An immune-related miRNA signature predicts the prognosis in nasopharyngeal carcinoma

Abstract

Background

MicroRNAs (miRNAs) play a significant role in the progression of nasopharyngeal carcinoma (NPC), particularly in relation to immune responses. This study aims to identify immune-related miRNAs that contribute to the development of a prognostic signature for overall survival (OS) in NPC patients.

Methods

Differentially expressed miRNAs between NPC and normal nasopharyngeal tissues were identified using GSE70970, GSE32960, and The Cancer Genome Atlas (TCGA)-NPC datasets. Univariate Cox regression analysis was performed to find miRNAs associated with OS. The Least Absolute Shrinkage and Selection Operator algorithm was conducted on the GSE70970 cohort to create a risk signature, stratifying patients into high- and low-risk groups based on the median risk score.

Results

The immune-related miRNA signature included six specific miRNAs: hsa-miR-523, hsa-miR-130a, hsa-miR-342-3p, hsa-miR-320b, hsa-miR-1181, and hsa-miR-150. Patients classified into the high-risk score group exhibited poorer prognoses. The prognostic signature demonstrated its significance as an independent predictor of OS. Furthermore, high-risk patients presented distinct gene mutation statuses and increased stemness scores.

Conclusions

The developed immune-related miRNA prognostic signature provided an accurate prediction of NPC prognosis. Additionally, the accompanying nomogram offered a practical tool for estimating patient OS, potentially aiding clinical decision-making.

Highlight box.

Key findings

• This study identified a novel immune-related miRNA prognostic signature comprising six miRNAs that robustly predicts overall survival in nasopharyngeal carcinoma (NPC).

What is known and what is new?

• miRNAs play a crucial role in NPC progression and immune regulation.

• This study introduced a clinically applicable nomogram combining the miRNA signature with standard prognostic factors for personalized survival estimation.

What is the implication, and what should change now?

• The miRNA signature offers a molecular tool to refine NPC risk stratification beyond conventional clinical staging.

Introduction

Nasopharyngeal carcinoma (NPC) is an epithelial malignancy arising from the nasopharyngeal mucosa, predominantly found in Southeast Asia (1-3). Treatment failure often results from local metastasis, with over 70% of patients experiencing local progression (1). Accurate prognosis prediction is essential for guiding treatment regimens. The tumor-node-metastasis (TNM) system is recognized as the gold standard for prognosis prediction (4). However, this system primarily focuses on anatomical characteristics while neglecting biological heterogeneity, leading to inaccurate prognosis estimates. Thus, there is a pressing demand for novel biomarkers to improve prognosis prediction alongside the TNM system.

MicroRNAs (miRNAs) are crucial in cancer development and treatment response (5-7). Certain miRNAs, such as miR-195, miR-29b, and miR-29c, can inhibit or promote various cancers and serve as potential biomarkers for diagnosis and prognosis (8-10). Studies have confirmed the predictive value of specific miRNAs for NPC prognosis (11-13), although the underlying molecular mechanisms remain poorly understood.

The tumor microenvironment (TME), consisting of tumor cells, stromal cells, and immune cells, plays a vital role in tumorigenesis and progression (14,15). Research indicates a connection between TME and miRNA dysregulation (16-18), but the relationship between TME-related miRNAs and prognosis in NPC patients is still unclear.

The least absolute shrinkage and selection operator (LASSO) has become a widely used method for constructing prognostic models (19,20). Drawing inspiration from this, we developed a risk profile based on immune-related miRNAs to assess overall survival (OS) in NPC patients. Using the LASSO Cox regression algorithm, we selected features and quantified Cox regression coefficients, resulting in a scoring model that accurately predicts NPC prognosis. Additionally, we provide a nomogram to assist clinicians in prognostic assessment for NPC patients. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-393/rc).

Methods

Data collection and preprocessing

Large microRNA microarray and sequencing cohorts of NPC patients were sourced from the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) and The Cancer Genome Atlas (TCGA; https://cancergenome.nih.gov/). Additionally, microRNA data for head and neck squamous cell carcinomas (HNSCs) were obtained from UCSC XENA (https://xena.ucsc.edu/). We specifically separated 90 pharynx cancer patients from the TCGA-HNSC cohort to create a new cohort named TCGA-NPC (Table S1), which was used for screening immune-related miRNAs. The GSE70970 dataset was selected for constructing the miRNA signature. Detailed information about all cohorts utilized is provided in Table 1. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Table 1. Baseline information for the selected cohorts.

