSE243933 cohort.

VMP1 is associated with poor prognosis, invasive molecular programs, and impaired antitumor immunity across multiple cohorts

Based on the candidate genes consistently upregulated across metastatic contexts, we next evaluated their expression patterns and clinical relevance in bulk cohorts. In the TCGA-HNSCC cohort, VMP1 and CDKN2A showed prominent upregulation in tumors compared with normal tissues (Fig. 3A). Kaplan–Meier analysis showed that high VMP1 expression was associated with worse overall survival, whereas higher CDKN2A expression was associated with a more favorable survival outcome (Fig. 3B,C). Therefore, subsequent analyses focused on VMP1.

Fig. 3.

Expression patterns, prognostic significance, and immune associations of VMP1 in bulk transcriptomic cohorts. (A) Heatmap showing the expression of candidate genes (VMP1, CALML5, LSP1, KRT15, and CDKN2A) in normal and tumor tissues from the TCGA-HNSCC cohort. (B) Kaplan–Meier overall survival (OS) curves for TCGA-HNSCC patients stratified into VMP1-high and VMP1-low groups using the optimal cutoff value. Statistical significance was evaluated using the log-rank test. (C) Kaplan–Meier OS curves for TCGA-HNSCC patients stratified into CDKN2A-high and CDKN2A-low groups using the optimal cutoff value. Statistical significance was evaluated using the log-rank test. (D–E) Kaplan–Meier OS curves for patients in the GEO validation cohorts GSE65858 (n = 270) (D) and GSE117973 (n = 77) (E), stratified into VMP1-high and VMP1-low groups using the optimal cutoff value. Statistical significance was evaluated using the log-rank test. (F) KEGG pathway enrichment analysis comparing VMP1-high versus VMP1-low groups in the TCGA-HNSCC cohort, with samples stratified according to the median VMP1 expression level. (G) Hallmark gene set enrichment analysis comparing VMP1-high versus VMP1-low groups in the TCGA-HNSCC cohort, with samples stratified according to the median VMP1 expression level. (H) Differences in immune cell infiltration proportions estimated by CIBERSORTx between VMP1-high and VMP1-low groups in the TCGA-HNSCC cohort, with groups defined by the median VMP1 expression level. Statistical significance was evaluated using the Wilcoxon rank-sum test. *P < 0.05. (I) Correlation analysis between VMP1 expression levels and ssGSEA scores of CD8⁺ T cells and macrophages in the TCGA-HNSCC cohort using Spearman’s correlation. (J) Hallmark pathway enrichment analysis comparing VMP1-high versus VMP1-low groups in the GSE65858 cohort, with samples stratified according to the median VMP1 expression level. (K) KEGG pathway enrichment analysis comparing VMP1-high versus VMP1-low groups in the GSE65858 cohort, with samples stratified according to the median VMP1 expression level. (L) Correlation analysis between VMP1 expression levels and ssGSEA scores of CD8⁺ T cells and macrophages in the GSE65858 cohort using Spearman’s correlation.

For transcriptomic and pathway-level comparisons, samples were stratified into VMP1-high and VMP1-low groups using the median VMP1 expression level, as described in the Methods. In TCGA-HNSCC, VMP1-high tumors were enriched for cell cycle/proliferation-related pathways and Hallmark programs including epithelial–mesenchymal transition (EMT) and TNFα/NF-κB signaling, whereas immune-related pathways were broadly suppressed (Fig. 3F,G). Consistently, CIBERSORTx immune infiltration analysis showed reduced CD8⁺ T-cell infiltration in VMP1-high tumors (Fig. 3H). ssGSEA analysis further showed a negative association between VMP1 expression and CD8⁺ T-cell-related immune signature scores, with additional changes in macrophage-related signature scores (Fig. 3I).

