Development of a quantitative genomic instability scoring system and a related competing endogenous RNA network in head and neck squamous cell carcinoma

Abstract

Background

Genomic instability (GI) is a hallmark of cancer and plays a crucial role in the progression of head and neck squamous cell carcinoma (HNSCC). This study aimed to quantitatively characterize GI features and construct a GI-related competing endogenous RNA (ceRNA) network in HNSCC.

Methods

Weighted gene co-expression network analysis (WGCNA) and differential gene expression analysis were conducted to compare genomically stable and unstable HNSCC samples. Thirty-six hub GI-related genes (GIGs) were identified and used to categorize patients into distinct clusters through consensus clustering analysis. A GI scoring (GIS) system was then developed to assess its relationship with somatic mutations, tumor mutational burden (TMB), and differential gene expression, including genes such as KRAS and TP53. In vitro experiments were performed to explore the functional mechanism of the GI-associated ceRNA axis—RNF216P1/let-7b-5p/DUSP9. The expression levels of RNF216P1, let-7b-5p, and DUSP9 were also validated using clinical samples from a local hospital.

Results

The identified 36 GIGs enabled the categorization of HNSCC patients into three distinct clusters, each exhibiting unique prognostic and immune profiles. The developed GIS system effectively distinguished between somatic mutations, TMB, and differential gene expression. Patients with higher GIS scores had better prognoses compared to those with lower scores. Additionally, GIS was positively correlated with overall immune cell infiltration and immune function, highlighting its potential in predicting responses to immunotherapy. The GI-associated ceRNA axis RNF216P1/let-7b-5p/DUSP9 was established, with The Cancer Genome Atlas (TCGA) analysis revealing upregulation of RNF216P1 and DUSP9 in tumor tissues, while let-7b-5p was downregulated. These expression trends were corroborated in clinical samples. In vitro experiments demonstrated that RNF216P1 functioned as a molecular sponge for let-7b-5p, leading to upregulation of DUSP9 and promoting oncogenesis in HNSCC.

Conclusions

The GIS system is an effective biomarker for evaluating GI, prognosis, and immune features in HNSCC. The findings also clarify the functional mechanism of the GI-related ceRNA axis RNF216P1/let-7b-5p/DUSP9, providing valuable insights for future research and the development of therapeutic strategies for HNSCC.

Highlight box.

Key findings

• A genomic instability (GI) scoring (GIS) system was developed to evaluate GI, prognosis, and immune features in head and neck squamous cell carcinoma (HNSCC).

• A GI-related competing endogenous RNA (ceRNA) axis, RNF216P1/let-7b-5p/DUSP9, was identified, with functional validation in clinical samples and in vitro experiments.

What is known and what is new?

• GI plays a key role in tumor progression and resistance in HNSCC, but its quantification and ceRNA involvement remain underexplored.

• This study introduces the GIS system for assessing GI and immune response, and identifies a ceRNA axis contributing to HNSCC oncogenesis.

What is the implication, and what should change now?

• The GIS system has potential clinical utility for patient stratification and prognosis prediction, as well as immunotherapy response assessment in HNSCC.

• Further research should focus on validating GIS and the ceRNA axis in larger cohorts and exploring targeted therapies based on these biomarkers.

Introduction

Head and neck squamous cell carcinoma (HNSCC) ranks as the sixth most common malignancy worldwide. Statistics from 2020 reveal a global incidence of approximately 750,000 new cases of HNSCC, accompanied by a mortality count of 365,000 (1). This represents an uptick from the 700,000 new cases and 360,000 fatalities recorded in 2018 (2), highlighting a concerning upward trend in both incidence and mortality rates for this malignancy. The complexity of HNSCC treatment partly stems from its high clinical heterogeneity, even among patients with similar tumor-node-metastasis (TNM) staging and human papillomavirus (HPV) infection status, the same treatment regimen can lead to vastly different prognoses (3). Additionally, the high risk of local recurrence and distant metastasis contributes to an overall 5-year survival rate for HNSCC patients that remains below 50% (4). Therefore, the exploration of novel biomarkers and therapeutic targets becomes crucial. These advancements promise deeper insights into the heterogeneity of HNSCC and offer more accurate predictions of clinical progression, thereby facilitating personalized treatment approaches for patients, while elucidating the molecular mechanisms of HNSCC.

Genomic instability (GI), characterized by DNA fragmentation and loss of genomic integrity, is not only pivotal in tumor progression and prognosis but also a hallmark of tumor (5,6). A recent study has underscored the significant role of GI in the pathogenesis of HNSCC (7). For instance, current research suggests that arecoline, a component of betel quid, may downregulate the expression of the ATM and BRCA1 genes, disrupting DNA damage response and repair mechanisms, and consequently leading to carcinogenesis (8). Furthermore, the increased mutational burden caused by GI can elicit the production of neoantigens, stimulating immune responses and modulating immune infiltration in the tumor microenvironment (TME) (9,10). This, in turn, enhances the sensitivity of tumors to immune checkpoint inhibitors, thereby potentially improving patient prognosis. Given the strong link between GI and HNSCC, studies have explored risk models based on GI-related long non-coding RNA (lncRNA) for HNSCC prognosis and immune infiltration (11,12). However, studies on GI-related genes (GIGs) in HNSCC are limited, particularly in understanding tumor heterogeneity or quantifying GI features using GIGs. Additionally, there is a lack of research on GI’s role in predicting immunotherapy response and its impact on the HNSCC immune microenvironment through single-cell sequencing.

The competing endogenous RNA (ceRNA) hypothesis provides a model for the regulation of gene expression, wherein transcripts such as messenger RNA (mRNA) and lncRNA modulate each other’s expression levels by competing for the same microRNA (miRNA) response elements (13). MiRNAs regulate the abundance of mRNA by binding to their target gene transcripts, typically resulting in translation inhibition. CeRNAs play a significant role in the prognosis and progression of HNSCC (14). However, the research on the involvement of ceRNAs associated with GI in HNSCC remains largely unexplored.

This study aims to utilize the integrative analysis of The Cancer Genome Atlas (TCGA) database for HNSCC patients, leveraging the mutation information therein to identify hub GIGs. Furthermore, by integrating data from both TCGA and Gene Expression Omnibus (GEO), the study aims to investigate the role of GI in the heterogeneity of HNSCC. Additionally, this research intends to construct a GI-related ceRNA regulatory network and preliminarily validate its mechanisms in HNSCC through in vitro experiments, thereby enhancing our understanding of the functional mechanisms of GI in HNSCC. We present this article in accordance with the MDAR reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-1925/rc).

