A total of 20 patients with differentiated squamous cell NPC were recruited prior to radiation therapy at the Institute of NPC, Affiliated with the People's Hospital of Guangxi Zhuang Autonomous Region (Nanning, China) between June 2003 and July 2004. Inclusion criteria were as follows: i) Histopathologically confirmed non-keratinizing squamous cell PC (WHO Type II); ii) age ≥18 years at diagnosis; iii) no prior history of radiotherapy, chemotherapy or surgical intervention for NPC; and iv) signed informed consent for tissue sample collection and follow-up procedures. Exclusion criteria were as follows: i) Presence of distant metastasis at initial diagnosis; ii) coexistence of other primary or metastatic malignancies; iii) severe comorbidities that may compromise treatment compliance or follow-up; iv) active autoimmune diseases requiring systemic immunosuppressive therapy; and v) loss to follow-up (incomplete clinical data or inability to assess treatment outcomes). These patients included 15 men (75%) and 5 women (25%), with a median age of 41.5 years (35.8, 51.0). Pre-treatment primary nasopharyngeal tumor tissues were obtained from patients with NPC through diagnostic biopsy and immediately stored at −80°C until required for analysis. Written informed consent was obtained from all patients prior to biopsy procedures. The present study was approved by the Institutional Ethics Committee of the People's Hospital of Guangxi Zhuang Autonomous Region (approval no. KY-KJT-2024-43).

Patients were categorized into ‘favorable’ and ‘unfavorable’ prognosis groups, based on survival times (>3 or <3 years). The relevant patient characteristics are listed in Table I.

Total RNA was isolated from the tissue samples of 20 patients with NPC using a TRIzol^™ reagent (Invitrogen^™; Thermo Fisher Scientific, Inc.), as per the manufacturer's guidelines. The concentration and purity of each RNA sample were assessed using agarose gel electrophoresis, whilst RNA integrity was ensured by measuring the 260/280 nm absorbance ratio. Subsequently, total RNA was reverse-transcribed into cDNA using Oligo (dT) primers and the Superscript^™ III RNase H-Reverse Transcriptase Kit (Invitrogen; Thermo Fisher Scientific, Inc.), according to the manufacturer's instructions. The synthesized cDNA was stored at −20°C for future use.

The construction of the microarray using the tissues from 20 patients with NPC and probe preparation were performed as previously described (17).

The probe was dissolved in 20 µl hybridization buffer (5X SSC, 0.75 M NaCl, 0.075 M sodium citrate, 0.4% SDS and 50% formamide). The microarrays were pre-treated in a hybridization buffer supplemented with 0.5 mg/ml denatured salmon sperm DNA at 42°C for 6 h. Subsequently, the fluorescent probe mixtures were denatured following heating at 95°C for 5 min and were then applied to the pre-hybridized microarray under a cover slip. Hybridization was performed at 42°C for 15–17 h. Post-hybridization, the chips were washed at 60°C for 10 min in the following sequential solutions: 2X SSC with 0.2% SDS, followed by 0.1X SSC with 0.2% SDS and then in 0.1X SSC. Finally, the chips were air-dried at room temperature.

Tissue specimens from 20 NPC patients with varying prognoses were used to construct cDNA microarrays (Biostar H-141s) (Shanghai Boxing Gene Chip, Co., Ltd.) following Brown's protocol, encompassing 9,646 genes. After normalizing ≥14 microarray datasets, detectable signals enabled evaluation of gene expression profiles across NPC patients with divergent prognoses. Differential gene expression analysis was performed using the limma package (version 3.54.1) (https://bioconductor.org/packages/limma/) in R software (version 4.2.1) (https://www.r-project.org/). To capture the emissions from Cy5 and Cy3, the microarrays were scanned using the ScanArray 4000 (GSI Lumonics, Inc.) at wavelengths of 635 and 532 nm, respectively. GenePix Pro 3.0 software (Molecular Devices, LLC) was utilized to process the captured images. The intensities for each spot at these two wavelengths corresponded to the quantities of Cy3-dUTP and Cy5-dUTP. The Cy5 to Cy3 ratios were calculated using the GenePix Pro 3.0 median ratio method. Data normalization was performed utilizing the default normalization factor provided by GenePix. Spots flagged as ‘bad’ or ‘not found’ by the software were excluded from the analysis. Only genes with raw intensity values of >200 counts for both Cy3 and Cy5 on each array were further analyzed.

