Single-cell and Mendelian analyses reveal shared mechanisms between head and neck neoplasms and aging

Abstract

Objective

This study aims to explore shared key genes between head and neck neoplasm (HNN) and aging.

Methods

Using single-cell RNA sequencing data of peripheral blood from HNSCC patients, aging individuals, and healthy controls, we identified cross-group co-expressed, downregulated cell subpopulations as core targets. Integrated pseudotime trajectory analysis and intercellular communication modeling were employed to investigate the dynamic evolution and functional interaction patterns of these subpopulations. Differentially expressed genes were identified, followed by Mendelian randomization (MR) analysis to assess their causal associations with HNN. Co-localization analysis was performed using GWAS data for HNN and expression quantitative trait loci (eQTL) datasets. Key genes were further subjected to metabolic pathway enrichment analysis.

Results

T cell subsets were found to be represented in both HNN and aging. Among them, CD4_naive T cells were down-regulated in both groups, leading to the identification of 24 differentially expressed genes. MR studies have shown that CCR, LEF1, NOSIP and FHIT have causal relationships with HNN. In the validation phase, however, only FHIT was retained, for which co-localization analysis revealed limited evidence of a shared causal variant between the GWAS and eQTL signals (H4 = 0.01). The metabolic enrichment highlighted metabolic pathways associated with these genes.

Conclusions

This study identified CD4_naive T cells down-regulation as a shared feature of HNN and aging and highlighted FHIT as potential molecular links. These findings may provide novel insights into the intersection of aging and tumorigenesis based on MR and single-cell analysis, offering potential targets for combined therapeutic strategies.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-04550-y.

Introduction

Head and neck neoplasm (HNN) is one of the most common malignant tumors worldwide [1], primarily occurring in the lips, oral cavity, pharynx, larynx, paranasal sinuses, as well as salivary gland tumors and mucosal melanoma [2]. Notably, head and neck cancer (HNC) ranks as the sixth most common cancer globally [3, 4]. In 2021, the International Agency for Research on Cancer (IARC) reported 840,000 new cases of HNC annually [5]. Smoking and alcohol consumption are considered the main risk factors [6]. Besides, cancer incidence generally increases with age, making aging a potential major risk factor [7, 8]. Thus, clarifying senescence-carcinogenesis molecular links is crucial for oncotherapy and prevention.

Aging involves physiological/pathological changes causing tissue/organ function decline [9]. Experimental data show senescence-mediated chemokine storm and cytokine imbalance are key pathways linking age-related pathologies to immune surveillance impairment [10, 11]. Research shows accelerated biological age increases cancer susceptibility (e.g. lung and colorectal) irrespective of chronological age, sex, or other risk factors [12].

In addition, mechanistic studies reveal Treg-derived cytokine networks critically regulate HNN tumorigenesis.These lymphocytes suppress immunity by secreting TGF-β, IL-10 and expressing CTLA-4, all linked to tumor progression [13]. Although preliminary studies have highlighted the involvement of immune cells in HNSCC pathology, the specific functions of different immune cell subsets —particularly T cells—and their association with clinical manifestations remain poorly defined. Recent advances in high-throughput sequencing technology provide unprecedented resolution to decode the clonal evolution of tumor-infiltrating lymphocytes during HNN progression and organismal aging processes.

Through a multi-cohort integration framework, we harmonized peripheral blood-derived scRNA-seq datasets encompassing HNN patients and age-stratified controls to systematically analyze the transcriptional heterogeneity of immune cell subsets, screen for key genes shared between aging and HNN, and analyze enriched pathways to delineate conserved immunoregulatory axes in HNN and senescence-associated immune remodeling. Furthermore, employing two-sample Mendelian randomization (MR), we conducted instrumental variable analyses to interrogate the causal key genes on HNN pathogenesis. This study’s characterization of pivotal molecular determinants and their regulatory networks offers critical mechanistic insights into HNN pathogenesis and aging-related biological processes.

Materials and methods

Analytical datasets were sourced from established open-access repositories.

