Identification of two distinct head and neck squamous cell carcinoma subtypes based on fatty acid metabolism-related signatures: Implications for immunotherapy and chemotherapy

Abstract

The dysregulation of lipid metabolism is a critical factor in the initiation and progression of tumors. In this investigation, we aim to characterize the molecular subtypes of head and neck squamous cell carcinoma (HNSCC) based on their association with fatty acid metabolism and develop a prognostic risk model. The transcriptomic and clinical data about HNSCC were obtained from public databases. Clustering analysis was conducted on fatty acid metabolism genes (FAMG) associated with prognosis, utilizing the non-negative matrix factorization algorithm. The immune infiltration, response to immune therapy, and drug sensitivity between molecular subtypes were evaluated. Differential expression genes were identified between subtypes, and a prognostic model was constructed using Cox regression analyses. A nomogram for HNSCC was constructed and evaluated. Thirty FAMGs have been found to exhibit differential expression in HNSCC, out of which three are associated with HNSCC prognosis. By performing clustering analysis on these 3 genes, 2 distinct molecular subtypes of HNSCC were identified that exhibit significant heterogeneity in prognosis, immune landscape, and treatment response. Using a set of 7778 genes that displayed differential expression between the 2 molecular subtypes, a prognostic risk model for HNSCC was constructed comprising 11 genes. This model has the ability to stratify HNSCC patients into high-risk and low-risk groups, which exhibit significant differences in prognosis, immune infiltration, and immune therapy response. Moreover, our data suggest that this risk model is negatively correlated with B cells and most T cells, but positively correlated with macrophages, mast cells, and dendritic cells. Ultimately, we constructed a nomogram incorporating both the risk signature and radiotherapy, which has demonstrated exceptional performance in predicting prognosis for HNSCC patients. A molecular classification system and prognostic risk models were developed for HNSCC based on FAMGs. This study revealed the potential involvement of FAMGs in modulating tumor immune microenvironment and response to treatment.

1. Introduction

Head and neck squamous cell carcinoma (HNSCC) is a prevalent malignancy originating from the epithelial tissues of the oral cavity, pharynx, and larynx.^[1] It affects a significant number of individuals worldwide, with approximately 600,000 new cases diagnosed annually, ranking it as the sixth most common type of cancer.^[2] Despite an increasing incidence of HNSCC, there is a lack of effective early detection or screening methods, making careful physical examination the primary means of detecting HNSCC at an early stage. Smoking and alcohol consumption are major risk factors for oral and laryngeal cancer, while human papillomavirus infection is the primary cause of pharyngeal cancer.^[3,4] Classifying HNSCC into human papillomavirus-positive and negative groups is clinically meaningful, with the former exhibiting a better prognosis than the latter.^[1] Current treatment options mainly consist of surgery, radiotherapy, chemotherapy, targeted therapy, and immunotherapy. However, targeted therapy drugs have only demonstrated moderate overall response rates in patients.^[5] Although combined surgery, radiotherapy, and chemotherapy are commonly employed clinical treatment strategies, patients still exhibit high recurrence rates and low survival rates.^[6,7] Consequently, identifying a new prognostic biomarker is a primary objective in the treatment of HNSCC.

The significance of fatty acid metabolism (FAM) in cancer has been increasingly recognized. FAM exerts a critical influence on the initiation, progression, and resistance to the treatment of cancer by modulating lipid synthesis, storage, and breakdown processes.^[8,9] In cancer cells, FAM is activated to facilitate the production of cell membranes, signaling molecules, and energy storage through de novo fatty acid synthesis.^[10] Reprogramming of FAM pathway-associated genes represents a pivotal mechanism for sustaining the continuous proliferation and migration of cancer cells. Furthermore, FAM can also exert immune-suppressive effects through modulation of the phenotype and function of tumor-infiltrating immune cells within the tumor microenvironment (TME),^[11] and may confer survival advantages to tumors while ameliorating cellular stress associated with the metastatic cascade.^[11] Therefore, targeting FAM-related pathways to regulate fatty acid levels may represent a promising new strategy for cancer treatment.^[12] However, the precise roles and mechanisms of FAM in various types of cancer warrant further research and investigation. Presently, Du et al^[13] and Yuan et al^[14] have endeavored to develop risk models based on fatty acid metabolism genes (FAMG). Despite some progress, systematic studies are still required to explore the prognostic and treatment response roles of FAMGs in HNSCC.