These patterns were further supported in the independent GEO cohort GSE65858. High VMP1 expression was associated with poorer survival, and VMP1-high tumors showed similar enrichment of cell cycle/stress- and EMT-related programs with reduced immune-related signatures (Fig. 3D,J,K). VMP1 expression also negatively correlated with CD8⁺ T-cell- and macrophage-related ssGSEA scores in this cohort (Fig. 3L). In GSE117973, VMP1-high tumors showed similar pathway-level features and lower CD8⁺ T-cell immune scores, whereas macrophage-related scores did not show a clear trend (Fig. S1A–D). Collectively, high VMP1 expression was associated with unfavorable survival, invasive molecular programs, and reduced CD8⁺ T-cell immune activity across multiple cohorts.

Single-cell analyses associate high VMP1 expression with impaired CD8⁺ T-cell effector states and myeloid remodeling

Based on these bulk transcriptomic findings, we next used single-cell data to examine VMP1-associated immune microenvironmental features, focusing on CD8⁺ T-cell states and macrophage remodeling. The GSE243933 cohort (n = 41) was stratified into VMP1-high and VMP1-low groups based on the median sample-level VMP1 expression in malignant tumor cells. Major immune populations were identified in both groups (Fig. 4A). VMP1-high tumors showed reduced T-cell abundance and myeloid remodeling (Fig. 4B-C).

Fig. 4.

Immune microenvironmental differences between VMP1-high and VMP1-low tumors in the GSE243933 single-cell cohort. (A) UMAP visualization of major immune cell populations in the GSE243933 cohort (n = 41). Tumor samples were stratified into VMP1-high and VMP1-low groups according to the median sample-level VMP1 expression in malignant tumor cells, yielding 20 VMP1-high and 21 VMP1-low samples. (B-C) Comparison of overall immune cell composition between VMP1-high and VMP1-low tumors defined by sample-level VMP1 expression in malignant tumor cells. (D) UMAP projection and composition analysis of T-cell subsets between VMP1-high and VMP1-low tumors. (E) Comparison of T-cell functional state scores between VMP1-high and VMP1-low tumors, including cytotoxicity-related scores, IFNG signaling, and exhaustion-associated gene signature scores. Scores were summarized according to tumor sample-level VMP1 grouping. (F) UMAP visualization and composition analysis of macrophage subsets between VMP1-high and VMP1-low tumors. (G) Comparison of macrophage polarization scores between VMP1-high and VMP1-low tumors defined by sample-level VMP1 expression in malignant tumor cells. Statistical significance was evaluated using the Wilcoxon rank-sum test. (H) UMAP visualization of major immune cell populations in metastatic samples from the GSE243933 cohort (n = 11). Metastatic tumor samples were stratified into VMP1-high and VMP1-low groups according to the median sample-level VMP1 expression in malignant tumor cells within the metastatic subset, yielding 5 VMP1-high and 6 VMP1-low metastatic samples. (I-J) Comparison of T-cell subset composition between VMP1-high and VMP1-low metastatic tumors defined by sample-level VMP1 expression in malignant tumor cells. (K-L) Comparison of macrophage subset composition between VMP1-high and VMP1-low metastatic tumors defined by sample-level VMP1 expression in malignant tumor cells. Cell-type composition was summarized at the tumor-sample level and is shown descriptively unless statistical annotations are indicated.

T-cell subclustering showed fewer CD8⁺ cytotoxic effector T cells, a relative increase in Tregs, decreased cytotoxicity/IFNG-related scores, and increased exhaustion-associated signatures in VMP1-high tumors (Fig. 4D–E). Macrophage analysis showed reduced CXCL9/10⁺ antigen-presenting macrophages, increased APOC1⁺CCL18⁺ TAMs, and higher M2 polarization scores in VMP1-high tumors (Fig. 4F-G). In the metastatic subset (n = 11), VMP1-high metastases showed increased macrophage abundance, reduced CD8⁺ cytotoxic effector T cells, and enrichment of APOC1⁺CCL18⁺ TAMs with reduction of CXCL9/10⁺ antigen-presenting macrophages, although overall T-cell proportions changed only minimally (Fig. 4H-L).