Data source

This investigation leveraged The Gene Expression Omnibus(GEO) to procure senescence-associated and HNN-linked gene expression signatures. Specifically, we searched using the keywords “neck malignant neoplasm”, “senescence” and “blood” to identify relevant data sets. Then, we selected two samples (GSM5017033 and GSM5017021) from HNSCC data set (GSE164690) for HNN analysis, and two samples (GSM4750298 and GSM4750299) from Aging data set (GSE157007) for senescence-related analysis (Supplementary Table 1). GSM4750303 and GSM4750304 were used as control samples from GSE157007 dataset. All HNSCC cases were confirmed by oncologists and met the established diagnostic criteria for HNSCC [14].

To complement our gene expression analysis, we further incorporated genome-wide association study (GWAS) data related to HNN from the GWAS Catalog [15]. The training data set for HNN included 492 African American or Afro-Caribbean cases and 121,151 controls of the same ancestry, 4,578 European ancestry cases and 442,171 controls, as well as 233 Hispanic or Latin American cases and 59,498 controls (GCST90479818). Additionally, the validation data set for HNN specifically consisted of 4,671 European participants, including 2,342 confirmed cases (GCST012234). Genes identified from the MR analysis were further validated using the bulk RNA expression dataset GSE39400.

scRNA-seq analysis

In this study, we integrated and analyzed scRNA-seq datasets derived from HNSCC tissues, peripheral blood of aging individuals, and samples from healthy controls (Supplementary Table 1). Cell type annotation was carried out using canonical marker genes in combination with the SingleR (Version 2.8.0) algorithm. We identified a specific cell subpopulation showing consistent down-regulation in both HNSCC and senescent samples compared to controls, defined as the core cellular subpopulation (Log2FC < −0.5). Comprehensive data processing was performed using the Seurat R package (Version 4.4.0). Quality control involved the removal of low-quality cells with fewer than 500 or more than 4,000 detected genes, or with mitochondrial gene content exceeding 20%. Normalization, variance stabilization, and selection of highly variable genes were performed, providing robust input for downstream analysis. Dimensionality reduction was conducted through principal component analysis (PCA), retaining the top 5 components, followed by visualization using Uniform Manifold Approximation and Projection (UMAP). Cell clustering was executed using the FindNeighbors and FindClusters functions. Differentially expressed genes within each cluster were identified via the Wilcoxon rank-sum test using the FindAllMarkers function. Annotation was validated by referencing curated databases including CellMarker. Trajectory inference was carried out using the Slingshot R package (Version 2.14.0) to reconstruct potential cellular developmental pathways and lineage relationships. Additionally, we evaluated intercellular communication to explore signaling interactions within and between clusters. To investigate the functional significance of the core cellular subpopulation, we performed functional enrichment analysis on its marker genes. Finally, we elucidated the putative roles of the core cellular subpopulation in both HNSCC progression and aging-related cellular reprogramming, highlighting its relevance to disease pathogenesis and age-associated immune or stromal changes.

MR

To explore causal gene links between aging and HNN pathogenesis, we used two-sample MR frameworks. Univariable MR analyses were conducted via the TwoSampleMR R package (V0.6.14), with the inverse variance weighted (IVW) method as the main causal estimation approach.

To enhance the accuracy and credibility of the instrumental variables (IVs) applied in this MR analysis, a strict screening protocol was adopted. Initially, we extracted single nucleotide polymorphisms (SNPs) that showed strong associations with aging-related traits (P < 5 × 10⁻⁸) from large-scale GWAS conducted in populations of European descent. Subsequently, we applied linkage disequilibrium (LD) pruning with a cutoff of R² < 0.01 to eliminate correlated variants and retain only mutually independent SNPs. To minimize the impact of weak instruments, we calculated the F-statistic for each SNP using the formula F = β²/SE², and only retained those with F > 10, thereby improving the strength and reliability of the instrumental variables [16]. The selected SNPs were further validated by comparison with external datasets to confirm their stability and relevance to the exposure. To mitigate horizontal pleiotropic effects, we conducted comprehensive sensitivity analyses comprising MR-Egger regression, weighted median estimation, and leave-one-out validation. To assess heterogeneity, we calculated Cochran’s Q statistic, and used the Pleiotropy Residual Sum and Outlier (MR-PRESSO) method to detect and correct for outliers [17]. After excluding these outliers, the causal estimates were recalculated. In order to further validate our results, we selected the GWAS data of another HNN as the validation set. Additionally, we conducted differential analysis of the relevant genes and immune cell communication at the T cell level among HNSCC, aging and healthy groups.

