Development and validation of a novel palmitoylation-related prognostic signature in head and neck squamous cell carcinoma

Introduction

Palmitoylation is a reversible post-translational modification (PTM) in which a palmitic acid (a 16-carbon saturated fatty acid) is covalently attached to specific cysteine residues through a thioester bond. This modification has been identified in a wide range of cancer-related proteins, including both oncoproteins (1) and tumor suppressors (2). The primary functions of palmitoylation are associated with membrane localization and protein stability, making it a crucial regulator in cancer biology (3,4). By influencing protein trafficking, signal transduction, and membrane interaction, palmitoylation plays a key role in driving cancer cell survival, proliferation, and drug resistance (5,6). Targeting palmitoylation-related enzymes or pathways may offer new therapeutic strategies for cancer treatment. Dysregulation of palmitoylation has been implicated in the progression of several cancers, including liver cancer (7), breast cancer (8), colorectal cancer (9,10), bladder cancer (11), pancreatic cancer (12), ovarian cancer (13), esophageal cancer (14), prostate cancer (15), and stomach cancer (16). However, no reports have documented the role of palmitoylation in head and neck squamous cell carcinoma (HNSCC), highlighting a potential area for future research.

HNSCC is a group of malignancies originating from the epithelial squamous cells lining the tissue of the head and neck region, including the oral cavity, pharynx, larynx, nasal cavity, and paranasal sinuses. As of 2023, HNSCC remains the seventh most common cancer worldwide, with approximately 890,000 new cases and 450,000 deaths reported annually, accounting for about 4.5% of all cancer diagnoses and deaths (17). The risk factors have been extensively studied. In developing countries, alcohol and tobacco are the primary risk factors for HNSCC. In developed countries, human papillomavirus (HPV) infection, particularly HPV-16 and HPV-18, is a leading cause of oropharyngeal HNSCC (18). The major molecular pathways implicated in HNSCC include: p53 signaling (19), PI3K/ARK/mTOR pathway (20), NOTCH signaling (21), EGFR signaling (22), Wnt/β-catenin signaling (23), HPV-related E6/E7 proteins (24). Given that palmitoylation regulates genome maintenance pathways, and genomic instability is a hallmark of HNSCC, dysregulated palmitoylation may represent a novel driver of HNSCC progression (25). EGFR, PIK3CA, and Ras are known to be palmitoylated in other solid cancers, in which palmitoylation regulates membrane localization, signal transduction, and protein stability. However, it remains unclear whether these proteins are dysregulated through palmitoylation in HNSCC. Interestingly, palmitoylation has been identified as a prognostic factor in cancers such as glioblastoma (26), liver cancer (27), and ovarian cancer (13). However, its potential as a prognostic marker in HNSCC remains unexplored.

The prognosis of HNSCC varies considerably due to its heterogeneous nature but is generally poor, with a 5-year overall survival (OS) rate of only 40–50%. Prognosis is influenced by factors such as HPV status, age, sex, and tumor stage. While various prognostic models have been developed, including clinical, demographic, and radiomics-based approaches, gene-based models are becoming increasingly important. Recent studies have identified potential biomarkers, such as the immune-related gene signature reported by Wei et al. (28) and the anoikis-related long non-coding RNA-based model proposed by Yuan et al. (29). However, the predictive performance of these models remains limited, with modest area under curve (AUC) values, suggesting that additional molecular factors may contribute to the pathogenesis and heterogeneity of HNSCC. Therefore, exploring novel gene signatures, such as those related to palmitoylation, may not only enhance our understanding of HNSCC etiology but also facilitate the development of more accurate prognostic models and personalized therapeutic strategies.

We systematically evaluated RNA-sequencing and clinical data to establish a palmitoylation-related prognostic model for HNSCC. The model was further used to predict individual prognoses for HNSCC patients, providing new insights into the clinical relevance of palmitoylation in HNSCC. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1228/rc).

Methods

Data acquisition and preprocessing

UCSC Xena (https://xena.ucsc.edu/) was used to acquire The Cancer Genome Atlas (TCGA)-HNSCC cohort. Star count (n=566) and ID/gene mapping were downloaded. Phenotype (n=604) and survival data (n=603) were also obtained. The discrepancy in sample counts across TCGA RNA sequencing (RNA-seq), clinical, and survival datasets reflects differences in data availability; only samples with complete RNA-seq and clinical information were retained for downstream analysis. After removing 2 metastatic samples, 564 samples were used for differential expression analysis. After removing cases with missing clinical information (age, sex, clinical stage, N stage, and T stage), 462 samples were included for nomogram construction. One independent microarray data set was downloaded from the GEO database (Accession No. GSE41613, n=97). The microarray data were processed using quantile normalization and z-score normalization with the R package “preprocessCore”. The specific clinical and pathological characteristics of the two cohorts are shown in Table 1. A total of 4,052 palmitoylation-related genes (PRGs) were retrieved from the Molecular Signatures Database (MsigDB) using the R package “msigdbr”. The complete PRGs are listed in table online: https://cdn.amegroups.cn/static/public/tcr-2025-1228-1.xls. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Differentially expressed mRNA analysis and functional analysis