These immune features were further supported in the independent GSE181919 cohort (n = 23). VMP1-high samples showed reduced cytotoxic CD8⁺ effector T cells, lower cytotoxicity/IFNG scores, increased exhaustion signatures, and a consistent M2 polarization trend (Fig. S2A-G). Analyses of the metastatic subset in this cohort (n = 4) were descriptive because of limited statistical power, but suggested a similar reduction in cytotoxic CD8⁺ effector T cells (Fig. S2H-J). Collectively, VMP1-high status was associated with reduced CD8⁺ effector features and myeloid remodeling across single-cell cohorts.

Elevated VMP1 expression in HNSCC tissues and its unfavorable clinical associations

Based on the convergent evidence from single-cell and transcriptomic analyses, we further assessed VMP1 expression and its clinical relevance at the tissue level using an institutional cohort of treatment-naïve HNSCC patients who underwent primary surgical resection. p16 status was available for 67 patients. Immunohistochemistry (IHC) showed markedly stronger VMP1 staining in HNSCC tumor tissues than in adjacent normal epithelium, where staining was weak or nearly undetectable. Tumor cells exhibited diffuse cytoplasmic VMP1 positivity (Fig. 5A). Semi-quantitative IHC scoring confirmed significantly higher VMP1 scores in tumors compared with normal controls (P < 0.0001) (Fig. 5B). Based on IHC scores, 88 HNSCC patients were stratified into VMP1-high and VMP1-low groups. Kaplan–Meier analysis showed significantly worse overall survival in the VMP1-high group (log-rank P = 0.0185; HR = 1.965; 95% CI, 1.057–3.653) (Fig. 5C). To further evaluate the prognostic relevance of VMP1, univariate and multivariate Cox regression analyses were performed. High VMP1 expression remained significantly associated with worse overall survival after adjustment for age, gender, and clinical stage, and the detailed results are provided in Supplementary Tables S1 and S2. Clinicopathological correlation analysis showed no significant association between VMP1 expression and sex, age, p16 status, or recurrence status. However, high VMP1 expression was more frequently observed in patients with advanced clinical stage (III–IV, P = 0.0324) and was significantly associated with higher T stage (T3–T4, P < 0.0001) and lymph node metastasis (P < 0.0001) (Fig. 5D). Collectively, tissue-level IHC analyses support elevated VMP1 expression in HNSCC and its association with unfavorable survival and advanced clinicopathological features.

Fig. 5.

Expression of VMP1 in HNSCC tissues and its clinical significance. (A) Representative immunohistochemical (IHC) staining images of VMP1 in adjacent normal tissues and HNSCC tumor tissues (original magnification, × 200; scale bar = 100 μm). (B) Semi-quantitative analysis of VMP1 IHC scores comparing adjacent normal tissues (n = 32) and HNSCC tumor tissues (n = 88). Statistical significance was evaluated using the Mann–Whitney U test. (C) Kaplan–Meier overall survival (OS) analysis of 88 patients with HNSCC, stratified into VMP1-high and VMP1-low groups according to the IHC score cutoff. Statistical significance was evaluated using the log-rank test. (D) Associations between VMP1 expression status (IHC high vs low) and clinicopathological features of 88 patients with HNSCC. Statistical analysis was performed using the Pearson χ² test. p16 status was available for 67 patients, and recurrence information was available for 51 patients. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

VMP1 promotes proliferative capacity of HNSCC cells

Building on the TCGA/GEO bulk transcriptomic findings that VMP1-high tumors were enriched for cell cycle/proliferation programs, together with tissue-level IHC associations between high VMP1 expression and advanced clinical stage and higher T classification, we next asked whether VMP1 modulates proliferative capacity in HNSCC cells. Western blotting and qPCR showed that VMP1 was upregulated across multiple HNSCC cell lines to varying degrees compared with normal human oral keratinocytes (HOK), with relatively higher expression in SCC25 and CAL27 (Fig. 6A,B). SCC25 and CAL27 cells were therefore used to generate stable VMP1 knockdown (shVMP1#1) and overexpression (oeVMP1) models, and successful modulation was confirmed by both Western blotting and qPCR (Fig. 6C,D). CCK-8 assays showed that VMP1 knockdown suppressed cell proliferation, whereas VMP1 overexpression enhanced proliferation in both cell lines (Fig. 6E). Consistently, colony formation assays demonstrated reduced clonogenicity upon VMP1 silencing and enhanced clonogenicity following VMP1 overexpression (Fig. 6F,G). EdU incorporation assays further showed decreased EdU-positive fractions after VMP1 knockdown and increased EdU incorporation upon VMP1 overexpression, indicating corresponding changes in DNA synthesis activity (Fig. 6H,I). Collectively, VMP1 promoted proliferative phenotypes in vitro, consistent with the proliferation-related transcriptomic features observed in VMP1-high tumors.