The present investigation employed bioinformatics analysis to identify the molecular subtypes and develop a prognostic assessment model for HNSCC, utilizing FAMGs. The study comprehensively evaluated the efficacy of FAMGs in HNSCC prognosis, tumor immune microenvironment, and treatment response, thereby providing novel perspectives on personalized therapy and offering new evidence for investigating potential therapeutic targets for HNSCC.

2. Materials and methods

2.1. Data collection and processing

To conduct molecular subtype identification and prognosis model building, transcriptome and clinical data of the TCGA-HNSCC cohort were acquired from The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/, accessed data, January 20, 2023), and patients with incomplete prognosis data and clinical information were excluded. Validation of the established model was performed using transcriptome data and clinical information obtained from the GSE41613 project in the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/gds/) database. The HALLMARK_FATTY_ACID_METABOLISM dataset of the Molecular Signatures Database, which comprises a total of 158 FAM-related genes, was utilized to derive the genes related to FAM.

2.2. Non-negative matrix factorization clustering

The limma package was employed to perform differential expression analysis to identify aberrantly expressed FAMGs in HNSCC. A threshold was determined using |log (fold change)|>1 and an adjusted P value < .05 and genes that met this criterion were selected. The prognostic relevance of these FAMGs was then assessed using univariate Cox regression analysis. Subsequently, clustering analysis was performed using the non-negative matrix factorization package on the relevant FAMGs. The classification performance of 2 to 6 clusters was evaluated using the Brunet method. The optimal number of clusters was determined based on the k value at which the correlation coefficient began to decrease.

2.3. Immune infiltration analysis

The Cell-type Identification By Estimating Relative Subsets Of RNA Transcripts algorithm is a computational method employed to perform immune cell type analysis on single-cell RNA sequencing data. Its capability to precisely identify various immune cell types, such as T-cell subsets, B-cells, macrophages, dendritic cells, and others, has been widely acknowledged. In the context of the TCGA-HNSCC dataset, the Cell-type Identification By Estimating Relative Subsets Of RNA Transcripts package was utilized to scrutinize the immune infiltration status and to investigate differences in immune infiltration across distinct molecular subtypes and risk model groups.

2.4. Analysis of immune therapeutic response and drug sensitivity

Tumor immune dysfunction and exclusion (TIDE) is a metric utilized for the evaluation of immune response to tumors by integrating molecular features and clinical characteristics of tumor samples. It is capable of predicting the probability of immunotherapy responsiveness in patients through the analysis of gene expression data. Essentially, the TIDE website (http://tide.dfci.harvard.edu) utilizes normalized transcriptome data to calculate the efficacy of immunotherapy. The pRRophetic package was employed to conduct drug sensitivity analysis in TCGA-HNSCC. This involves the comparison of immune infiltration differences across distinct molecular subtypes and risk model groups.

2.5. Gene set enrichment analysis

Gene set enrichment analysis was a bioinformatic approach utilized to determine the association between genes in a gene set and specific biological processes, pathways, or functions present in gene expression or proteomics data. The functional enrichment of gene ontology and pathways associated with diverse subtypes can be studied using the clusterprofiler R software package. Statistical significance was considered when the enrichment threshold reached P < .05.

2.6. Prognostic model construction and evaluation

The study employed a molecular subtype classification of FAMGs to perform differential gene expression analysis. The goal was to identify genes that were differentially expressed and might be useful as independent prognostic markers. Univariate Cox regression analysis was used to screen for these genes, followed by the least absolute shrinkage and selection operator Cox regression analysis to prevent overfitting. Multivariate Cox regression was subsequently employed to determine the independent prognostic genes. A prognostic risk model was then built using the following prognostic risk model formula: risk score = $\sum \underset{i}{coef}\times \underset{i}{\exp }$, which considered the identified independent prognostic genes. Finally, the predictive performance of this risk model for prognosis was evaluated using receiver operating characteristic curve analysis.