Differentially expressed genes (DEGs) were identified using the R package “limma” by comparing the tumor group (n=520) with the normal group (n=44) after excluding two metastatic samples. Genes are considered as differentially expressed if they have a log₂(fold change) >log₂(1.5) or <−log₂(1.5), and a P value <0.05. A Venn diagram was generated using the R package “VennDiagram” to visualize the shared genes between DEGs and PRGs.

Functional enrichment analysis of the shared genes was performed by using the R package “clusterProfiler” to identify enrichment for specific biological terms.

Establishment and validation of a prognostic model

The TCGA-HNSCC cohort was used as the training set. Using a univariate Cox regression analysis, we identified the candidate PRGs associated with prognosis. Next, feature selection was conducted using the least absolute shrinkage and selection operator (LASSO) with the R package “glmnet”. The lambda 1se criterion was used in LASSO to prioritize model simplicity and generalizability, balancing predictive performance with reduced risk of overfitting compared to lambda min. Subsequently, multivariate Cox regression was implemented with the R package “survival” and “survminer”. Non-significant genes with P value >0.05 were removed. The coefficients obtained from the multivariate Cox analysis were used to calculate the risk score for each patient, following the formula: Risk score = Σ (βᵢ × Exprᵢ), where βᵢ represents the regression coefficient for each gene, and Exprᵢ is the corresponding gene expression level. Using this approach, a risk scoring model was established, and each patient was assigned an individual risk score. Patients in the TCGA dataset were then stratified into high- and low-risk groups based on the median risk score. The survival differences between these groups were evaluated using the Kaplan-Meier method and the log-rank test. The diagnostic accuracy was demonstrated by developing time-dependent receiver operating characteristic (ROC) curves and by computing the area under the curve (AUC) values. The predictive performance of the risk score model was further validated in an independent external cohort (GSE41613), using the same set of prognostic signature genes and following the same procedure.

Development of a nomogram

Clinical predictors (age, sex, clinical stage, T stage, and N stage) were first evaluated using univariate Cox regression, and those with significant prognostic value (P<0.05) were further incorporated into a multivariate Cox regression model together with the PRG-based risk score. Independent predictors identified from this analysis were then integrated into a nomogram using the R package “rms”. The predictive performance of the nomogram was assessed by ROC analysis and the corresponding AUC values.

Risk model-based analysis of immune cell infiltration

The R package “estimate” was used to compute the ImmuneScore, StromalScore, EstimateScore, and tumor purity of each sample. The results were visualized by the R package “ggpubr”. Additionally, the expression level of immune checkpoint genes between risk groups was explored. The CIBERSORT algorithm was used to estimate the proportions of immune cell types between the high-risk (n=281) and low-risk (n=282) groups. Immune infiltration was calculated from log₂-transformed, normalized RNA-seq data, with both groups exceeding the recommended minimum sample size of 50.

Gene set enrichment analysis (GSEA)

The R package “limma” was used to calculate the log₂(fold change) values of the gene list between the high- and low-risk groups. Hallmark gene set was retrieved from Molecular Signagure Database (MsigDB) using the R package “msigdbr”. GSEA was performed by the R package “clusterProfiler” and the results were visualized by “enrichplot” and “ggplot2”.

Risk model-based analysis of drug sensitivity

The R package “oncoPredict” was used to calculate the half maximum inhibitory concentration (IC₅₀) between the high- and low-risk groups. Group comparisons were analyzed using non-parametric Wilcoxon rank-sum tests. The lower the IC₅₀ value, the better the drug sensitivity.

Statistical analysis

The R programming language (version 4.5.0) was utilized to perform a series of bioinformatics analyses. P<0.05 was identified to indicate statistical significance. The predictive performance of the risk model and nomogram was evaluated using ROC curves and AUC values. According to established criteria, an AUC of 0.5 indicates no discrimination, 0.6–0.7 fair discrimination, 0.7–0.8 good discrimination, and >0.8 excellent discrimination. However, AUC interpretation is context-dependent, and moderate values may still hold clinical significance in heterogeneous diseases such as HNSCC.