Fig. 6.

In vitro functional validation of VMP1 effects on HNSCC cell proliferation and DNA synthesis. (A) Western blot analysis of VMP1 protein expression in normal human oral keratinocytes (HOK) and multiple HNSCC cell lines, including HSCC2, SCC15, SCC25, UM1, CAL27, HN6, HN30, and FADU. (B) qPCR analysis of relative VMP1 mRNA expression levels in HOK and the indicated HNSCC cell lines. (C) Establishment of stable VMP1 knockdown and overexpression models in SCC25 and CAL27 cells. Western blot analysis shows VMP1 expression in shVMP1#1 versus shNC cells and in oeVMP1 versus oeNC cells. Vinculin was used as a loading control. (D) qPCR validation of VMP1 knockdown and overexpression efficiency in SCC25 and CAL27 cells, comparing shVMP1#1 with shNC and oeVMP1 with oeNC groups. (E) CCK-8 assays assessing cell proliferation in SCC25 and CAL27 cells after VMP1 knockdown or overexpression. Growth curves compare shVMP1#1 versus shNC and oeVMP1 versus oeNC groups. (F) Colony formation assays and quantitative analysis showing the effects of VMP1 knockdown (shVMP1#1 vs shNC) on clonogenic ability in SCC25 and CAL27 cells. (G) Colony formation assays and quantitative analysis showing the effects of VMP1 overexpression (oeVMP1 vs oeNC) on clonogenic ability in SCC25 and CAL27 cells. (H) EdU incorporation assays and quantitative analysis evaluating the effects of VMP1 knockdown (shVMP1#1 vs shNC) on DNA synthesis in SCC25 and CAL27 cells. (I) EdU incorporation assays and quantitative analysis evaluating the effects of VMP1 overexpression (oeVMP1 vs oeNC) on DNA synthesis in SCC25 and CAL27 cells. Scale bar = 100 μm. All experiments were independently repeated three times. For quantitative analyses in panels B and D–I, statistical analyses were performed using a two-tailed Student’s t-test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

VMP1 promotes migratory and invasive phenotypes accompanied by EMT-like changes and NF-κB/p65/PD-L1-related alterations

Based on the above bulk transcriptomic findings, VMP1-high tumors were enriched not only for proliferation-related programs but also for invasion-associated features. Consistently, clinicopathological analyses showed that high VMP1 expression was associated with advanced stage and lymph node metastasis. We therefore next investigated whether VMP1 modulates the migratory and invasive behavior of HNSCC cells in vitro. Given that EMT is closely related to tumor cell motility and invasive capacity, we further examined EMT-related marker changes. In wound-healing assays, VMP1 knockdown (shVMP1#1) markedly impaired wound closure within 24 h in both SCC25 and CAL27 cells, whereas VMP1 overexpression accelerated migration (Fig. 7A,B). Transwell assays further showed that VMP1 knockdown reduced, while VMP1 overexpression increased, both migratory and Matrigel-invasive cell numbers in both cell lines (Fig. 7C,D). Consistently, Western blot analysis of EMT markers showed that VMP1 silencing decreased N-cadherin and increased E-cadherin, whereas VMP1 overexpression produced the opposite pattern (Fig. 7E). In addition, given that VMP1-high status was recurrently associated with enrichment of TNFα/NF-κB signaling across multiple bulk cohorts, we further assessed NF-κB/p65 activity. Based on the bulk and single-cell analyses showing that VMP1-high tumors were accompanied by reduced CD8⁺ T-cell infiltration and impaired effector features, we also examined PD-L1 expression. Western blot analysis showed that VMP1 overexpression increased phosphorylated p65 (p-p65) protein levels, whereas VMP1 knockdown decreased p-p65 expression (Fig. 7F). Similarly, VMP1 overexpression upregulated PD-L1 protein levels, whereas VMP1 knockdown reduced PD-L1 expression (Fig. 7F). To further assess whether NF-κB/p65 is involved in VMP1-associated PD-L1 regulation, we performed p65 knockdown experiments. Western blot analysis confirmed effective silencing of p65 by siRNA in both SCC25 and CAL27 cells (Fig. 7G). Notably, p65 knockdown attenuated VMP1 overexpression-induced p-p65 and PD-L1 upregulation (Fig. 7H). Collectively, VMP1 promoted migration and invasion of HNSCC cells in vitro, accompanied by EMT-like marker switching and NF-κB/p65/PD-L1-related alterations. The attenuation of VMP1-induced p-p65 and PD-L1 upregulation by p65 knockdown supports the potential involvement of NF-κB/p65 signaling in VMP1-associated PD-L1 regulation.