2.7. Nomogram construction and evaluation

A multivariate Cox regression analysis was employed in the present investigation to ascertain independent prognostic factors for HNSCC. Subsequently, a nomogram model incorporating clinical and pathological characteristics was constructed by integrating the risk score with these independent prognostic factors. The rms package was utilized to develop this nomogram model. Moreover, calibration curve and decision curve analysis were conducted to assess the prognostic accuracy of the nomogram model.

2.8. Statistical analysis

The data analysis and visualization were performed using the R software (version 4.2.0, Vienna, Austria). Inter-group comparisons were conducted utilizing the Wilcoxon test, whereas survival analysis was carried out employing Kaplan–Meier survival curves and log-rank tests. The determination of statistical significance was based on a two-tailed P value of less than .05.

3. Results

3.1. Molecular subtypes of HNSCC based on FAMGs

Figure 1A presents the expression profile of FAMGs in HNSCC, where a total of 30 FAMGs have been detected to exhibit abnormal expression patterns in HNSCC. Among these genes, 14 genes are upregulated while 16 genes are downregulated. Univariate Cox regression analysis was carried out on these 16 genes, which identified galectin 1 and spermine synthase as risk genes, and vanin 1 as a protective gene, with significant association with HNSCC prognosis (Fig. 1B).

Figure 1.

Identification of HNSCC molecular subtypes based on prognosis-related FAMGs. (A) Volcano plot of differential expression of FAMGs in HNSCC. (B) Forest plot of prognosis-related FAMGs. (C and D) HNSCC molecular subtyping based on NMF algorithm. (E) PCA plot based on prognosis-related FAMGs. (F) K–M survival analysis between HNSCC molecular subtypes derived from FAMGs. FAMGs = fatty acid metabolism genes, HNSCC = head and neck squamous cell carcinoma, K–M = Kaplan–Meier, NMF = non-negative matrix factorization.

To further investigate the prognostic significance of these 3 FAMGs, a non-negative matrix factorization clustering analysis was performed based on their expression profiles, dividing the TCGA-HNSCC cohort into 2 molecular subtypes (Fig. 1C and D). Additionally, the PCA plot of these 3 FAMGs effectively distinguished the 2 subtypes (Fig. 1E). Further survival analysis showed that there were significant differences in the prognostic features between the 2 FAMG-derived subtypes, where cluster1 had a significantly better prognosis than cluster2 (Fig. 1F, P = .045).

3.2. Differences in immune landscape and treatment response among molecular subtypes derived from FAMGs

The immune microenvironment of tumors plays a critical role in predicting the tumor prognosis and response to therapy. Therefore, an investigation was conducted to assess whether there is heterogeneity in the immune landscape and treatment response among molecular subtypes derived from FAMGs. The results revealed significant variations in the infiltration of immune cells between cluster 1 and cluster 2 patients. Cluster 1 exhibited higher numbers of naive B cells, plasma cells, follicular helper T cells, and neutrophils, but lower numbers of resting CD4 memory T cells compared to cluster 2 (Fig. 2A). Drug sensitivity analysis showed that cluster 1 had lower sensitivity to methotrexate but higher sensitivity to cisplatin and docetaxel (Fig. 2B). Immune cell infiltration is recognized as one of the key influencing factors for immunotherapy response. The observed differences in immune cell infiltration among molecular subtypes derived from FAMGs suggest potential differences in immune therapeutic response. Furthermore, the analysis demonstrated that cluster 2 had a better TIDE score, indicating higher immune therapy resistance (Fig. 2C). Additionally, it was found that cluster 2 had better cancer-associated fibroblast and tumor-associated macrophage M2 scores but lower Merck18 scores (Fig. 2C).

Figure 2.

Differential immune infiltration and treatment response among molecular subtypes derived from FAMGs. (A) Comparison of immune cell infiltration among different molecular subtypes derived from FAMGs. (B) Comparison of drug sensitivity among different molecular subtypes derived from FAMGs. (C) Comparison of TIDE, Merck18, CAF, and TAM.M2 scores among molecular subtypes derived from FAMGs. CAF = cancer-associated fibroblast, FAMGs = fatty acid metabolism genes, TAM = tumor-associated macrophage, TIDE = tumor immune dysfunction and exclusion.