Colocalization analysis and expression quantitative trait locus (eQTL) analyses.

We executed genomic colocalization assessments employing the coloc R package [18], aiming to identify potential shared association signals between GWAS data related to HNN and aging-related eQTL datasets. The H4 posterior probability, generated by the coloc.abf function, was used as the primary indicator of colocalization. In addition, we carried out regional association mapping to further investigate the relationships between SNP genotypes and the expression levels of key genes, in the context of the biological link between aging and head and neck tumorigenesis.

Moreover, to examine potential colocalization between the exposure (eQTL) and outcome (GWAS) datasets, we utilized the locuscomparer R package (Version 1.0.0), which visualized the relationship between the datasets. Finally, the results and datasets were saved for further investigation.

Metabolic enrichment analysis and bulk RNA validation

To gain deeper insights into the relationship between HNSCC and aging, we performed metabolic enrichment analysis to identify key biological pathways. The scMetabolism package (version 0.2.1), which integrates curated pathways from the KEGG and Reactome databases, was used to compute metabolic activity scores across different cell types. Single-cell metabolic states were evaluated using the AUCell algorithm, enabling the detection of subtle pathway activity changes. Finally, bulk RNA validation was conducted and the expression patterns of selected genes were visualized through heatmaps.

Statistical analysis

All statistical analyses were conducted using R version 4.1.3 (https://www.r-project.org/), with a two-sided P-value < 0.05 considered statistically significant. Mann–Whitney U test was used to analyse the differences between healthy and HNSCC group in bulk datasets. For each gene, expression values were normalized by calculating the log2 fold-change (log2FC) relative to the mean expression of the healthy group.

Results

Single-cell RNA-seq analysis

The proportions of the different cell type and T-cell subsets in three groups were shown in Fig. 1a and b, respectively, which show a total decrease in various immune cells in both HNSCC and aging groups compared to healthy group. A dot plot showing the markers used to identify each cell type for clarity and a UMAP plot to show the corresponding T-cell clusters in Supplementary Fig. 1a and b. The decrease in CD4_naive T cells in the HNSCC and aging groups relative to healthy controls was evident after sample aggregation (Fig. 1c), with the distribution of each individual sample provided in Supplementary Fig. 2. UMAP analysis revealed the global differentiation landscape and cellular heterogeneity of four distinct T cell subsets: CD4_EM, CD4_Naive, CD8_CM, and CD4_REG (Fig. 1d). The cellular differentiation trajectory is characterized by a sequential order from right to left: CD4-Naive, CD4-EM, and CD4-REG. As shown in Fig. 1e-g, in the three groups, the communication strength between CD4_Naive and B cells is basically consistent (involving 2 ligand-receptor pathways); the interaction strengths between CD4_Naive and CD4_EM, CD4_Naive and CD4_REG, CD4_Naive and CD8_CM are 1 or 2 in the HNSCC and healthy groups, but not manifested in the aging group. The cell communication pathway of CD4_naive T cells between other cells is heterogeneous in aging, HNSCC, and healthy group. In the aging and HNSCC groups, the communication is concentrated in the ANXA1-FPR1 pathway, while in the healthy group, it is concentrated in the MIF-(CD74 + CD44) pathway (Other specific contents are presented in supplementary Fig. 3). Finally, we identified 181 potential marker genes by contrasting CD4_naive T cells with other T-cell subsets, and 26 markers through CD4_naive T subset with other non-T cell types (Monocyte, NK-cell, B-cell). The intersection yielded 24 consensus differentially expressed genes (Supplementary Table 2).

Fig. 1.

Analysis of single-cell RNA-seq data for HNSCC, aging and healthy groups. (a) The proportion of different immune cell types in each group. (b) The different T-cell subpopulations in HNSCC, aging and healthy. (c) Bar graphs of the proportion of different T-cell subpopulations in HNSCC, aging, and healthy groups. (d) The global differentiation landscape and cellular heterogeneity of four distinct T cell subsets. e-g) The cell-cell communication of CD4_naive T cells relative to other non-T cell in HNSCC, aging and healthy group, respectively. (CM, Central Memory T Cell; EM, Effector Memory T Cell; REG, Regulatory T cells. e-g)

Fig. 3.