To assess the expression levels of stomatin like (STOML)1, ras-related protein 25 (RAB25), brain acid soluble protein 1 (BASP1), RAD50-interacting protein 1 (RINT1) and U2 snRNP associated SURP domain containing (U2SURP), a tissue microarray containing 107 primary NPC samples was purchased from Shanghai Outdo Biotech Co. Ltd. (cat. no. HNasN110Su01). Out of the 107 patients, there were 77 men (71.96%) and 30 women (28.04%), with a median age was 47.0 years (41.0–56.0). Ethical approval was obtained from the Ethics Committee of Shanghai Outdo Biotech Company (Shanghai, China; approval no. YBM-05-01).

Immunohistochemistry was performed on the 107 paraffin-embedded NPC tissue sections (4-µm thickness). Tissues were fixed in 10% neutral buffered formalin at room temperature for 24 h prior to paraffin embedding, as previously described (17). Briefly, antigen retrieval was performed by microwaving sections in 10 mM citrate buffer (pH 6.0) at 95–100°C for 5 min. Prior to this step, tissue sections underwent deparaffinization and rehydration through a descending ethanol series (100% down to 70%), followed by endogenous peroxidase blocking. Endogenous peroxidase activity was blocked using 3% hydrogen peroxide in methanol for 10 min at room temperature. To prevent non-specific binding, slides were incubated with 10% FBS/PBS for 30 min at room temperature. The sections were then incubated with antibodies against STOML1 (1:100; cat. no. HPA042353; Sigma-Aldrich; Merck KGaA), RAB25 (1:100; cat. no. Ml086507; Shanghai Enzyme-linked Biotechnology Co., Ltd.), BASP1 (1:200; cat. no. HPA045218; Sigma-Aldrich; Merck KGaA), RINT1 (1:100; cat. no. HPA019875; Sigma-Aldrich; Merck KGaA) and U2SURP (1:500; cat. no. HPA037546; Sigma-Aldrich; Merck KGaA) at 4°C overnight. After incubation, the sections were washed three times with PBS (pH 7.4) and incubated in HRP-conjugated anti-rabbit (cat. no. A0239; Beyotime Institute of Biotechnology) secondary antibodies for 1 h at room temperature, followed by staining with DAB according to the GTVision^™ II Detection System/Mo&Rb Kit (cat. no. GK500705; Shanghai GenTech Co., Ltd.) instructions. After DAB staining, sections were counterstained with Mayer's Hematoxylin Solution (cat. no. G1004; Wuhan Servicebio Technology Co., Ltd.) for 30 sec at room temperature, followed by differentiation in 1% hydrochloric acid-alcohol for 5 sec and bluing in saturated lithium carbonate solution for 30 sec. The sections were then dehydrated through an ascending ethanol series (70, 95 and 100%) and cleared in xylene, the sections were mounted and visualized under a light microscope (Leica DM2500; Leica Microsystems, Inc.).

Gene expression comparisons between head and neck squamous cell carcinoma (HNSC) and normal tissues were performed using mRNA data from the TCGA-HNSC project (https://portal.gdc.cancer.gov/projects/TCGA-HNSC), Survival prognosis was evaluated through Kaplan-Meier analysis using the Kaplan-Meier Plotter database (https://kmplot.com/analysis). Additionally, the Tumor Immune Estimation Resource (TIMER) 2.0 database (http://timer.cistrome.org) was employed to predict the correlations between gene mRNA expression levels.

GO and KEGG pathway analyses were performed using the clusterProfiler package (version 4.4.4) (https://bioconductor.org/packages/clusterProfiler/) in R software (version 4.2.1) (https://www.r-project.org/) with data from the microarray of tissues from 20 patients with NPC. ID conversion for the molecular input lists was facilitated using the org.Hs.eg.db package from the ID conversion library (https://bioconductor.org/packages/org.Hs.eg.db/). The enrichment results were visualized with ggplot2 (version 3.3.6) (https://cran.r-project.org/package=ggplot2), along with igraph (version 1.4.1) (https://cran.r-project.org/package=igraph) and ggraph (version 2.1.0) for enhanced graphical representation (https://cran.r-project.org/package=ggraph).

ROC curve analysis, performed using data from the microarray of 107 primary NPC samples in GraphPad Prism 9 software (Dotmatics), was employed to evaluate the predictive performance of different clinical variables, such as TNM stage and protein expression levels, in NPC prognosis.

Statistical analyses were performed using SPSS software (v22.0; IBM Corps.) and GraphPad Prism v.9.0. (Dotmatics). Data are expressed as the mean ± standard deviation. Unpaired student's t-test was applied to evaluate significant differences between two groups. Kaplan-Meier survival curves were plotted to assess the association between gene expression and overall survival (OS) and disease-free survival (DFS) in patients with NPC. Prognostic cut-off values for BASP1 and STOML1 expression were determined using X-tile software v3.6.1 (http://www.tissuearray.org/rimmlab/Xtile.htm). The significance was then determined using the log-rank test. Cox regression analyses were performed to identify prognostic factors that could affect OS. P<0.05 was considered to indicate a statistically significant difference.