Methods

Study design

Figure 1 illustrates the method for collecting and analyzing data. The following subsections provide details on each step.

Figure 1.

HNSCC data gathering and analysis flowchart. CeRNA, competing endogenous RNA; GIGs, genomic instability-related genes; GIS, genomic instability scoring; HNSCC, head and neck squamous cell carcinoma; MCC, maximal clique centrality; PCA, principal component analysis; qPCR, quantitative polymerase chain reaction; TCGA, The Cancer Genome Atlas; TMB, tumor mutational burden; WB, western blot; WGCNA, weighted gene co-expression network analysis.

Data acquisition and sample collection

This research incorporated a dataset from TCGA (https://portal.gdc.cancer.gov/), encompassing 547 samples that included 503 cases of HNSCC and 44 normal samples. Each sample provided a detailed genomic profile with data on copy number variations (CNVs), mutations, tumor mutational burden (TMB), gene expression, and survival data. After excluding cases with survival times less than 30 days, the study proceeded with 500 HNSCC samples. Additionally, the dataset was enriched with 270 HNSCC samples from the GEO (https://www.ncbi.nlm.nih.gov/geo/) dataset GSE65858. The gene expression matrices from TCGA were transformed from fragments per kilobase million (FPKM) to transcripts per kilobase million (TPM) values, followed by batch normalization. Subsequently, these data were merged with the GSE65858 dataset to create a unified dataset for in-depth clustering and analytical exploration. Single-cell RNA sequencing data were obtained from the GEO database (GSE139324), including 26 HNSCC samples and 5 normal samples. RNA methylation-related genes were sourced from previous literature (15). Moreover, from November to December 2023, clinical samples from 10 patients (8 males and 2 females, aged between 45 and 67 years, with a median age of 56 years) (Table 1) were collected from the Department of Head and Neck Oncology at Hubei Cancer Hospital. These included tumor tissues and adjacent non-tumor tissues, adhered to specific inclusion criteria: (I) pathologically confirmed HNSCC diagnosis; (II) no prior anti-tumor treatments; and (III) first occurrence of cancer. Exclusion criteria were (I) patients with distant metastases or local recurrence; (II) concurrent other malignancies; and (III) severe chronic diseases. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Hubei Cancer Hospital (No. LLHBCH2023YN-033), and all patients signed informed consent forms.

Table 1. Demographic characteristics of the 10 patients included in the study.

Patient ID	Gender	Age (years)
1	Male	45
2	Male	49
3	Male	52
4	Male	55
5	Male	60
6	Male	63
7	Male	67
8	Male	56
9	Female	53
10	Female	58

Identification of GIGs

Leveraging the TCGA-HNSCC database, we employed a systematic approach to identify GIGs as follows: (I) total mutations per patient were calculated using the Wilcoxon test, and patients were ranked in descending order of mutation count; (II) we compared the mRNA expression matrix between the top 25% genomic unstable (GU)-like group and the bottom 25% genomic stable (GS)-like group; (III) differentially expressed genes (DEGs) with a |log₂foldchange| >0.585 and a false discovery rate (FDR) adjusted P value <0.05 were identified; (IV) weighted gene co-expression network analysis (WGCNA) was conducted using the R ‘WGCNA’ package (version 1.71) (16), focusing on genes within the module that exhibited the highest correlation with HNSCC tumors for subsequent analysis; and (V) the intersection of genes identified in steps 3 and 4 was taken, resulting in the final list of hub GIGs.

Functional enrichment analysis

The biological functions of the GIGs were elucidated through Gene Ontology (GO) terms, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, and Disease Ontology (DO) analysis. This analysis was conducted using a suite of R packages: ‘clusterProfiler’ (version 3.25.0.002) (17), ‘org.Hs.eg.db’ (version 3.15.0), ‘enrichplot’ (version 1.18.4), and ‘DOSE’ (version 3.25.0.002) (18). Additionally, the biological functions across different clusters and groups were characterized using gene set variation analysis (GSVA) with the ‘GSVA’ package (version 1.44.0). The activity levels of pathways were quantified using the single-sample gene set enrichment analysis (ssGSEA) algorithm.

Unsupervised clustering analysis and construction of the GIS system

Our study undertook two independent rounds of unsupervised clustering analysis on the merge dataset. The initial analysis was conducted based on 36 GIGs, while the subsequent analysis focused on the intersecting DEGs among the three clusters identified from the first round. Both analyses employed the ‘ConsensusClusterPlus’ R package (version: 1.60.0) and the hierarchical clustering method, executed 1,000 times to ensure the stability of the clusters (19). Following the clustering, we developed a GIS. This system was formulated by first identifying DEGs between clusters derived from the initial clustering, followed by obtaining those DEGs associated with prognosis through univariate Cox regression analysis. Subsequently, the principal component analysis (PCA) method was applied, where the GIS for each sample was calculated based on the sum of scores from the first and second principal components of the expression levels of prognostically significant DEGs.

Immunity analysis

To explore the immune cell infiltration abundances, the activity levels of immune functions, and the state of the TME in each sample, we applied CIBERSORTx (20), ESTIMATE, and ssGSEA algorithm. The analysis was facilitated by the ‘GSVA’ R package (version 1.44.0) (21) and the ‘estimate’ R package (version 1.0.13) (22). Furthermore, the tumor immune dysfunction and exclusion (TIDE) algorithm was utilized to evaluate the potential immunotherapy responses across different groups. Single-cell RNA sequencing analysis was performed using ’seurat’ R package (version 4.3.0) for quality control and other analyses. Cell annotation markers were derived from previous literature (23). The activity scores of 36 GIGs in cells were analyzed using ‘AUCcell’ R package (version 1.18.1). Cell communication was analyzed using ‘CellChat’ R package (version 1.6.1).

Cell culture

The human HNSCC cell lines, FaDu (ATCC HTB-43, Homo sapiens) and Tca-8113 (CBP60426, Homo sapiens), were sourced from the American Type Culture Collection (ATCC, Baltimore, MD, USA) and the Cell Bank of the Chinese Academy of Sciences (Shanghai, China), respectively. These cell lines were cultured in Dulbecco’s modified Eagle medium (DMEM) and supplemented with 10% fetal bovine serum (FBS), both obtained from Thermo Fisher Scientific, Waltham, MA, USA. Additionally, the culture medium was enhanced with 100 units/mL of penicillin and 100 µg/mL of streptomycin to prevent bacterial contamination. The cells were maintained at 37 ℃ in a humidified atmosphere containing 5% CO₂, ensuring optimal conditions for cell growth and proliferation. Subculturing was conducted when cells reached 80–90% confluency to maintain healthy and viable cell populations for experimental analyses.