Fig. 7.

VMP1 promotes migration, invasion, EMT-like changes, and NF-κB/p65/PD-L1-related alterations in HNSCC cells in vitro. (A) Wound-healing assays showing the effects of VMP1 knockdown (shVMP1#1 vs shNC) on the migratory capacity of SCC25 and CAL27 cells at 0 h and 24 h. Quantification of wound closure is shown on the right. Scale bar = 400 μm. (B) Wound-healing assays showing the effects of VMP1 overexpression (oeVMP1 vs oeNC) on cell migration in SCC25 and CAL27 cells, with corresponding quantitative analysis. Scale bar = 400 μm. (C) Transwell migration and invasion assays assessing the effects of VMP1 knockdown (shVMP1#1 vs shNC) on SCC25 and CAL27 cells, with quantitative analysis of migrated and invaded cell numbers. Scale bar = 100 μm. (D) Transwell migration and invasion assays evaluating the effects of VMP1 overexpression (oeVMP1 vs oeNC) on SCC25 and CAL27 cells, with quantitative analysis of migrated and invaded cell numbers. Scale bar = 100 μm. (E) Western blot analysis of EMT-related markers (N-cadherin and E-cadherin) following VMP1 knockdown (shVMP1#1 vs shNC) or VMP1 overexpression (oeVMP1 vs oeNC) in SCC25 and CAL27 cells. Vinculin was used as a loading control. (F) Western blot analysis showing changes in VMP1, p65, p-p65, and PD-L1 expression following VMP1 knockdown (shVMP1#1 vs shNC) or VMP1 overexpression (oeVMP1 vs oeNC) in SCC25 and CAL27 cells. Vinculin was used as a loading control. (G) Western blot analysis confirming the knockdown efficiency of p65 using three independent siRNAs in SCC25 and CAL27 cells. Vinculin was used as a loading control. (H) Western blot analysis of p65, p-p65, and PD-L1 expression in SCC25 and CAL27 cells transfected with oeNC, oeVMP1, or oeVMP1 combined with si-p65-mediated p65 knockdown. Vinculin was used as a loading control. All in vitro experiments were independently repeated three times. For quantitative analyses in panels A-D, statistical analyses were performed using a two-tailed Student’s t-test. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Taken together, our findings are consistent with a proposed model in which VMP1 upregulation is associated with a metastasis-related tumor-intrinsic state characterized by enhanced proliferative, migratory, and invasive phenotypes, accompanied by EMT-like changes, increased NF-κB activity, PD-L1 upregulation, and impaired CD8⁺ T-cell function in HNSCC (Fig. 8).

Fig. 8.