Result of Constructing cell communication and pseudo-time analyses. (a) The cells communication pathways between CD4_naive T cells and other non-T cell subsets with high and low FHIT expression. (b) The cell communication direction and proportion between CD4_Naive T cells relative to other subpopulations and other non-T cell subsets with high and low FHIT expression. (c) Pseudotime trajectory analysis of distinct temporal expression patterns of cell cycle-associated genes. (d) Pseudotime trajectory analysis of FHIT gene

MR

Further MR Analysis of the 24 differentially expressed genes identified above revealed four genes with causal effects on HNN risk. Figure 2a identifies aging-associated genes within CD4_naive T cells that demonstrate causal relationships with HNN. Specifically, as shown in Fig. 2b, the IVW methodology shows that CCR7 (nsnp = 5, OR = 0.0001, 95%CI:0.000–0.0425.000.0425, p = 0.002), LEF1 (nsnp = 3, OR = 0.0001, 95%CI:0.000–0.0674.000.0674, p = 0.005), NOSIP (nsnp = 5, OR = 0.0089, 95%CI: 0.0003–0.2424, p = 0.005), FHIT (nsnp = 13, OR = 114.76, 95%CI:1.3866–9495.4434.3866.4434, p = 0.035) are causally associated with HNN. Sensitivity analyses, including tests for heterogeneity and horizontal pleiotropy, supported the validity of our MR approach. Furthermore, MR-Egger regression revealed no evidence of horizontal pleiotropy, with all intercepts non-significant (P > 0.05).

Fig. 2.

Results of MR Analysis and Colocalization analysis. (a) Genes associated with aging in the CD4_naive T cells subpopulations and causally related to HNN. (b) Forest plot showing MR of key genes with HNN. (c) Results of key genes in the validation set with HNN. (d) Results of the colocalization analysis, showing the regional association maps for FHIT

To independently validate our findings, we applied the same analysis to an additional HNN GWAS dataset as the validation set, results are presented in Fig. 2c. The IVW methodology shows that FHIT (nsnp = 10, OR = 0.7326, 95%CI:0.5970–0.8991, p = 0.0029) is causally associated with HNN, while CCR7 (nsnp = 5, OR = 0.5732, 95%CI: 0.3057–1.0746, p = 0.08), LEF1 (nsnp = 3, OR = 0.5479, 95%CI:0.2678–1.1212, p = 0.0996), NOSIP (nsnp = 4, OR = 0.7359, 95%CI: 0.5358–1.0107, p = 0.058) are not causally associated with HNN. Additionally, the relevant genes at the T cell level and immune cell communication among HNSCC, anging and healthy are displayed in Sunpplementray Fig. 4–6.

Fig. 4.

Enrichment metabolism pathway analysis results of key genes. a-b) The FHIT are predominantly expressed in CD4_naive T cells. c) The major enrichment pathways of CD4_naive T cells with high and low FHIT expression. d) The FHIT NOSIP, CCR7, and LEF1 gene expression levels in healthy and HNSCC groups

Colocalization analysis

The results are shown in Fig. 2d, showing the results of colocalization analysis of HNN and FHIT: H0 = 4.97e-299, H1 = 2.90e-299, H2 = 0.63, H3 = 0.37, H4 = 0.01. The posterior probability (H4) of a shared causal variant between HNN and FHIT was estimated at 1%.

Constructing cell communication and pseudo-time analyses

As shown in Fig. 3a, the cells communication pathways between CD4_naive T cells relative to other subpopulations and other non-T cell subsets (including NK-cells, monocytes, B cells, CD8-CM, CD4-REG, CD4-EM) were different with high and low FHIT expression. The expression status of FHIT significantly influences the differentiation lineage of CD4_Naive T cells. The high-FHIT expression CD4_naive T cells tend to differentiate into monocytes and regulatory CD4⁺ T cells, while the low-FHIT expression CD4_naive T cells tend to differentiate into monocytes, CD4_REG, B cells, and NK cells (Fig. 3b).