Series accession numbers	Number of NPC tissues	Number of normal tissues	Platform
GSE70970	246	17	nCounter® Human miRNA Assay (v1.0, Nanostring)
GSE32960	312	18	MicroRNA array
TCGA-HNSC	512	–	Illumina miRNAseq
TCGA-NPC	90	–	Illumina miRNAseq

HNSC, head and neck squamous cell carcinoma; NPC, nasopharyngeal carcinoma; TCGA, The Cancer Genome Atlas.

Establishment and validation of the miRNAs signature

GSE70970, GSE32960 and TCGA-NPC were utilized to identify differentially expressed miRNAs (DEMs) between NPC tissues and normal nasopharyngeal tissues. Dysregulated DEMs were selected using a false discovery rate (FDR) <0.01 and |log₂ fold change (FC)| >1.2 as criteria. Univariate Cox regression analysis was conducted to identify miRNAs significantly related to OS of NPC patients. Subsequently, LASSO analysis was performed on the GSE70970 cohort, employing tenfold cross-validation to prevent overfitting. Using the resulting risk signature, NPC patients from the GSE70970 dataset were stratified into high- and low-risk groups based on the median risk score. The prognostic significance of the signature was evaluated using Kaplan-Meier survival analysis, while its predictive accuracy was determined via time-dependent receiver operating characteristic (ROC) curves. Finally, the TCGA-NPC dataset served as an independent validation cohort to confirm the robustness of the signature.

Functional annotation of the miRNA signature

To assess the functions of the miRNA signature, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses for the selected miRNAs were conducted using miRPath (https://dianalab.e-ce.uth.gr/html/mirpathv3/index.php?r=mirpath).

Evaluation of immune-infiltrating cells in NPC

To evaluate immune-infiltrating cells in NPC, the abundances of 22 immune cell types for each patient in TCGA-NPC were estimated using CIBERSORTx (https://cibersortx.stanford.edu/). We then assessed the correlations between the prognostic signature and the abundances of these immune-infiltrating cells.

Assessing tumor heterogeneity in HNSC

Gene mutation status and tumor mutation burden (TMB) in NPC patients were assessed using data from the TCGA database, while stemness scores were calculated using DNA methylation signature probes in the TCGA-NPC cohort. The patients with NPC were divided into high-risk and low-risk groups based on the median risk score. We compared the proportions of gene mutations between these two groups and calculated the correlations between TMB and the signature, as well as between stemness scores and the signature.

Statistical analysis

All statistical analyses were conducted using R software (version 4.4.2). A P value of less than 0.05 was considered statistically significant.

Results

Identification of immune-related DEMs

Microarray data from GSE70970 (Figure 1A) and GSE32960 (Figure 1B) were utilized to identify DEMs. Using a cut-off criterion of FDR <0.05 and |log₂ FC| >1.2, 26 DEMs were deemed statistically significant between NPC tissues and normal tissues in both datasets. TCGA-NPC was further analyzed to identify DEMs (Figure 1C), revealing 11 overlapping miRNAs for further study (Figure 1D).

Figure 1.

Screening and analysis of differentially expressed miRNAs in NPC. Upregulated and downregulated miRNAs identified in (A) GSE70970, (B) GSE32960, and (C) TCGA-NPC. (D) Venn diagram illustrating common differentially expressed miRNAs across the three datasets. NPC, nasopharyngeal carcinoma; TCGA, The Cancer Genome Atlas.

Feature selection and prognostic signature construction

To identify miRNAs associated with OS, we conducted univariate Cox regression analysis on the DEMs, selecting six microRNAs (hsa-miR-523, hsa-miR-130a, hsa-miR-342-3p, hsa-miR-320b, hsa-miR-1181, and hsa-miR-150) as candidate features (Figure 2). We then employed the LASSO Cox regression algorithm to construct a prognostic signature (Figure 3A,3B). The scoring model was calculated as follows: risk score =1.909 * hsa-miR-1181 + 0.148 * hsa-miR-130a − 0.106 * hsa-miR-150 + 0.121 * hsa-miR-320b - 0.066 * hsa-miR-342-3p + 0.195 * hsa-miR-523, where each miRNA symbol represents its expression level.