Quantitative polymerase chain reaction (qPCR)

Total RNA was extracted from HNSCC cell lines or tissues using TRIzol (Thermo Fisher Scientific) and reverse transcribed with a kit from the same company. qPCR was conducted using SYBR Green on an Applied Biosystems system, with β-actin as the normalization control. Primers for RNF216P1, let-7b-5p, DUSP9, and β-actin are listed in Table 2. Gene expression was quantified using the 2^−ΔΔCT method.

Table 2. Primer sequences for target genes analysis.

Gene	Forward	Reverse
DUSP9	CAGCCGTTCTGTCACCGTC	CAAGCTGCGCTCAAAGTCC
hsa-let-7b-5p	CCTCGCCTTTGCCGA TCC	GGATCTTCATGAGGTAGTCAGTC
RNF216P1	AACAACAATGAAGAGGTAA	TTTAGGCAACTTTGTCTC
β-actin	TCAAGAAGGTGGTGAAGCAG	TCAAAGGTGGAGGAGTGGGT

Western blot (WB)

Proteins from cells and tumor tissues were extracted using radioimmunoprecipitation assay (RIPA) buffer, and equal amounts of protein were loaded for sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), followed by transfer to polyvinylidene fluoride (PVDF) membranes. Membranes were blocked with 5% non-fat milk in Tween-20 (TBST) for 90 minutes. Primary antibodies against DUSP9 (1:2,000) and β-actin (1:10,000) from Abcam, Cambridge, UK, were incubated overnight at 4 ℃. After washing, membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibody (1:4,000, Abcam) for 2 hours at 25 ℃. Following TBST washes, membranes were developed using enhanced chemiluminescence (ECL) detection in a dark room, and protein expression was quantitatively analyzed.

Cell counting kit-8 (CCK-8) assay

Cell viability was assessed using the CCK-8 assay (Beyotime Institute of Biotechnology, Shanghai, China). Briefly, cells were seeded in 96-well plates and treated as indicated. CCK-8 solution was added to each well, and plates were incubated for 2 hours at 37 ℃. Absorbance was measured at 450 nm using a microplate reader.

Apoptosis assay

Apoptosis analysis was conducted using the Annexin V-fluorescein isothiocyanate (FITC)/propidium iodide (PI) apoptosis detection kit (Beyotime Institute of Biotechnology) following treatment under various conditions. Cells were harvested and centrifuged at 1,000 rpm for 3 minutes at room temperature. The cell pellet was washed twice with cold phosphate-buffered saline (PBS) and then resuspended in 100 µL of 1× binding buffer. To this suspension, 5 µL of Annexin V-FITC and 5 µL of PI were added. The cells were incubated for 20 minutes at room temperature in the dark. After incubation, 400 µL of binding buffer was added, and the samples were immediately analyzed by flow cytometry to determine the percentage of apoptotic cells.

Clonogenic assay

For clonogenic assays, a certain number of cells were seeded in six-well plates and allowed to grow for 2 weeks. Medium was changed every 3 days. Colonies were fixed with methanol, stained with 0.1% crystal violet, and counted.

Transwell assay

Invasion capabilities of the cells were assessed using transwell chambers (Corning Inc., Corning, NY, USA). The upper compartment of the chamber was coated with a mixture of DMEM and Matrigel (Thermo Fisher Scientific) in a 1:1 ratio. A total of 5×10⁴ cells were seeded into the upper chamber after transfection and incubated at 37 ℃ for 4 hours. The lower chamber was filled with 600 µL of DMEM containing 10% FBS, which served as a chemoattractant. The system was incubated at 37 ℃ in a 5% CO₂ atmosphere for 72 hours. Subsequently, chambers were washed twice with PBS, fixed in 4% paraformaldehyde for 20 minutes, and stained with 0.1% crystal violet for 30 minutes.

Statistical analysis

Graphical illustrations and statistical analyses were performed using the R statistical software (version 4.2.1) and GraphPad Prism (version 9.0.0). Continuous variables were expressed as mean ± standard deviation (SD). Clinical samples and cell experiments were repeated three times. An independent t-test was used to compare two groups, while one-way analysis of variance (ANOVA) was employed for comparisons among multiple groups. Components in bioinformatics data were compared using the Wilcoxon test. Survival analysis was conducted using the Kaplan-Meier method, and differences in survival rates were evaluated using the log-rank test. The Cox proportional hazards model was used for univariate analyses. Correlation analyses were performed using Pearson’s correlation coefficients. For all tests, a P value of less than 0.05 was considered statistically significant.

Result

Identification of hub GIGs in HNSCC patients

We initiated our study by employing WGCNA to obtain key genes intimately linked with HNSCC. After setting the soft threshold at 8, our analysis identified 14 gene expression modules, with the brown module exhibiting the highest correlation with HNSCC tumors, marked by a correlation coefficient of 0.45 (Figure 2A-2D). Subsequent to quantifying somatic mutations, we categorized the top 25% (124 samples) with the highest mutation count as GU-like, and the bottom 25% (127 samples) as GS-like. Furthermore, stringent criteria were applied to discern DEGs, selecting only those with a log-fold change (logFC) value exceeding 0.585 and an FDR-adjusted P value below 0.05 (Figure 2E). This process yielded 1,139 DEGs, termed GIGs. A deeper intersectional analysis between these 1,139 GIGs and the brown module led to the identification of 36 hub GIGs (Figure 2F). An extensive exploration of the biological functionality of these 36 hub GIGs highlighted their predominant enrichment in pathways related to GI and tumorigenesis, as delineated by GO and KEGG analyses. These pathways notably included homologous chromosome segregation, chromocenter formation, cyclin binding, and viral carcinogenesis (Figure 2G,2H). DO analysis further substantiated the close association of these genes with a spectrum of tumors, including HNSCC (Figure 2I). In summary, our comprehensive analysis identified 36 hub GIGs closely associated with GI and potentially crucial in HNSCC progression.

Figure 2.