Pseudotime trajectory analysis revealed distinct temporal expression patterns of cell cycle-associated genes: FHIT exhibited prominent expression during the early phase (0–10), while NOSIP, CCR7, and LEF1 were preferentially upregulated in the terminal phase (Fig. 3c), indicating potential involvement in cell cycle exit or checkpoint transitions. These temporally segregated expression profiles highlight their putative regulatory roles in different stages of cell cycle progression. As shown in Fig. 3d, the expression of the FHIT gene significantly decreased over pseudotime and this negative correlation is highly statistically significant (Pearson’s r = −0.13, p = 3.85e − 19), suggesting its potential involvement in early stages of the cell trajectory as well. Meanwhile, The green histogram at the top shows that most data points are concentrated in the early time interval (approximately 10–25 range). This indicates that the FHIT gene is more dense in the “early phases” of pseudotime, which may be a key interval for cell state transition.

Enrichment metabolism pathway analysis and bulk RNA validation of key genes

Figure 4a shows that FHIT is predominantly expressed in T cells subsets. The violin plot analysis results show that FHIT is significantly highly expressed in CD4_naive T cells, compared with the other three T cell subsets(Fig. 4b). The major CD4_naive T cells enrichment pathways were different with high and low FHIT expression. As shown in Fig. 4c: Compared to low- FHIT expression CD4_naive T cells, high-FHIT expression CD4_naive T cells were enriched with four metabolic pathways (glycosphingolipid biosynthesis-globo and isoglobo series, glycosphingolipid biosynthesis-ganglio series, glycosaminoglycan biosynthesis-keratan sulfate, other glycan degradation. For bulk RNA validation, as shown in Fig. 4d, the FHIT was no significant difference consistently in both healthy and HNSCC (P = 0.469), while NOSIP, CCR7 and LEF1 showed significantly elevated expression in healthy compared with HNSCC samples (P = 0.011, P = 008, and P = 0.015, respectively).

Discussion

HNN are a heterogeneous group of cancers, with HNSCC accounting for over 90% of cases, characterized by high global incidence and poor prognosis. To explore connections between HNN and aging, we focused on shared mechanisms and immunological mechanisms. Initially, we integrated multiple single-cell RNA sequencing datasets using established methodologies. Guided by prior studies [19–21], we prioritized T-cell populations for analysis. Both the HNSCC and the aging cohorts displayed significant alterations in T-cell subset distribution compared to healthy controls, notably marked by a pronounced reduction in CD4_naive T cells proportions. To further pinpoint central genes with potential causal roles in HNN pathogenesis, MR analysis was performed. Finally, metabolic enrichment analysis was conducted to investigate how these key genes interact with immune cells and metabolic pathways in both HNN and aging contexts. This integrated approach highlights shared molecular and immunological features between HNN and aging, offering insights into potential therapeutic targets for precision interventions.

CD4_naive T cells may participate in the immune alterations observed in both HNN and aging. Our data suggest potential functional changes in this subset, although further experimental validation is required to determine their direct roles in immune evasion or inflammation. IL-6 and TGF-β drive differentiation toward Th17 cells [22], promoting inflammation and tumor progression in HNSCC and chronic inflammation in aging [23, 24]. Additionally, IL-12 deficiency impairs Th1 differentiation, while Treg-mediated suppression weakens anti-tumor immunity and pathogen defense [25, 26]. Both conditions involve the presence of immunosuppressive factors (e.g., PD-L1, TGF-β) that induce Treg differentiation or exhaustion [27, 28], hindering CD8 + T and NK cell function [27, 29, 30]. The accumulation of Tregs correlates with advanced stages and poor outcomes [26]. In aging, thymic involution reduces CD4_naive T cells production, and senescent CD4_naive T cells exhibit impaired proliferation and cytokine production [31]. These cells express exhaustion markers like PD-1 and CTLA-4, contributing to chronic inflammation, impaired immune surveillance, and the accumulation of senescent cells [32–34]. Aging also skews CD4_naive T cells toward the Th17 lineage [35], exacerbating inflammation despite reduced immune efficacy [36]. Aging-related immune dysregulation involving CD4_naive T cells may promote HNN development [37, 38], suggesting aging as a potential risk factor.