Results

Identification of 282 differentially expressed PRGs (DE-PRGs) in HNSCC

A study flowchart is presented in Figure 1. A total of 4,138 DEGs were identified from the TCGA-HNSCC cohort by comparing tumor and normal tissues using the R package “limma”. The volcano plot of these DEGs is shown in Figure 2A. By intersecting the 4,138 HNSCC DEGs with 4,052 PRGs, we identified 282 DE-PRGs (Figure 2B). A heatmap of the top 100 DE-PRGs is displayed in Figure 2C, and the volcano plot of all 282 DE-PRGs is shown in Figure 2D. The complete list of DE-PRGs is provided in table online: https://cdn.amegroups.cn/static/public/tcr-2025-1228-2.xls.

Figure 1 Flow chart of the current analysis. GO, Gene Ontology; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; HNSCC, head and neck squamous cell carcinoma; LASSO, least absolute shrinkage and selection operator; PRGs, palmitoylation-related genes; ROC, receiver operating characteristics; TCGA, The Cancer Genome Atlas.

Figure 2 Differentially expressed palmitoylation-related genes and functional analysis. (A) Volcano plot showing a total of 4,138 DEGs in the TCGA-HNSCC tumor and normal group. (B) Venn diagram showing that the overlap of 4,138 DEGs and 4,052 PRGs led to 282 DE-PRGs being identified. (C) Heat map showing the top 100 DE-PRGs. (D) Volcano plot showing the distribution of 282 DE-PRGs. (E) Circular visualization of GO enrichment and pathway analysis of 282 DE-PRGs. (F) Boxplot showing the KEGG enrichment of 282 DE-PRGs. DEGs, differentially expressed genes; DE-PRGs, differentially expressed palmitoylation-related genes; FC, fold change; GO, Gene Ontology; HNSCC, head and neck squamous cell carcinoma; KEGG, Kyoto Encyclopedia of Genes and Genomes; TCGA, The Cancer Genome Atlas.

Functional analysis of DE-PRGs

The potential biological function of these 282 DE-PRGs was explored through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis (Figure 2E,2F). GO analysis reveals significant enrichment in the fatty acid metabolic process, the regulation of lipid metabolic processes, and lipid transport. These findings confirm that these genes are indeed involved in palmitoylation-related pathways. KEGG analysis identified enrichment in retinol metabolism, where the activity and stability of the key proteins are regulated by palmitoylation (30), and steroid hormone biosynthesis, which has also been linked to palmitoylation-related regulation (31).

Identification of novel prognostic genes

We conducted univariate Cox screening on the 282 DE-PRGs using the R packages “survival” and “survminer”. A total of 603 survival data points were used as input, with a significance threshold of P value <0.05. After screening, 49 genes associated with the prognosis were identified. The results of univariate Cox screening are summarized in table online: https://cdn.amegroups.cn/static/public/tcr-2025-1228-3.xls. These 49 genes were further subjected to feature selection using LASSO regression with the R package “glmnet”. A total of 16 genes were identified based on the lambda 1se value (Figure 3A,3B). The 16 genes were analyzed through multivariate Cox analysis. The results are presented in Table 2. After removing 8 non-significant genes with P values greater than 0.05, a total of 8 novel prognostic genes were identified and used for further analysis.

Figure 3 Construction of a PRGs prognostic model using the TCGA-HNSCC training set. (A,B) LASSO analysis identified 16 key DE-PRGs according to the lambda 1se value. (C) Forest plot illustrates the multivariate Cox analysis. (D) Risk scores for HNSCC patients are plotted in ascending order. (E) The relationship between survival time and increasing risk score. (F) Heat map showing the expression of eight signature genes between the high and low risk groups. (G) Kaplan-Meier curve showing the different survival probabilities between two risk groups. (H) ROC curves showing the predictive capacity of the risk score for 1-, 3-, and 5-year overall survival. Model accuracy was evaluated using the AUC. (I) Kaplan-Meier curves of single PRGs. AIC, Akaike Information Criterion; AUC, area under the curve; CI, confidence interval; DE-PRGs, differentially expressed palmitoylation-related genes; HNSCC, head and neck squamous cell carcinoma; HR, hazard ratio; LASSO, least absolute shrinkage and selection operator; PRGs, palmitoylation-related genes; ROC, receiver operating characteristics; TCGA, The Cancer Genome Atlas.