Identification of GIGs. (A) Selection of the optimal power for the WGCNA, illustrated by the scale-free fit index (Y-axis) as a function of the soft-thresholding power (X-axis). (B) Dendrogram of all detected genes clustered based on a dissimilarity measure, where each color below represents a gene module. (C) Heatmap displaying the correlation between module eigengenes and clinical traits (tumor vs. normal), with each cell containing the corresponding correlation and P value. (D) Visualization of GS for HNSCC in the brown module, indicating the module’s relevance to the disease. (E) Heatmap representing DEGs between GS and GU samples, highlighting gene expression patterns. (F) Venn diagram illustrating the intersection of genes from the brown module and DEGs between GS and GU, identifying key overlapping genes. (G-I) Functional enrichment analyses including GO (G), KEGG (H), and DO (I), providing insights into the biological processes, pathways, and diseases associated with the 36 identified GIGs. BP, biological process; CC, cellular component; DEGs, differentially expressed genes; DO, Disease Ontology; GIGs, genomic instability-related genes; GO, Gene Ontology; GS, genomic stable; GU, genomic unstable; HNSCC, head and neck squamous cell carcinoma; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function; TCGA, The Cancer Genome Atlas; WGCNA, weighted gene co-expression network analysis.

Stratification of HNSCC based on the 36 GIGs

Next, we commenced with an in-depth analysis of genetic variations in GIGs. CNV frequency analysis revealed that the majority of GIGs exhibited CNV loss (Figure 3A). Among 433 samples, 115 (accounting for 26.56%) were found to harbor mutations in GIGs, with CDKN2A emerging as the most frequently mutated gene (Figure 3B). Additionally, the chromosomal distribution of GIGs was also delineated (Figure 3C). A subsequent analysis of the expression differences of 36 GIGs in tumor and normal tissues showed significant disparities in most GIGs, such as IGF2BP1 and SOX30 (Figure 3D), shedding light on the potential mechanisms of GIGs in HNSCC. Building upon these observations, we further applied a consensus cluster algorithm (Figure 4A). By integrating data from 770 HNSCC patient samples from the TCGA and GSE65858 datasets, we defined three distinct GI clusters based on these 36 GIGs. The study revealed significant prognostic differences among these clusters, with the best prognosis observed in cluster C and the worst in cluster B, indicating unique biological characteristics of each cluster (Figure 4B). GSVA-KEGG analysis demonstrated that pathways related to GI functions and immunity were activated in cluster C, including cell cycle, DNA replication, T-cell receptor signaling, and the immunoglobulin A (IgA) intestinal immune network. In contrast, pathways associated with tumor progression and immune suppression, such as JAK/STAT, NOD-like receptor signaling, and graft-versus-host disease, were activated in cluster A, with cluster B’s functional characteristics lying between those of clusters A and C (Figure 4C). To further elucidate the immune characteristics of these clusters, we conducted an ssGSEA analysis. The results indicated significant differences in immune features among the three GI clusters. Overall, cluster C exhibited a higher degree of immune cell infiltration and more active immune functions, whereas cluster B showed the opposite (Figure 4D,4E). Notably, aside from CD70, whose expression did not differ significantly among the three GI clusters, the other 42 immune checkpoint-related genes displayed marked differences, with the majority expressing highest in cluster C and lowest in cluster B (Figure 4F). The consistency of survival curves, GSVA-KEGG, ssGSEA, and immune checkpoint gene expression analyses further confirmed cluster C as an immune-activated GI-related cluster, cluster A as an immune-suppressed tumor-related cluster, and cluster B as an intermediate state cluster. In conclusion, these findings highlight the pivotal role of these 36 GIGs in unraveling the heterogeneity of HNSCC and provide valuable guidance for treatment decisions and prognostic assessment in HNSCC.

Figure 3.

Genetic variations in 36 GIGs. (A) Frequency of CNVs of the 36 GIGs in the TCGA cohort. (B) Mutation frequency of the 36 GIGs in the TCGA cohort. (C) Chromosomal distribution of CNV alterations in the 36 GIGs. (D) Comparative expression analysis of the 36 GIGs between normal and tumor tissues. ns, not significant (P>0.05); *, P<0.05; **, P<0.01; ***, P<0.001. CNVs, copy number variations; GIGs, genomic instability-related genes; TCGA, The Cancer Genome Atlas; TMB, tumor mutational burden.

Figure 4.

Prognostic and immune characteristics of the constructed three GI clusters. (A) The consensus matrix depicting the stability and agreement of patient classification into the three GI clusters. (B) Kaplan-Meier survival curves of the three GI clusters in the TCGA and GSE65858 cohorts, illustrating differences in patient survival among the clusters. (C) Heatmap of the activation status of biological behaviors among the three GI clusters, showing the activity of various biological processes and pathways. (D-F) Differences in immune-related features among the three GI clusters: immune-related functional activation (D), immune cell infiltration abundances (E), and expression of immune checkpoint genes (F), demonstrating the distinct immune landscapes of each cluster. ns, not significant (P>0.05); **, P<0.01; ***, P<0.001. GI, genomic instability; KEGG, Kyoto Encyclopedia of Genes and Genomes; TCGA, The Cancer Genome Atlas.

Construction of a quantitative GIS system based on 36 GIGs

In the context of HNSCC, where a specific system for quantifying GI is lacking, we initiated a comprehensive study. Initially, an intersection analysis among GI clusters A vs. B, A vs. C, and B vs. C yielded 1,969 overlapping DEGs (Figure 5A). Subsequently, unsupervised clustering analysis based on these DEGs identified three gene clusters with distinct prognostic differences (Figure 5B,5C). Further, univariate Cox regression analysis on these 1,969 DEGs singled out 276 DEGs with P values less than 0.05. These significant DEGs were employed to develop a quantitative GIS system, termed GIS, using the PCA algorithm. Our study revealed that patients with higher GIS scores had significantly better prognoses than those with lower scores (Figure 5D). The high consistency between GIS scores, GI clusters, and gene clusters was evident through gene expression heatmaps, Sankey diagram analysis, and the differential expression of GIS across various GI and gene clusters. Specifically, the highest level of GIGs expression and GIS scores were observed in GI cluster C and gene cluster C, while the lowest levels were noted in GI cluster A and gene cluster B (Figure 5E-5G). This finding underscores the reliability and consistency of our analysis method in distinguishing different states of GI. To further validate the efficacy of GIS in quantifying GI, we analyzed variations in GIGs and several mutational indicators between high and low GIS groups. Significant differences were observed in somatic mutation count, TMB, ARID1A, UBQLN4, KRAS, TP53, PIK3CA, and EGFR, indicating distinct genetic landscapes (Figure 6A,6B). GSVA analysis revealed that certain GI-related pathways, including DNA replication, cell cycle, and mismatch repair, were activated in the high GIS group, reflecting the system’s capacity to accurately represent the GI status in HNSCC patients (Figure 6C). Additionally, the ssGSEA algorithm was employed to illustrate differences in pathways between the two groups, aiming to identify potential targets related to GI (Figure 6D). In summary, we successfully developed a scoring system based on 36 GIGs, offering a quantitative approach to assess the GI status in HNSCC patients and potentially guiding targeted therapeutic strategies.