The study cohort included 12 patients with favorable outcomes and eight with unfavorable ones. Differentially expressed genes were identified using the following criteria: |Log₂ fold change|>1.5, P<0.05 and false discovery rate <0.01. Data analysis was performed using the Limma package in R. A total of six genes were identified to be associated with NPC prognosis. Among them, epidermal growth factor receptor (EGFR), RINT1, BASP1, U2SURP and RAB25 were upregulated in patients with unfavorable outcomes, whilst STOML1 was downregulated in this group (Fig. 1). As EGFR is a well-characterized prognostic marker in several types of cancer, it was excluded from further analysis (18).

The TCGA database was used to assess the differential expression of five genes in HNSC tissues compared with normal ones, and the results revealed that the mRNA expression levels of RINT1, BASP1, U2SURP and STOML1 were increased in HNSC tissues, whereas RAB25 expression was decreased (Fig. 2). Additionally, the Kaplan-Meier plotter online database was utilized to assess the associations between the mRNA expression levels of the aforementioned five genes in HNSC tumor tissues and OS. The results revealed that patients with high BASP1 mRNA expression levels had significantly shorter OS rates compared with those with low BASP1 expression. By contrast, elevated STOML1 mRNA expression was significantly associated with markedly longer OS compared with low STOML1 expression (P<0.05; Fig. 3).

To further assess the reliability of DNA microarray technology and database-predicted results, an immunohistochemistry analysis was performed on tissue samples from 107 patients with NPC (Fig. S1, Fig. S2, Fig. S3, Fig. S4, Fig. S5). Due to their aforementioned OS results, the analysis focused on assessing the expression levels of STOML1 and BASP1. Additionally, patients were categorized into high and low expression groups for both proteins using the X-tile tool (19), which provided optimal cut-off values. Kaplan-Meier survival curves were utilized to evaluate patient survival, whilst the log-rank test was applied to assess the effect of BASP1 and STOML1 expression on OS and DFS in patients with NPC. The results demonstrated that high STOML1 expression was significantly associated with both improved OS and DFS rates compared with low STOML1 expression (P<0.05). By contrast, high BASP1 expression was significantly associated with worse DFS rates compared with low BASP1 expression, but not OS (P<0.05; Fig. 4). Furthermore, Cox univariate and multivariate analyses were performed to assess the association between protein expression, clinical characteristics, OS and DFS. The analyses revealed that STOML1 expression was an independent prognostic factor for both OS (Table II) and DFS (Table III) in patients with NPC. These preliminary findings suggest that STOML1 could serve a crucial role in the development and progression of NPC.

ROC curve analysis was performed to assess the predictive performance of different models in 5-year OS in 107 patients with NPC. More specifically, the predictive significance of TNM staging, the current gold standard for NPC prognosis (20), the protein expression levels of STOML1 and that of a combined model incorporating both factors were evaluated. The results revealed an AUC value of 0.774 [95% confidence interval (CI), 0.580–0.968; P=0.010] for STOML1, 0.715 (95% CI, 0.556–0.875; P=0.043) for TNM staging and 0.874 (95% CI, 0.766–0.982; P<0.001) for the combined model (Fig. 5). These findings indicate that the protein expression levels of STOML1, whether analyzed independently or in combination with TNM staging, demonstrate improved prognostic accuracy compared with TNM staging alone for predicting the outcomes of patients with NPC.

GO and KEGG pathway enrichment analyses were performed on the differentially expressed genes identified through the DNA microarray analysis of tissues from 20 patients with NPC to assess the potential molecular mechanisms underlying the effect of STOML1 on NPC prognosis. GO functional enrichment analysis revealed that NPC prognosis was associated with several biological processes, including ‘inactivation of epidermal growth factor’, ‘K63-linked polyubiquitination-dependent protein binding’ and ‘negative regulation of the TLR4 signaling pathway’. In addition, KEGG pathway enrichment analysis demonstrated that NPC prognosis may be closely associated with several pathways, such as the ‘PI3K-Akt signaling pathway’, ‘EBV infection’ and ‘microRNAs in cancer’ (Fig. 6). Furthermore, an analysis using the TIMER 2.0 database revealed a significant positive correlation between STOML1 and the mRNA expression levels of BAX, FAS and caspase-9, which are key genes involved in the apoptosis pathway (21) (Fig. 7). The aforementioned results suggest that STOML1 could affect the apoptosis-related signaling pathways, thus potentially suppressing NPC progression.