The FHIT gene at chromosome 3p14.2 encodes a tumor suppressor regulating apoptosis, cell cycle, oxidative stress, and genomic stability [39–41]. FHIT loss (50–70% in HNN via deletion/skipping/aberrant transcripts) correlates with poor prognosis, metastasis, and therapy resistance [42–45]. Mechanistically, FHIT deficiency promotes genomic instability, impairs DNA repair, disrupts p53 signaling, enhances oncogene-induced senescence escape [46], and accelerates cancer progression [41, 47]. FHIT regulates mitochondrial ROS via FdxR/Hsp60 interactions, and its loss worsens oxidative damage and cellular aging [41, 46]. Clinically, restoring FHIT induces apoptosis/inhibits tumor growth, serving as a therapeutic target and response predictor [46–49], especially in heavy tobacco users [50]. Additionally, FHIT methylation has been linked to aging, though its relationship with protein expression remains unclear [39, 51]. While FHIT is not traditionally studied in CD4_naive T cells, its role in maintaining genomic integrity and oxidative balance warrants further investigation in the context of immune aging and cancer immunology. Although bulk RNA-seq analysis(Fig. 4d) did not show significant differential expression of FHIT between healthy and HNN samples, this does not contradict our conclusion that FHIT may serve as a molecular link between HNN and aging. Bulk RNA-seq represents an averaged signal across all cell types, which can mask gene-level alterations occurring in specific subpopulations. In contrast, our single-cell analysis revealed that FHIT is predominantly upregulated in CD4_naive T cells, a cell type strongly associated with immune aging. This observation is further supported by MR analysis, which suggests a potential causal effect of FHIT expression on HNN-related traits. Therefore, the combined evidence from single-cell resolved expression, cell-cell communication analysis, and MR-based causal inference supports the role of FHIT as a potential molecular link between HNN and aging.

Furthermore, our pseudotime analysis revealed that FHIT expression significantly decreases over pseudotime (Pearson’s r = −0.13, p = 3.85e − 19), as shown in Fig. 3c-d. This pattern suggests that FHIT is more highly expressed in the early pseudotime phase (approximately 10–25 range) and may play a regulatory role in cell fate determination or state transition. The declining expression of FHIT along the developmental trajectory aligns with its potential involvement in aging-related processes and supports its role as a molecular bridge between HNN and immunosenescence. A deeper understanding of FHIT-mediated pathways may offer novel strategies for managing HNN and age-related immune dysfunction.

In addition, we noticed changes in the effector memory (EM) T cell subset in our aggregated data (Fig. 1c), although we did not perform a detailed analysis of this population. Previous studies have shown that EM T cells expand with age and chronic antigen exposure, and a relative increase in EM T cells has also been reported in HNSCC tumor-infiltrating lymphocytes [52–54]. These observations are consistent with the concept of immunosenescence and suggest that EM T cell dynamics may also contribute to the parallel immune remodeling observed in HNSCC and aging.

In the CellChat network analysis, some cell types (e.g., CD4_REG or CD8_CM) do not show direct interactions with CD4_Naive(Fig. 1f). This is because Cellchat only displays statistically significant predicted interactions, which are filtered based on two criteria: a communication probability threshold (default 0.1) calculated from ligand and receptor expression levels, and a minimum cell number per cell type (here min.cells = 10). Therefore, lack of edges in the network does not imply absence of biological communication, but reflects that the predicted interaction probability did not pass the significance threshold under these criteria.

Interestingly, we also found that Fig. 1e shows the cell-cell communication intensity between CD4 naïve cells and CD4 EM cells based on all ligand-receptor interactions expressed in the entire CD4-naive population. In contrast, Fig. 3b stratifies CD4_naive T cells into FHIT⁺and FHIT⁻subpopulations, and the communication strength is calculated only using ligand-receptor pairs expressed within each FHIT⁻defined subgroup. Because FHIT⁺cells represent only a subset of CD4_naive T cells and have a distinct transcriptional profile, many ligand-receptor pairs that contribute to the strong communication seen in Fig. 1e are not expressed in the FHIT⁺subpopulation. As a result, the communication network in Fig. 3b appears weaker or even absent. Therefore, the difference between Figs. 1e and 3b does not represent a contradiction; instead, it reflects the heterogeneity within CD4_naive T cells and shows that FHIT expression defines a subpopulation with distinct communication behavior.