Figure 2.

Forest plots presenting the results of Cox univariate regression analysis for six prognostic miRNAs. CI, confidence interval.

Figure 3.

Development of a prognostic signature using the LASSO Cox regression algorithm in GSE70970. (A,B) Cvfit and lambda curves demonstrate the LASSO regression performed with minimum criteria. (C) Receiver operating characteristic curve indicating the prognostic signature’s ability to predict 1-, 3-, and 5-year overall survival. (D) Kaplan-Meier curves depicting survival status and duration. (E) Analysis of the relationship between risk score and survival status in NPC patients. LASSO, least absolute shrinkage and selection operator; NPC, nasopharyngeal carcinoma.

NPC patients in the GSE70970 dataset were stratified into high- and low-risk groups based on the median risk score. The six-miRNA signature exhibited robust predictive performance for OS, with AUCs of 0.67, 0.67, and 0.69 for 1-, 3-, and 5-year OS, respectively (Figure 3C). Furthermore, the signature was significantly associated with OS, yielding a hazard ratio (HR) of 2.57 [95% confidence interval (CI): 1.54–4.28; Figure 3D]. A heatmap visualization revealed that expression levels of the six miRNAs followed a distinct trend corresponding to increasing risk scores (Figure 3E). External validation using the TCGA-NPC dataset confirmed the prognostic reliability of the signature, with AUCs of 0.61, 0.69, and 0.77 for 1-, 3-, and 5-year OS, respectively (Figure S1). These findings collectively support the clinical utility of the six-miRNA signature in predicting NPC prognosis.

Functional annotation of the miRNAs signature

To identify the biological functions of the miRNA signature, we conducted GO and KEGG enrichment analyses. The biological processes associated with the miRNAs were primarily enriched in immune system processes, epidermal growth factor receptor (EGFR) signaling, cell death, energy reserve metabolism, and nucleoplasm (Figure 4A, Table S2). The cellular components of the associated mRNAs focused on catabolic processes, enzyme binding, cellular protein metabolism, component assembly, and transcription factor activity (Figure 4B, Table S3). In terms of molecular functions, mRNAs were significantly enriched in protein complexes, gene expression, neurotrophin TRK receptor signaling, nucleic acid binding transcription factor activity, and viral processes (Figure 4C, Table S4). The five most significantly enriched KEGG pathways included axon guidance, gap junctions, signaling pathways regulating pluripotency of stem cells, prolactin signaling, and leukocyte transendothelial migration (Figure 4D, Table S5).

Figure 4.

Biological functional and pathway enrichment analysis of the prognostic signature. Enrichment in (A) biological processes, (B) cellular components, (C) molecular functions, and (D) KEGG pathways for the prognostic miRNAs. (E) Relationship between risk score and immune infiltration in NPC tissues. -, P>0.05; *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. KEGG, Kyoto Encyclopedia of Genes and Genomes; NPC, nasopharyngeal carcinoma.

To further elucidate the underlying mechanisms of the risk score, we examined its relationship with immune cell infiltration in TCGA-NPC (Figure 4E). Immune infiltrates were estimated using CIBERSORTx, revealing a positive correlation between the risk score and the abundance of M1 macrophages and resting CD4⁺ T cells.

The relationship between miRNAs signature and tumor heterogeneity in NPC

To further investigate the functions of the miRNA signature in NPC, we compared tumor mutation statuses between the high-risk and low-risk groups. The high-risk group showed a higher proportion of low-density lipoprotein receptor-related protein 1B (LRP1B) mutations, a biomarker for predicting immunotherapy efficacy (21-23), while the proportion of tumor protein p53 (TP53) mutations was lower (Figure 5A). Additionally, analysis of stemness scores and tumor mutational burden (TMB) in TCGA-HNSC revealed a significant negative correlation between risk scores and stemness scores (Figure 5B), but no correlation with TMB (Figure 5C).

Figure 5.

Correlation between the miRNA signature and tumor heterogeneity in HNSC. (A) Comparison of tumor mutation statuses between high-risk and low-risk groups, with higher mutations in the high-risk cohort. Relationships between (B) stemness scores, (C) TMB, and risk score are explored. HNSC, head and neck squamous cell carcinoma; TMB, tumor mutation burden.