Figure 5.

Construction and characterization of the GIS system in HNSCC. (A) Intersection analysis revealing 1,969 overlapping DEGs among GI clusters A vs. B, A vs. C, and B vs. C. (B) The consensus matrix illustrating the stability and agreement in patient classification into three gene clusters. (C,D) Kaplan-Meier survival curves for the three gene clusters (C) and high vs. low GIS groups (D) in the TCGA and GSE65858 cohorts. (E-G) Comprehensive visual representations of the relationship between GIS, GI clusters, and gene clusters: gene expression and clinical feature heatmaps (E), Sankey diagram analysis (F), and patterns of GIS differential expression across GI and gene clusters (G), elucidating the intricate genomic interplay and its impact on patient stratification. DEGs, differentially expressed genes; GI, genomic instability; GIS, genomic instability scoring; HNSCC, head and neck squamous cell carcinoma; TCGA, The Cancer Genome Atlas.

Figure 6.

Characterization and pathway insights of the GIS system in HNSCC. (A,B) Comparative analysis uncovering significant variances in somatic mutation count, TMB (A), as well as expression levels of critical genes including ARID1A, UBQLN4, KRAS, TP53, PIK3CA, and EGFR (B), distinctly contrasting between the high and low GIS groups. (C) Heatmap of biological function differences between high and low GIS groups based on GSVA-KEGG pathway analysis. (D) Visualization of pathway activation disparities between high and low GIS groups, determined through ssGSEA. ns, not significant (P>0.05); *, P<0.05; **, P<0.01; ***, P<0.001. GEO, Gene Expression Omnibus; GIS, genomic instability scoring; GSVA, gene set variation analysis; HNSCC, head and neck squamous cell carcinoma; KEGG, Kyoto Encyclopedia of Genes and Genomes; ssGSEA, single-sample gene set enrichment analysis; TCGA, The Cancer Genome Atlas; TMB, tumor mutational burden.

Close association of the GIS system with immune characteristics

Recognizing the crucial role of immune cell infiltration in the TME, we delved deeper into the association between the GIS and immune status. Employing the CIBERSORTx algorithm, we discerned a positive correlation between GIS and the infiltration abundance of certain immune cells like CD8 T cells and M1 macrophages, while a negative correlation was observed with neutrophils (Figure 7A). The ssGSEA analysis not only corroborated these observations but also further indicated a close relationship between GIS and a majority of immune functions (P<0.05) (Figure 7B), suggesting a profound connection between GIS and the patient’s TME. Further analysis using the ESTIMATE algorithm demonstrated GIS’s capability to differentiate various states within the TME and the extent of stromal cell infiltration (Figure 7C), underscoring the significant role of GIS in deciphering the complexities of the TME. Moreover, the TIDE algorithm’s outcomes indicated that apart from dysfunction and M2-tumor-associated macrophage (TAM), which showed no significant variation between high and low GIS groups, other markers like TIDE score, interferon gamma (IFNG), and immune microsatellites exhibited notable differences (Figure 7D). Given the potential limitations of bulk RNA analysis in assessing heterogeneity, we further analyzed single-cell RNA sequencing data. By examining cells derived from 99,261 HNSCC patients and 22,086 normal samples (Figure 8A), we annotated 10 distinct immune cell types (Figure 8B). Notable differences in gene expression profiles between tumor and normal tissues were observed, such as changes in the marker genes of CD4⁺ T conv cells, with LTB, NOSIP, and RPS6 being replaced by NOSIP, CCR7, and GIMAP7 (Figure 8C,8D). To further investigate the role of GIS in the HNSCC immune microenvironment, we employed the ‘Auccell’ algorithm to calculate the scores of 36 GIGs in immune cells. The results revealed significant AUC score differences between normal and tumor-derived cells, with the most pronounced changes observed in plasmacytoid dendritic cells (pDCs) (Figure 8E). Based on this, we performed cell-cell communication analysis to explore the interaction of GIGs in immune cell signaling. The results indicated that a decrease in area under the curve (AUC) scores inhibited the communication between pDCs and B cells through the CD70-CD27 signaling pathway (Figure 8F). These analyses further confirm the role of GIS in the HNSCC immune microenvironment and support its potential in predicting responses to immunotherapy.

Figure 7.

Immune landscape analysis and correlation of GIS with immune features in HNSCC. (A) Bar graph illustrating the correlation of GIS with immune cell infiltration abundance based on the CIBERSORTx algorithm. (B) Heatmap representation of the association between GIS and immune functions, as well as immune cell infiltration abundance, determined by the ssGSEA algorithm. (C) Violin plot depicting the differences in TME scores between high and low GIS groups. (D) Bar graph showing the disparities in various indicators from the TIDE algorithm between high and low GIS groups. ns, not significant (P>0.05); *, P<0.05; **, P<0.01; ***, P<0.001. CAF, cancer-associated fibroblast; GIS, genomic instability scoring; HNSCC, head and neck squamous cell carcinoma; IFNG, interferon gamma; MDSC, myeloid-derived suppressor cell; MSI Expr Sig, microsatellite instability expression signature; ssGSEA, single-sample gene set enrichment analysis; TAM, tumor-associated macrophage; TIDE, tumor immune dysfunction and exclusion; TME, tumor microenvironment.

Figure 8.

Single-cell RNA sequencing analysis of immune cells in HNSCC and normal tissues. (A) UMAP visualization of sample origin. (B) UMAP visualization of cell types in the samples. (C) Dot plot showing the expression of various immune markers across different immune cell types. (D) Dot plot highlighting immune markers in tumor and normal-derived immune cell populations. (E) Dot plot displaying AUC score differences in 10 immune cell types between different samples. (F) Differences in receptor-ligand communication among 11 immune cell types across different samples. AUC, area under the curve; HNSCC, head and neck squamous cell carcinoma; NK, natural killer; pDC, plasmacytoid dendritic cell; UMAP, Uniform Manifold Approximation and Projection.