This study employs several innovative analytical strategies. First, by comparing differential genes between HNN and aging cohorts, we used gene heatmaps to visualize age-related expression patterns. Integrated with single-cell transcriptomic data, these heatmaps identified co-dysregulated genes (e.g., FHIT, NOSIP, CCR7, LEF1) as key targets for further investigation. Second, time-course analyses across HNN progression and aging stages revealed stage-specific gene expression and declining intercellular communication, linking these trends to immunosenescence and chronic inflammation. This temporal profiling also highlighted prognostically relevant, time-dependent genes with potential therapeutic value. Finally, cell communication modeling uncovered regulatory networks centered on FHIT, including its influence on pathways such as the Akt-survivin axis, offering a systems-level view for identifying targets of combinatorial therapy.

Despite these strengths of this study, there are limitations. Our conclusions are hypothesis-generating and based primarily on bioinformatic data, which require experimental validation. The primary dataset used was GEO, which focused on HNSCC, but the MR results were not replicated in an external HNN cohort, necessitating further study. While HNSCC accounts for over 90% of HNN, our findings may not fully apply to other HNN subtypes. Future research should explore non-HNSCC HNN subtypes, such as hematolymphoid malignancies, mesenchymal sarcomas, and neuroendocrine tumors. Furthermore, due to data limitations, reverse MR was not conducted to further verify the effects of HNN and aging on related genes.

Conclusion

This study leveraged single-cell transcriptomics and MR to explore the relationship between HNN and aging. Through MR and HNN-related transcriptomic analyses, we identified 4 candidate genes (NOSIP, CCR7, LEF1, and FHIT) that may contribute to the potential connection between aging and HNN. However, these findings require further experimental validation. Time-course analyses revealed stage-specific gene expression patterns and age-related declines in intercellular communication, linked to immunosenescence and chronic inflammation. Additionally, cell communication modeling emphasized FHIT regulatory role in tumor microenvironment signaling, suggesting new therapeutic targets. This research contributes to understanding the connection between aging and HNN, paving the way for precision interventions.

Author contributions

Chen Sun contributed to the conceptualization, data curation, formal analysis, investigation, methodology, resources, software, supervision, validation, visualization, writing – original draft, writing –review, and editing. Xinlei Chen contributed to the conceptualization, formal analysis, investigation, methodology, software, validation, visualization, writing – original draft, writing –review, and editing. Jinzhao Li contributed to the conceptualization, formal analysis, investigation, methodology, software, validation, visualization, writing – original draft, writing –review, and editing. Lirong Hu contributed to the conceptualization, formal analysis, investigation, methodology, software, validation, visualization, writing – original draft, writing –review, and editing. Rui Shi contributed to the conceptualization, data curation, funding acquisition, methodology, project administration, resources, supervision, writing – review, and editing. Chunhui Li contributed to the conceptualization, methodology, supervision, writing – review, and editing. Canli Wang contributed to the conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, writing – original draft, writing – review, and editing.

Funding

This work was supported by projects of Stomatology Hospital Affiliated to Southwest Medical University (NO. 201905 and NO. 2023Y06).

Data availability

Single-cell RNA sequencing data were obtained from GEO database: GSE164690)(https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE164690),(GSE157007)(https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE157007). Genome-wide association study (GWAS) data related to HNN from the GWAS Catalog: (GCST90479818)(https://www.ebi.ac.uk/gwas/studies/GCST90479818), (GCST012234) (https://www.ebi.ac.uk/gwas/studies/GCST012234), and GSE39400(https://www.ebi.ac.uk/gwas/studies/GSE39400).

Declarations

Ethics approval and consent to participate

This information is not available.

Consent for publication

All authors have consented to the publication of this manuscript.

Competing interests

The authors declare no competing interests.