Proposed schematic model of VMP1-associated tumor-intrinsic and immune-related changes in HNSCC. VMP1 upregulation in tumor cells is associated with enhanced proliferative, migratory, and invasive phenotypes, accompanied by EMT-like changes characterized by reduced E-cadherin and increased N-cadherin expression. In parallel, VMP1 upregulation is associated with increased NF-κB activity, as reflected by enhanced p65 phosphorylation, together with PD-L1 upregulation. These changes may be associated with reduced CD8⁺ T-cell infiltration, impaired CD8⁺ T-cell effector function, and myeloid remodeling within the tumor immune microenvironment, thereby potentially linking VMP1-high tumor states to metastatic progression in HNSCC. Solid lines indicate relationships supported by experimental data in this study, whereas dashed lines indicate putative associations or inferred effects.

Discussion

This study integrated multi-cohort single-cell transcriptomics, bulk transcriptomic analyses, clinical specimens, and in vitro functional assays to characterize metastasis-associated immune microenvironment changes in head and neck squamous cell carcinoma (HNSCC) [12,30,31]. We identified VMP1 as a tumor cell–intrinsic molecule that is consistently upregulated across distinct metastatic contexts. Elevated VMP1 expression was significantly associated with poor outcomes and aggressive tumor progression. It was also accompanied by an immunosuppressive microenvironment characterized by impaired CD8⁺ T-cell effector function. These findings suggest that VMP1 may be associated with both tumor invasion and immune-suppressive microenvironmental features.

Previous studies have established that tumor metastasis is frequently accompanied by profound immune remodeling, typically characterized by reduced CD8⁺ T-cell activity and enrichment of immunosuppressive myeloid populations [32,33]. Such an immunosuppressive state is considered a major biological basis for tumor immune escape and limited responses to immunotherapy [[34], [35], [36]]. At single-cell resolution, our data further supported these features in metastatic HNSCC, where impaired CD8⁺ T-cell function and accumulation of immunosuppressive myeloid cells were commonly observed. This remodeling was more evident in distant metastatic lesions, whereas lymph node metastases showed relatively preserved immune effector activity. These findings indicate that although immune heterogeneity exists across metastatic sites, metastasis-associated immunosuppression may involve shared molecular determinants.

Despite extensive characterization of immune phenotypes, most studies have primarily focused on immune cell composition or functional states. In contrast, the tumor cell–intrinsic molecular factors associated with immune remodeling remain poorly understood. In particular, whether a set of tumor-intrinsic factors that are stably upregulated across diverse metastatic environments, independent of specific immune contexts, remains unclear. By intersecting two independent single-cell cohorts, we identified VMP1 as a consistently upregulated candidate, thereby providing a molecular entry point for understanding metastasis-associated immune remodeling.

VMP1 has been recognized as a regulator of autophagosome formation and cellular stress adaptation and has been linked to enhanced tumor cell survival, invasion, and therapeutic resistance in multiple malignancies [[37], [38], [39], [40]]. However, its potential role in shaping the tumor immune microenvironment has been largely overlooked. Here, our data suggest that VMP1 is consistently upregulated across metastatic settings in HNSCC and is associated with both invasive phenotypes and immunosuppressive features. Multi-cohort bulk transcriptomic analyses revealed that VMP1-high tumors were enriched for invasion programs, including EMT and TNFα/NF-κB signaling. By contrast, antitumor immune pathways were broadly suppressed. Consistently, in vitro functional assays showed that VMP1 promotes tumor cell proliferation, migration, and invasion. These functional effects were accompanied by EMT-associated molecular alterations, supporting a role for VMP1 in malignant progression.

The NF-κB pathway is widely regarded as a central inflammatory hub involved in both tumor invasion and immune regulation [41,42]. Sustained NF-κB activation not only facilitates EMT and tumor cell motility but also suppresses CD8⁺ T cell-mediated antitumor immunity through transcriptional upregulation of immune checkpoint molecules such as PD-L1 [[43], [44], [45]]. In our study, VMP1 upregulation was accompanied by increased p65 phosphorylation and elevated PD-L1 expression, whereas VMP1 knockdown produced the opposite pattern. Moreover, p65 knockdown attenuated VMP1 overexpression-induced PD-L1 upregulation. These findings support a potential VMP1-associated NF-κB/p65–PD-L1 regulatory link. Nevertheless, the current data do not fully define the direct molecular mechanism by which VMP1 influences NF-κB activation. Therefore, this relationship should be interpreted as a functionally supported but not yet fully established regulatory pathway.