3.3. Gene expression characteristics of molecular subtypes derived from FAMGs

The gene set enrichment analysis has unveiled the gene regulatory mechanisms that account for the substantial molecular heterogeneity between FAMG-derived subtypes. The results demonstrated that cluster2 exhibits significantly activated processes related to muscle organ development, muscle contraction, and muscle cell development compared to cluster1 (Fig. 3A). In contrast, cluster2 displayed significant suppression of pathways associated with humoral immune response, immunoglobulin production, and antigen receptor-mediated signaling pathway. Moreover, cluster2 showed significant activation of pathways such as the ribosome, focal adhesion, and calcium signaling pathway, while cytokine-cytokine receptor interaction, Th17 cell differentiation, and antigen processing and presentation are significantly suppressed (Fig. 3B). These findings suggested that immune responses are considerably suppressed in cluster2, which could contribute to its inferior prognosis.

Figure 3.

GSEA analysis of molecular subtypes derived from FAMGs. (A) Significant GO terms activated and inhibited among molecular subtypes derived from FAMGs. (B) Significant KEGG pathways activated and inhibited among molecular subtypes derived from FAMGs. FAMGs = fatty acid metabolism genes, GO = gene ontology, GSEA = gene set enrichment analysis, KEGG = kyoto encyclopedia of genes and genomes.

3.4. Construction and evaluation of a prognostic risk model derived from FAMGs

To enhance the quantification of prognostic risk in HNSCC, we endeavored to develop a prognostic model utilizing molecular subtypes derived from FAMGs. The process involved differential expression analysis which identified 7778 genes with significant differences in expression, among which 133 genes were found to be associated with HNSCC prognosis. A least absolute shrinkage and selection operator Cox regression analysis was then applied to further reduce the number of genes to 27 (Fig. 4A and B), all of which were related to HNSCC prognosis. Ultimately, through multivariate Cox regression analysis, 11 independent prognostic genes were selected from these 27 genes for developing a prognostic model for HNSCC (Fig. 4C). The formula for the risk score model is as follows: risk score = 0.4668195 * CORO2B + 0.3267448 * HPRT1 - 9.3623536 * KIR3DX1 + 3.8100350 * MAGEB3 + 0.4994142 * METTL5 + 0.8191056 * PCGF2 + 0.3078017 * RRAGA − 0.8961419 * SLC25A45 − 0.9028340 * USP49 − 0.5220005 * ZNF700 − 0.5686048 * ZNF775.

Figure 4.

Independent prognosis gene identification based on molecular subtypes derived from FAMGs. (A) LASSO analysis of the 133 prognostic DEGs with 10-fold cross-validation. (B) Coefficient profile plots of 133 DEGs are shown, with vertical dashed lines indicating the optimal lambda values. (C) Forest plot illustrating the details of multivariate Cox regression analysis on the 27 DEGs. FAMGs = fatty acid metabolism genes, LASSO = least absolute shrinkage and selection operator.

Utilizing the aforementioned models, both TCGA-HNSCC and GSE41613 cohorts were partitioned into high-risk and low-risk groups (Fig. 5A and B). The survival time and outcomes of each patient in both cohorts were depicted in Figure 5C and D. Furthermore, the expression heatmap of 11 independent prognostic genes in both cohorts was illustrated in Figure 5E and F. Survival analysis demonstrated that patients in the low-risk group had a significantly better prognosis than those in the high-risk group, based on both the TCGA-HNSCC cohort (Fig. 5G, P < .0001) and the GSE41613 cohort (Fig. 5H, P = .011). Receiver operating characteristic curve analysis revealed that in the TCGA-HNSCC cohort, the area under the curve values of risk score for predicting 1-, 3-, and 5-year overall survival were 0.749, 0.804, and 0.751, respectively (Fig. 5I). In the GSE41613 cohort, the area under the curve values of risk score for predicting 1-, 3-, and 5-year overall survival (OS) were 0.687, 0.671, and 0.663, respectively (Fig. 5J).

Figure 5.