Construction of a GI-associated ceRNA network

Recognizing the pivotal role of ceRNA in the progression of various diseases, including cancer, we embarked on a detailed study of a GI-related ceRNA network. Initially, utilizing the maximal clique centrality (MCC) algorithm, we identified the top 10 genes with the highest nodes among 36 genes (Figure 9A). Subsequent univariate Cox regression analysis pinpointed SOX30, ZFR2, TCP11, and DUSP9 as significant prognostic factors (Figure 9B). Among these, only SOX30 and DUSP9 exhibited statistically significant differential expression between tumor and normal tissues. Given existing research on the role of SOX30 in HNSCC (24), our focus shifted towards DUSP9. Survival curve analysis indicated that patients with low DUSP9 expression had a better prognosis than those with high expression (Figure 9C). Moreover, data from TIMER2.0 (http://timer.cistrome.org/) also revealed significant expression differences of DUSP9 in various cancers like bladder urothelial carcinoma and colon adenocarcinoma, underscoring the potential importance of DUSP9 in cancer research (Figure 9D). Next, using the starBase platform (https://rnasysu.com/encori/), we identified the upstream regulatory miRNAs of DUSP9 (Figure 9E), among which only let-7b-5p showed a negative correlation with DUSP9 and was downregulated in tumor tissues compared to normal tissues (Figure 9F). Consequently, we selected let-7b-5p as the upstream miRNA of DUSP9. Continuing with starBase, we identified the upstream regulatory lncRNAs of let-7b-5p (Figure 9G). Through correlation and prognostic analysis, only RNF216P1 was found to be negatively correlated with the expression of let-7b-5p but positively correlated with DUSP9 (Figure 9H). Furthermore, RNF216P1 was more highly expressed in tumor tissues than in normal tissues, and patients with low RNF216P1 expression had a better prognosis than those with high expression (Figure 9H,9I). Further analysis of the relationship between cancer stemness and RNF216P1/let-7b-5p/DUSP9 revealed that, except for the lack of correlation between RNF216P1 and RNAss, other analyses showed statistical significance (P<0.05, Figure 10A,10B). Additionally, methylation-related genes such as TRMT61A were positively correlated with RNF216P1 and DUSP9, while negatively correlated with let-7b-5p (P<0.05, Figure 10C).

Figure 9.

Constructing the GI-related ceRNA network in HNSCC. (A) Application of the MCC algorithm to identify the top 10 node genes among 36 GIGs. (B) Univariate Cox regression analysis of the 36 GIGs identifying potential prognostic factors in HNSCC. (C) Kaplan-Meier survival curves comparing high vs. low DUSP9 expression groups. (D) Pancancer expression analysis of DUSP9, red indicates tumor tissues and blue indicates normal tissues. (E) Utilization of the starBase platform for the identification and correlation analysis of upstream regulatory miRNAs of DUSP9. (F) Correlation analysis between let-7b-5p and DUSP9 expressions and its differential expression between tumor and normal tissues. (G) Elucidation of the ceRNA network. (H) Analysis of the correlation between RNF216P1 with let-7b-5p and DUSP9, and its differential expression between tumor and normal tissues. (I) Kaplan-Meier survival curves for high vs. low RNF216P1 expression groups. **, P<0.01; ***, P<0.001. CeRNA, competing endogenous RNA; CI, confidence interval; GI, genomic instability; GIGs, genomic instability-related genes; HNSCC, head and neck squamous cell carcinoma; MCC, maximal clique centrality; miRNA, microRNA; TPM, transcripts per kilobase million.

Figure 10.

Potential mechanism of the RNF216P1/let-7b-5p/DUSP9 axis in HNSCC. (A,B) Scatter plots showing the correlation between RNF216P1, hsa-let-7b-5p, and DUSP9 expression levels with RNAss (A) and DNAss (B) expression. (C) Heatmap depicting the pairwise correlation coefficients of RNA methylation-related genes in the RNF216P1/let-7b-5p/DUSP9 axis. (D-F) Scatter plots showing the correlation between RNF216P1 (D), let-7b-5p (E), and DUSP9 (F) expression and the efficacy of various drugs. HNSCC, head and neck squamous cell carcinoma.

Analysis of the CellMiner database (https://discover.nci.nih.gov/cellminer/) showed that simvastatin was positively correlated with RNF216P1, palbociclib was negatively correlated with let-7b-5p, and irofulven was positively correlated with DUSP9 (P<0.05, Figure 10D-10F). These findings provide valuable insights into the potential of the RNF216P1/let-7b-5p/DUSP9 network for cancer therapy.

Exploration of the cellular mechanism underlying the GI-related ceRNA axis

To validate the expression trends of the RNF216P1/let-7b-5p/DUSP9 trio in HNSCC tumors, we conducted experiments on clinical samples from a local hospital. qPCR analysis of 10 samples confirmed that the mRNA expression trend of RNF216P1/let-7b-5p/DUSP9 aligned with TCGA database findings (Figure 11A). WB assays also verified consistent expression of DUSP9 in tumor tissues and adjacent non-tumor tissues (Figure 11A). Luciferase assays co-transfecting RNF216P1-wild type (WT)/mutant (MUT) and let-7b-5p mimic or NC into HNSCC cells indicated that let-7b-5p significantly reduced luciferase activity in RNF216P1-WT, but not in RNF216P1-MUT (Figure 11B). A similar trend was noted with DUSP9-WT and DUSP9-MUT (Figure 11C), confirming specific interactions between RNF216P1, let-7b-5p, and DUSP9. Further, to decipher the role of RNF216P1/let-7b-5p/DUSP9 in HNSCC, we engaged two HNSCC cell lines (FaDu and Tca-8113) in a series of experiments. qPCR revealed differential expression of RNF216P1 and let-7b-5p mRNA, along with DUSP9 protein, post-transfection with various plasmids (Figure 11D,11E). Notably, suppressing RNF216P1 increased let-7b-5p expression but decreased DUSP9 levels, a trend reversed by transfecting let-7b-5p mimics. Similarly, transfecting short hairpin (sh)-RNF216P1 plasmids reduced cellular vitality, clonogenicity, and invasiveness of HNSCC cell lines while inducing apoptosis (Figure 11F-11I). In contrast, overexpressing let-7b-5p reversed these effects, suggesting RNF216P1 may contribute to oncogenesis in HNSCC by inhibiting let-7b-5p. Additionally, transfecting let-7b-5p mimics decreased DUSP9 protein levels, a trend reversed by overexpressing DUSP9 plasmids (Figure 12A,12B). Also, post-transfection with let-7b-5p mimics, HNSCC cell vitality, clonogenicity, and invasiveness were suppressed, and apoptosis increased, effects reversed by transfecting DUSP9 plasmids (Figure 12C-12F). In summary, RNF216P1 likely acts as a sponge for let-7b-5p to enhance DUSP9 expression, contributing to HNSCC carcinogenesis, thereby unveiling a potential oncogenic GI-related ceRNA network.