Integrating prior evidence of VMP1 involvement in cellular stress and inflammatory responses with our multi-omics and functional observations, VMP1 upregulation may reflect a pro-invasive tumor-intrinsic state. This state is accompanied by EMT activation, enhanced NF-κB/p65 activity, and increased PD-L1 expression. This model may partially explain the association between VMP1-high tumors and immune-suppressive microenvironmental features. However, it requires further validation in more mechanistically defined experimental systems.

Several limitations should be acknowledged. First, although our in vitro data support the involvement of NF-κB/p65 signaling in VMP1-associated PD-L1 regulation, the upstream molecular events linking VMP1 to NF-κB activation remain to be clarified. Future studies incorporating rescue experiments, promoter reporter assays, chromatin immunoprecipitation, and analysis of stress- or autophagy-related pathways would help define this regulatory mechanism more precisely. Second, the present study lacks in vivo functional validation of VMP1, which limits direct assessment of its role in metastatic progression and immune microenvironment remodeling under physiological tumor conditions. Third, the immune-related findings were mainly derived from bulk and single-cell transcriptomic analyses, together with correlative clinical observations. Additional validation using spatial profiling, flow cytometry, or in vivo immune profiling would strengthen these conclusions. Fourth, p16 status was recorded when available in the institutional cohort, but p16 information was not available for all cases, and HPV DNA/RNA testing was not routinely performed. Detailed HPV-related status and treatment background were also not uniformly available across all public datasets. Finally, the potential biomarker relevance of VMP1, including its relationship with immunotherapy response, requires further evaluation in larger, well-annotated prospective cohorts and immunotherapy-related datasets.

In summary, our study identifies VMP1 as a metastasis-associated tumor cell–intrinsic factor linked to invasive progression and immune-suppressive microenvironmental features in HNSCC. These findings expand the current understanding of VMP1 in tumor biology and provide a basis for further mechanistic validation and clinical investigation of its biomarker relevance.

Conclusion

In summary, by integrating multi-cohort single-cell transcriptomics with bulk transcriptomic analyses, clinical validation, and in vitro functional assays, we systematically delineated immune microenvironment remodeling during HNSCC metastasis and identified VMP1 as a metastasis-associated, tumor cell–intrinsic factor consistently upregulated across distinct metastatic contexts. VMP1-high tumors were characterized by impaired CD8⁺ T-cell effector states accompanied by myeloid remodeling, and elevated VMP1 expression was associated with unfavorable clinical outcomes and aggressive clinicopathological features. Functionally, VMP1 promoted HNSCC cell proliferation, migration, and invasion, in parallel with EMT marker switching as well as increased NF-κB activity and PD-L1 expression. Collectively, these findings nominate VMP1 as a potential molecular node associated with malignant progression and immune modulation in metastatic HNSCC, providing a rationale for further mechanistic validation and clinical investigation of its biomarker relevance in HNSCC.

Ethics approval and informed consent

This study was approved by the Ethics Committee of Tianjin Medical University Cancer Institute & Hospital (Approval No bc2025060). All procedures were performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients before inclusion in the study.

Consent for publication

Not applicable. This study does not contain any identifiable individual person’s data, images, or videos.

Data availability statement

Publicly available datasets were analyzed in this study. Transcriptomic and clinical data of head and neck squamous cell carcinoma (HNSCC) were obtained from The Cancer Genome Atlas (TCGA-HNSCC). Single-cell RNA-seq datasets GSE243933 and GSE181919, as well as validation cohorts GSE65858 and GSE117973, were downloaded from the Gene Expression Omnibus (GEO) database. All processed data and materials generated during this study are available from the corresponding author upon reasonable request.

CRediT authorship contribution statement

Xinhua Li: Writing – original draft, Software, Formal analysis, Data curation, Conceptualization. Ailin Yang: Validation, Investigation. Leidong Yue: Investigation. Chao Jing: Writing – review & editing, Supervision, Funding acquisition. Xudong Wang: Writing – review & editing, Supervision, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.