Construction and evaluation of the risk model derived from FAMGs. (A and B) Risk scores and grouping in TCGA-HNSCC and GSE41613 cohorts. (C and D) Scatter plots of overall survival and survival outcomes for all patients in TCGA-HNSCC and GSE41613 cohorts. (E and F) Heatmaps of gene expression for genes within the risk model in TCGA-HNSCC and GSE41613 cohorts. (G and H) Survival analysis evaluating prognostic differences between high- and low-risk groups in TCGA-HNSCC and GSE41613 cohorts. (I and J) ROC analysis evaluating the prognostic predictive performance of risk score in TCGA-HNSCC and GSE41613 cohorts. FAMGs = fatty acid metabolism genes, HNSCC = head and neck squamous cell carcinoma, ROC = receiver operating characteristic, TCGA = the cancer genome atlas.

3.5. Correlation of clinical pathological features with the risk model derived from FAMGs

To determine the potential associations between clinical and pathological features and a risk model developed from FAMGs, we conducted a comparison of risk scores across various groups with distinct clinical and pathological characteristics. The results of our analysis, presented in Figure 6A–I, demonstrated that patients who ultimately passed away had significantly higher risk scores compared to those who survived (P < 2.22e-16). Moreover, we observed that patients who received radiotherapy exhibited significantly lower risk scores than those who did not receive radiotherapy (P = .0046), patients with G4 grade had significantly lower risk scores than those with G1 (P = .0053), G2 (P = .00032), and G3 grade (P = .029), and patients who received chemotherapy also had significantly lower risk scores than those who did not receive chemotherapy (P = .0028). Our analysis also revealed that patients with the T4 stage had significantly higher risk scores than those with the T1 stage (P = .03). However, in other subgroups, we failed to observe statistically significant differences in risk scores.

Figure 6.

Differences in risk scores between different clinical and pathological feature subgroups. (A) Age, (B) survival outcome, (C) radiotherapy, (D) grade, (E) chemotherapy, (F) clinical stage, (G) T stage, (H) N stage, and (I) M stage.

3.6. Association of the risk model derived from FAMGs with immune landscape and treatment response

Given the substantial heterogeneity of molecular subtypes that arise from FAMGs in the immune landscape and response to therapy, we conducted an assessment of the relationship between risk score and both the immune landscape and treatment response. Our findings indicated that there were differences in cellular infiltration between patients in the high-risk group and low-risk group, primarily showing decreased levels of B cells, CD8 T cells, plasma cells, follicular helper T cells, and Tregs, and increased levels of M0 and M2 macrophages among those in the high-risk group (Fig. 7A). Furthermore, correlation analysis showed a negative association between the risk score and B cells and most T cells, while a positive association was found with macrophages, mast cells, and dendritic cells (Fig. 7B). We also observed that patients who responded to immunotherapy had significantly lower risk scores than non-responders (Fig. 7C). Among patients in the high-risk group, both TIDE and cancer-associated fibroblast scores were significantly elevated compared to those in the low-risk group (Fig. 7D and E), whereas Merck18 score was significantly lower in patients in the high-risk group (Fig. 7F). However, our sensitivity analysis of the 6 chemotherapy drugs mentioned above demonstrated no significant differences in drug sensitivity between patients in the high-risk and low-risk groups (data not presented).

Figure 7.

Relationship between FAMGs-derived risk model and immune landscape and immunotherapy response. (A) Comparison of infiltration patterns of 22 immune cells between high- and low-risk groups. (B) Correlation analysis between risk score and infiltration of 22 immune cells. (C) Comparison of risk scores between responders and non-responders to immunotherapy. (D–F) Comparison of TIDE, CAF, and Merck18 scores between high- and low-risk groups. CAF = cancer-associated fibroblast, FAMGs = fatty acid metabolism genes, TIDE = tumor immune dysfunction and exclusion.

3.7. Nomogram for the risk model derived from FAMGs

To enhance the practicality of FAMG-derived models in clinical settings, a nomogram model was developed to evaluate clinical prognosis. Initially, multivariate Cox regression analysis revealed (Fig. 8A) that 2 independent prognostic factors for HNSCC were the FAMGs-derived risk model (hazard ratio = 2.71, 95% confidence interval: 2.219–3.32, P < .001) and radiotherapy (hazard ratio = 0.56, 95% confidence interval: 0.382–0.83, P = .004). Consequently, a nomogram was constructed, which incorporated the risk score and radiotherapy to assess OS at 1, 3, and 5-year intervals for HNSCC patients (Fig. 8B). The c-index obtained for predicting 1-year OS was 0.81, indicating excellent discrimination ability. Additionally, the calibration curve of the model demonstrated high consistency between the predicted and observed survival rates at 1, 3, and 5 years (Fig. 8C). Decision curve analysis indicated that compared with other strategies in the TCGA cohort, the nomogram yielded better net benefit in predicting 1-year OS for HNSCC patients (Fig. 8D). Taken together, these results suggested that our risk model and nomogram could serve as valuable tools to forecast overall survival for HNSCC patients and facilitate informed clinical decision-making.