Figure 11.

Expression analysis and molecular interactions within the GI-related ceRNA axis in HNSCC. (A) The mRNA expression of RNF216P1, let-7b-5p, and protein expression of DUSP9 between tumor tissues and adjacent non-tumor tissues from 10 clinical samples. (B,C) Luciferase reporter assays in HNSCC cells transfected with WT or MUT constructs of RNF216P1 (B) and DUSP9 (C) along with let-7b-5p mimic-NC or mimic. (D,E) qPCR analysis post-transfection illustrating the differential expression of RNF216P1, let-7b-5p mRNA, and DUSP9 protein. (F) Cell viability analysis using CCK-8 assay across different groups and time points. (G) Flow cytometry analysis depicting apoptosis rates in different groups. (H) Clonogenic assay results depicting the colony-forming abilities across groups (0.1% crystal violet). (I) Transwell assay outcomes indicating variations in cellular invasiveness among groups (cells were fixed with methanol and stained with 0.1% crystal violet; scale bar, 50 µm). **, P<0.01. CA, clonogenic assay; CCK-8, cell counting kit-8; ceRNA, competing endogenous RNA; FITC, fluorescein isothiocyanate; GI, genomic instability; HNSCC, head and neck squamous cell carcinoma; mRNA, messenger RNA; MUT, mutant; NC, negative control; PA, peritumoral adjacent; PI, propidium iodide; qPCR, quantitative polymerase chain reaction; sh, short hairpin; WT, wild type.

Figure 12.

Mechanistic insights into the GI-related ceRNA axis impacting cellular processes in HNSCC. (A,B) Impact of let-7b-5p mimic transfection on DUSP9 protein levels (A) and let-7b-5p mRNA levels (B). (C) Flow cytometry analysis depicting apoptosis rates across different experimental groups. (D) CCK-8 assays illustrating cellular viability across groups at various time points. (E) Clonogenic assays assessing the colony-forming abilities of HNSCC cells in different groups (0.1% crystal violet). (F) Transwell assays evaluating the invasiveness of HNSCC cells in various transfection conditions (0.1% crystal violet; scale bar, 50 µm). **, P<0.01. CCK-8, cell counting kit-8; ceRNA, competing endogenous RNA; FITC, fluorescein isothiocyanate; GI, genomic instability; HNSCC, head and neck squamous cell carcinoma; mRNA, messenger RNA; NC, negative control; PI, propidium iodide.

Discussion

GI can lead to enhanced degree of tumor malignancy (25), as well as resistance to radiation and chemotherapy (26), and has been found to have a close relationship with immunotherapy efficacy (27). Consequently, GI has emerged as a crucial and promising avenue in oncological research. With the advent of high-throughput sequencing technologies, numerous studies have utilized bioinformatics approaches to identify key targets throughout tumor progression, thereby enriching our understanding of the origins and development of cancer (28,29). Inspired by these advancements, extensive research has been conducted to explore the role of GI-related regulatory factors in tumorigenesis. For instance, studies investigating the networks of GI-related lncRNAs have further elucidated their impact on prognosis and immune infiltration in HNSCC (11,12). However, the selection of genes used in constructing current models is often limited, and most models merely categorize patients into high- or low-risk groups, failing to comprehensively reflect the complexity and heterogeneity of tumors. A more inclusive integration of genes may accurately portray the true biological state of tumors. In this context, we drew inspiration from a study that defined hypoxia groups using an extensive set of hypoxia-related genes, followed by the development of a scoring system to quantify the hypoxic condition of patients (23). Following a similar rationale, we employed the WGCNA algorithm and meticulously identified genes closely associated with GI in HNSCC under stringent selection criteria. Leveraging these meticulously selected genes, we then developed a GIS system, displaying unprecedented precision in identifying patient-specific GI characteristics. Particularly, the GIS scoring system revealed significant variations across different GIS score groups in an extensive set of up to eight genes and biomarkers closely associated with mutations, including but not limited to KRAS. This precision underscores the system’s capability in delineating the intricate landscape of GI within tumors and sets the stage for nuanced patient stratification and targeted therapeutic approaches.

Additionally, our study employed a variety of algorithms to precisely delineate the close nexus between the GIS and the density of immune cell infiltration, as well as the immunotherapy response. The results revealed that genomic GI has a dual impact on the tumor immune microenvironment: on one hand, it promotes immune responses by generating neoantigens and enhancing immune cell infiltration; on the other hand, it contributes to immune escape by upregulating immune checkpoint molecules and recruiting immune-suppressive cells. Specifically, immune scores, reflecting immune cell infiltration abundance, were significantly lower in the low GIS group compared to the high GIS group, with patients in the low GIS group having worse prognoses. This suggests that low GIS scores are associated with an immunosuppressive TME. Specifically, GIS scores were positively correlated with the infiltration of anti-tumor immune cells, such as CD8⁺ T cells and M1 macrophages, which are crucial for tumor cell recognition and elimination. This indicates that higher GIS may represent a TME with stronger immunogenicity and less immune suppression. GI has been shown to generate neoantigens through mutations. These neoantigens are recognized by the immune system, particularly by CD8⁺ T cells, which promote immune cell infiltration and tumor cell elimination (30). This explains the positive correlation observed between GIS and CD8⁺ T cell infiltration in our study. Moreover, GI enhances immune responses by fostering a pro-inflammatory TME (31). Tumor cells with high GI typically release pro-inflammatory cytokines and chemokines, such as interleukins and interferons, which recruit immune cells like M1 macrophages (32). This process strengthens the immune response and may facilitate tumor cell clearance, which is consistent with the positive correlation between high GIS and M1 macrophage infiltration. However, GI also contributes to immune escape (33). Accumulating mutations can upregulate immune checkpoint molecules, such as programmed death-ligand 1 (PD-L1), on tumor cells, which suppress T cell activation and enable tumor cells to evade immune surveillance (34). This may explain why TIDE scores were higher in the low GIS group compared to the high GIS group. Additionally, the immune-suppressive TME driven by GI may recruit immune-suppressive cells, such as myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs), further impairing the anti-tumor immune response (35,36). Single-cell RNA sequencing analysis revealed differences in the GI characteristics of immune cells between tumor and normal tissues, particularly in pDCs. Our study also found that GI may inhibit the CD70-CD27 receptor-ligand interaction, limiting communication between pDCs and B cells. This disruption may result from GI altering the expression of surface molecules or secreted factors in tumor cells, thereby interfering with immune cell signaling. The CD70-CD27 axis has been extensively studied in cancer, with research showing that CD27/CD70 is associated with immunotherapy resistance in renal cell carcinoma, and its role has also been explored in pancreatic adenocarcinoma (37-39). Therefore, our findings suggest that disruption of the CD70-CD27 interaction may contribute to immune escape and tumor progression in HNSCC. Moreover, we also identified pathways intimately linked with the GIS scoring system, including some pathways like KRAS and TP53, which have been established as closely related to GI (40-42). Hence, our research not only underscores the critical role of GI in tumor prognostication, immune characteristic analysis, and prediction of immunotherapy response but also successfully developed a highly precise GI quantification system. This system serves as a powerful tool for intricately mapping the GI features within the tumor pathology, thereby enhancing our understanding and potential intervention strategies in the complex landscape of HNSCC.