Construction of the nomogram

Univariate and multivariate Cox regression analyses were conducted on GSE70970 to evaluate prognostic factors. The risk score (HR =2.58, 95% CI: 1.81–3.67), age (HR =1.04, 95% CI: 1.02–1.06), N stage (HR =1.71, 95% CI: 1.26–2.32), and chemotherapy (HR =0.54, 95% CI: 0.30–0.97) independently predicted OS in NPC patients (Figure 6A,6B). Based on these findings, a nomogram was developed to estimate 1-, 3-, and 5-year OS for each NPC patient (Figure 6C). The calibration curve showed excellent consistency between nomogram predictions and actual observations (Figure 6D).

Figure 6.

Establishment of a nomogram. Results from (A) univariate and (B) multivariate Cox regression analysis. (C) Nomogram predicting 1-, 3-, and 5-year overall survival rates for NPC patients. (D) Calibration curve assessing the accuracy of the nomogram model. CI, confidence interval; NPC, nasopharyngeal carcinoma.

Discussion

miRNAs play a critical role in tumor development, progression, and therapy response (24-28). Recent research has highlighted their significant effects on NPC (29,30). Advancements in understanding the pathogenesis, development, and prognosis of miRNAs in NPC have been rapid (25,29). A meta-analysis revealed 65 miRNAs as potential prognostic markers for NPC, offering new directions for future research (31).

In this study, an immune miRNA signature was developed utilizing LASSO Cox regression. This signature exhibited significant prognostic value, supported by a HR of 2.57. Furthermore, the signature demonstrated moderate predictive accuracy for OS, as indicated by time-dependent ROC analysis, which produced AUCs ranging from 0.67 to 0.69 for 1- to 5-year survival rates. External validation in an independent cohort confirmed these results, with AUCs spanning from 0.61 to 0.77, thereby reinforcing the model’s generalizability. However, several limitations merit consideration. First, the reliance on microarray-based expression profiling may not adequately capture the dynamic regulation of miRNAs, potentially introducing technical biases. Second, the signature’s moderate predictive performance raises concerns regarding its clinical utility, particularly in heterogeneous patient populations. Although the integration of the risk score into a clinically applicable nomogram exhibited favorable calibration, prospective validation in larger, multi-institutional cohorts is essential to establish its robustness and translational relevance.

In the immune system, miRNAs are key regulators of gene expression (32,33). Investigating miRNAs within the immunomodulatory network is crucial for elucidating molecular mechanisms and improving clinical management of NPC (34-36). In our study, immune scores were calculated for each patient in the TCGA-NPC cohort, leading to the identification of immune-related miRNAs that impact patient prognosis. Functional analysis revealed that higher risk scores were associated with the activation of immune processes and EGFR signaling pathways. This suggests a potential treatment strategy for high-risk NPC patients, who may benefit from immune checkpoint inhibitors or EGFR-targeted therapies. Further research is needed to validate this hypothesis and will be a focus of our future work.

Gene mutations critically influence treatment outcomes in cancer (37-39). For example, alterations in tumor suppressor genes like TP53 are linked to chemotherapy resistance and aggressive progression (40), necessitating alternative therapeutic strategies. However, tumor heterogeneity complicates universal application. Comprehensive genomic profiling thus remains vital to tailor treatments, improve survival, and overcome resistance in NPC. Our analysis revealed distinct mutation profiles between high-risk and low-risk NPC groups stratified by the miRNA signature. Notably, the high-risk group exhibited a higher prevalence of LRP1B mutations, a genomic alteration increasingly recognized as a biomarker for predicting immunotherapy efficacy (21-23). This finding suggests that the miRNA signature may indirectly reflect an immunosuppressive TME or altered DNA repair mechanisms, both of which could influence sensitivity to immune checkpoint inhibitors. Conversely, the lower proportion of TP53 mutations in the high-risk group is intriguing, as TP53 loss typically correlates with aggressive tumor behavior. This paradox may imply that miRNA-driven pathways in high-risk NPC bypass classical TP53-mediated tumor suppression, potentially through dysregulation of alternative oncogenic or anti-apoptotic pathways.

Conclusions

In conclusion, we developed an immune miRNA signature utilizing multiple datasets, which accurately predicts the prognosis and progression of NPC, while also proposing a potential treatment strategy based on risk scores.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.