Figure 8.

Construction and evaluation of the HNSCC clinical prognostic nomogram model. (A) Multivariate Cox regression analysis of risk score and other clinical and pathological features. (B) Nomogram consisting of risk score and radiotherapy for predicting 1-, 3-, and 5-year overall survival in HNSCC. (C) Calibration curve using the nomogram to predict 1-, 3-, and 5-year overall survival rates. (D) Decision curve analysis comparing the predictive performance of the nomogram and other strategies for 1-year overall survival in HNSCC. HNSCC = head and neck squamous cell carcinoma.

4. Discussion

In this study, we systematically explored the expression and prognostic implications of FAMGs in HNSCC. Our findings led to the identification of 2 distinct molecular subtypes within HNSCC, characterized by varying degrees of immune infiltration and drug sensitivity. A FAMG-based risk model was constructed using 3 key FAMGs with strong prognostic relevance, which accurately predicted patient outcomes, including responses to chemotherapy and immunotherapy. The nomogram developed from our risk model showed excellent predictive power for 1-year OS, with a c-index of 0.81, and demonstrated high consistency between predicted and observed survival rates at 1, 3, and 5 years.

During the process of tumorigenesis, tumor cells undergo metabolic reprogramming that frequently results in alterations in FAM. To support their rapid proliferation and growth, tumor cells increase the synthesis and storage of lipids to meet their energy and biosynthetic demands. Moreover, by modulating the metabolic products of FAM, tumor cells can manipulate the function of immune cells, thereby suppressing immune responses and evading immune surveillance. HNSCC is a molecularly heterogeneous malignancy^[15] that has been closely linked to the biological functions of FAM.^[14] Despite this knowledge, the precise relationship between genes related to FAM and HNSCC heterogeneity and prognosis remains incompletely understood to date.

In this study, 2 molecular subtypes of HNSCC were identified based on prognostically relevant FAMGs. Galectin 1 was found to be a prognostic marker for HNSCC, and its inhibition could potentially improve prognosis through vascular normalization and increasing overall oxygenation of HNSCC.^[16] Spermine synthase was also identified as a prognostic biomarker for HNSCC,^[17] and it may co-maintain the survival of colorectal cancer cells with MYC by inhibiting Bim expression.^[18] In addition, overexpression of vanin 1 was associated with poor preoperative chemoradiotherapy response and prognosis in rectal cancer patients.^[19] Given the significant heterogeneity in prognosis and treatment response of molecular subtypes derived from FAMGs, further investigation of these genes’ roles and mechanisms in the tumor immune microenvironment and drug sensitivity is warranted.

Additionally, gene set enrichment analysis revealed that processes related to the humoral immune response, immunoglobulin production, antigen receptor-mediated signaling pathway, and B cell receptor signaling pathway were significantly suppressed in molecular subtypes with better prognosis, while processes related to muscle development were significantly activated. Furthermore, the cytokine-cytokine receptor interaction pathway was significantly suppressed, while pathways such as ribosome, focal adhesion, and motor protein were significantly activated. The humoral immune response is a critical component of tumor immune surveillance and clearance, capable of recognizing tumor-associated antigens and activating B cells to produce specific antibodies, thereby mediating cytotoxicity, promoting complement system activation, and more.^[20] Additionally, the humoral immune response can eliminate tumor cells through cell-mediated immune responses by activating T cells.^[21] Cytokine-receptor interactions also play an important regulatory role in tumor development - in some cases, they may promote tumor growth, progression, and metastasis, while in other cases, they may antagonize these outcomes.^[22] Therefore, it is possible that FAMGs may have a potential role in regulating immune responses. Further investigation is needed to explore the molecular mechanisms of FAMGs involved in the tumor immune microenvironment.