In light of the exceptional performance of the GIS system, we adopted a more rigorous approach for in-depth analysis. The MCC algorithm was employed to pinpoint key genes, and subsequent univariate Cox regression analysis identified prognostically significant factors, particularly highlighting DUSP9 as a critical gene. DUSP9 plays a role in the progression of various cancers and is associated with GI (43-45). It acts as a component of the ceRNA axis in several cancers (46,47), yet its role in HNSCC has not been fully explored. Our research addresses this gap, providing new insights into its function and significance in HNSCC. Our study validated the expression trend of DUSP9 in HNSCC using clinical samples and preliminary cellular experiments confirmed its oncogenic potential. Current research on GI-related ceRNA networks in HNSCC is scarce. Our work progressed further by constructing a ceRNA network centered around DUSP9, aiming to shed light on the role of GI-related ceRNA in HNSCC. We also identified the miRNA let-7b-5p as upstream of DUSP9, previously acknowledged for its tumor-suppressive function in HNSCC, aligning with our observations (48). Our study confirmed the downregulation of let-7b-5p in tumor tissues relative to normal tissues, bolstering the credibility of our findings. Additionally, we identified RNF216P1 as an upstream lncRNA of this let-7b-5p. While RNF216P1 has been studied in hepatocellular carcinoma (49,50), its involvement in HNSCC was unknown, marking our study as the first to report a GI-related ceRNA axis in HNSCC. Prior studies have noted the significant role of DUSP9 in tumorigenesis through various pathways, including mTOR and JNK signaling (51,52). Building on this foundation, our future research will delve deeper into the functional mechanisms of this axis and its precise relationship with GI, paving the way for novel insights and interventions in cancer therapy. Tumor stemness is a key factor in tumor progression, linked to invasiveness, metastasis, and therapy resistance (53). We analyzed the correlations between RNF216P1, let-7b-5p, and DUSP9 with tumor stemness markers (RNAss and DNAss). The results showed significant associations, suggesting that this ceRNA axis may regulate tumor stemness. For instance, tumor-derived exosomal lnc-Sox2ot has been shown to promote tumor stemness by acting as a ceRNA in pancreatic ductal adenocarcinoma (54), offering a new perspective for understanding ceRNA and stemness in HNSCC. Additionally, RNA methylation plays a critical role in tumor development (55). We found that RNF216P1, let-7b-5p, and DUSP9 are correlated with key RNA methyltransferases, such as METTL3 and METTL11, indicating that this ceRNA axis may influence HNSCC oncogenesis via RNA methylation. A previous study has demonstrated that METTL3 affects RNA methylation and tumor metastasis, supporting our hypothesis of this axis’ involvement in epigenetic regulation (56). Lastly, we explored the relationship between the RNF216P1/let-7b-5p/DUSP9 axis and drug resistance. Several drugs, such as simvastatin, palbociclib, and dasatinib, correlated with the ceRNA axis expression, suggesting their potential role in HNSCC therapy (57-59). Palbociclib, for example, inhibits HNSCC cell proliferation by regulating the cell cycle (60), which may synergize with RNF216P1 and let-7b-5p expression changes, supporting the axis’ potential role in drug resistance. These findings deepen our understanding of the RNF216P1/let-7b-5p/DUSP9 axis in HNSCC and offer theoretical insights for developing targeted therapies, particularly for overcoming tumor stemness and drug resistance.

While our research has made significant strides in the realm of HNSCC, including the pioneering development of a quantitative GIS system and the discovery of a GI-related ceRNA axis, along with preliminary validation of its mechanisms through cellular experiments and clinical samples, we acknowledge certain limitations of our study. Firstly, the clinical sample size our research relies on is relatively limited, and the absence of extended follow-up constrains our ability to thoroughly assess the prognostic accuracy and practical utility of the GIS system. Moreover, our study does not encompass an analysis of the response to various treatment modalities, such as chemotherapy or surgery, thus limiting the breadth of clinical decision-making applications for the scoring system. Secondly, while our cellular experiments have yielded valuable insights, the current findings are preliminary. We have not delved deeply into the specific mechanisms of the ceRNA axis, nor have we nor have we investigated the relationship between GI-related markers and ceRNA. Additionally, animal models have not been utilized to confirm the precise interaction with the TME, which is a crucial step in understanding its role in tumor biology. Given these limitations, future research will focus on expanding the clinical sample size and prolonging the follow-up period to enhance the accuracy of the GIS system in prognostic assessment. Furthermore, analyzing the response to diverse treatment modalities will allow a more comprehensive understanding of the clinical utility of the GIS system. Additionally, we plan to conduct more in-depth cellular, molecular, and animal studies to elucidate the specific mechanisms of the ceRNA axis and its interplay with the TME, laying the groundwork for the development of novel therapeutic strategies.

Conclusions

By utilizing the WGCNA algorithm and analyzing DEGs between the GS and GU groups, we identified 36 GIGs. Based on these genes, we delineated three clusters with distinct prognostic and immune characteristics. Moreover, we developed a scoring system with outstanding performance and identified a GI-related ceRNA axis RNF216P1/let-7b-5p/DUSP9, thereby significantly enhancing our understanding of the role of GI in the prognosis, heterogeneity, and immune features of HNSCC.

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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Hubei Cancer Hospital (No. LLHBCH2023YN-033), and all patients signed informed consent forms.

Data Sharing Statement

Available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-1925/dss

tcr-14-07-4115-dss.pdf ^{(70.8KB, pdf)}

DOI: 10.21037/tcr-24-1925