Recently, there has been a growing interest in developing prognostic features based on specific biological characteristics and gene sets in cancer research. These features can help clinicians classify patients into high-risk or low-risk groups and improve the accuracy of personalized treatment decisions. In this study, the differentially expressed genes between molecular subtypes, as determined by using FAMG, were used to construct a prognostic risk feature. The researchers evaluated the clinical significance and application value of this risk feature in various areas, including tumor immune microenvironment, immune therapy response, and drug sensitivity assessment. They also investigated the association between this risk feature and various clinical parameters such as survival status, grade, radiotherapy and chemotherapy, and T staging. The results showed that the risk features derived from FAMGs and radiotherapy were independent prognostic indicators for nasopharyngeal carcinoma. Furthermore, a predictive nomogram containing radiotherapy and risk features was developed, and its effectiveness was assessed through C-index, calibration plot, and decision curve analysis. This nomogram can be utilized to monitor the predictive discriminatory ability of OS in HNSCC patients.

The high recurrence and metastasis rates of HNSCC can be attributed to the interaction between the TME’s extracellular matrix and immune cells. This study explored the infiltration patterns of various immune cells, including B cells, natural killer cells, T lymphocytes, and macrophages, among different HNSCC molecular subtypes and risk groups. Prior research has established a connection between FAM and immune cell regulation. For example, an increase in circulating or TME-free fatty acids can hinder the activity of CD8+ cytotoxic T lymphocytes,^[23] while exogenous lipids may affect metabolic programs during natural killer cell activation, leading to reduced responsiveness to stimuli and impaired effector function and protective anti-tumor immunity.^[24] Moreover, lipid accumulation within tumor-associated dendritic cells may reduce antigen presentation, causing dendritic cells dysfunction and insufficient T-cell response stimulation, eventually resulting in an immunosuppressive TME.^[25,26] Macrophages are classified as either classically activated (M1) or activated (M2).^[27] Studies have shown that M2 macrophages predict poor prognosis in HNSCC,^[28] whereas the molecular subtype associated with a better prognosis showed lower infiltration of M2 macrophages. Although M1 macrophages exhibit anti-tumor properties through their inflammatory response to cancer cells,^[29] no difference was observed among molecular subtypes in this study. Generally, HNSCC subgroups with worse prognoses exhibit a more immunosuppressive TME. As immunotherapy gains importance in various cancers, including HNSCC, further observation is necessary to identify suitable patients for treatment.^[30] The HNSCC molecular subtypes and prognostic risk models identified in this study hold promise in evaluating immunotherapy due to the differential immune cell infiltration among these tumor immune microenvironment subtypes.

Despite the valuable findings obtained from this study, certain limitations exist. Firstly, the performance of the risk model developed has not been validated through clinical data, and future multi-center large sample verification experiments will be necessary. Furthermore, additional evaluation is required to fully understand the underlying mechanisms of the marker genes utilized in constructing molecular subtypes and prognostic risk models for HNSCC, particularly their influence on the formation of the TME and the development of treatment resistance.

5. Conclusion

In summary, a systematic characterization of FAMGs’ expression and prognostic significance was conducted in HNSCC through our study. Additionally, 2 molecular subtypes with substantial heterogeneity were identified, and a FAMG-based prognostic risk model employing 3 prognostic-related FAMGs was established. Although the prognostic model developed in this study provided an accurate and dependable prediction of HNSCC prognosis, including its response to chemotherapy and immunotherapy, further research is necessary to elucidate the mechanisms and functions of key genes, as well as the clinical applicability of the molecular subtypes and prognostic models.

Author contributions

Conceptualization: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Data curation: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Formal analysis: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Funding acquisition: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Investigation: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Methodology: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Project administration: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Resources: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Software: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Supervision: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Validation: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang, Yufeng Wang.

Visualization: Jianjun Zou, Yanbi Dai, Guangbo Xu, Yilong Kai, Lingfeng Lan, Junkun Zhang.

Writing – original draft: Jianjun Zou, Yanbi Dai.

Writing – review & editing: Jianjun Zou, Yanbi Dai.