Constructing a model of PRG signature

We built a multivariate Cox model using the “coxph” function from the “survival” package based on 8 genes: CYP3A4, UNC13C, BRINP1, TIMP4, HCAR1, RSPO1, CYP4A11, and HTR2C. A forest plot demonstrates the robustness of the model (Figure 3C). Five of the eight genes (HTR2C, TIMP4, BRINP1, UNC13C, and CYP3A4) have a hazard ratio (HR) greater than 1, indicating an association with poor prognosis and suggesting that they may act as risk genes. Three genes (CYP4A11, RSPO1, and HCAR1) have an HR less than 1, suggesting that they may act as protective factors. The risk score was calculated for each patient in the TCGA-HNSCC cohort based on the coefficients and expression levels of these genes.

Patients were stratified into high- and low-risk groups based on the median risk score. The distribution of risk scores among patients is shown in Figure 3D, where the high-risk group exhibits higher risk scores. The relationship between survival time and increasing risk score is shown in Figure 3E. A higher risk score was observed among deceased patients, while those who survived had lower risk scores. The expression profiles of the 8 signature genes between high- and low-risk groups are shown in Figure 3F. The expression of protective factors (CYP4A11, RSPO1, and HCAR1) was higher in the low-risk group compared to the high-risk group. In contrast, the expression of risk genes (HTR2C, TIMP4, BRINP1, UNC13C, and CYP3A4) was significantly increased in the high-risk group.

Survival analysis using the Kaplan-Meier plot demonstrated that low-risk patients have significantly better survival probabilities (P value <0.001) (Figure 3G). We conducted a time-dependent ROC curve analysis to estimate the prognostic power of the model for 1-, 3-, or 5-year OS. The area under the curve (AUC) values of the ROC curve analysis were calculated to be 0.682, 0.697, and 0.666, respectively (Figure 3H). Additionally, the relationship between the expression levels of the 8 signature genes and survival probabilities is shown in Figure 3I.

Validation of the PRGs-based risk score in GSE41613 cohort

Prognostic model performance was validated in 97 patients of the GSE41613 microarray cohort. Patients were divided into high and low-risk groups based on the 8 signature genes of the multivariate Cox model. The risk score distribution, OS, and signature gene expression between risk groups are shown in Figure 4A-4C. The Kaplan-Meier curve shows higher survival probability for low-risk patients with a log-rank test P value of 0.01 (Figure 4D). The AUC for 1-, 3-, and 5-year was 0.618, 0657, and 0.678, respectively (Figure 4E).

Figure 4 Validation of the PRGs-based risk score in an independent external cohort, GSE41613. (A) Risk scores for HNSCC patients are plotted in ascending order. (B) The relationship between survival time and increasing risk score. (C) Heat map showing the expression of eight signature genes between high- and low-risk groups. (D) Kaplan-Meier curve showing the different survival probabilities between two risk groups. (E) ROC curves showing the predictive capacity of the risk score for 1-, 3-, and 5-year overall survival. AUC, area under the curve; HNSCC, head and neck squamous cell carcinoma; PRGs, palmitoylation-related genes; ROC, receiver operating characteristics.

Construction of the nomogram

We examined whether the PRGs-based prognostic model was independent of other clinical and pathological variables, including age, sex, stage, N-stage, and T-stage. Since there was no significant difference in tumor M-stage among all patients, M-stage was excluded from the analysis. Additionally, tumor grade data were not available for this analysis. A nomogram was constructed to integrate the risk score and other key clinical features, providing a quantitative tool for predicting patient prognosis and guiding clinical decision-making (Figure 5A). The AUC of the risk score was 0.709, outperforming all other clinicopathological parameters and demonstrating superior predictive performance (Figure 5B).

Figure 5 Independent prognostic performance of the risk score. (A) The nomogram integrates clinical features and the risk score to provide personalized survival predictions. (B) ROC curves comparing the prognostic accuracy of the risk score to that of other clinical parameters. Model accuracy was evaluated using the AUC. AUC, area under the curve; F, female; M, male; N, node; ROC, receiver operating characteristics; T, tumor.

Tumor immune environment of the two risk groups

Our results showed that the low-risk group exhibits significantly higher ImmuneScore and ESTIMATEScore. In contrast, there was no significant difference in the StromalScore between risk groups. The high-risk group has higher tumor purity (Figure 6A). Tumor purity is often inversely related to immune scores because a higher presence of non-tumor components, such as immune cells, reduces the overall proportion of tumor cells. These results suggest that the OS prognosis is positively correlated with immune components.

Figure 6 Tumor immune environments in two risk groups from the TCGA-HNSCC cohort. (A) Stromal, immune, ESTIMATE score, and tumor purity differences between the high- and low-risk groups using the ESTIMATE algorithm. (B) The discrepancy in the expression level of immune checkpoint genes between high- and low-risk groups. (C) Analysis of immune cell infiltration using the CIBERSORT algorithm. *, P<0.05; **, P<0.01; ***, P<0.001; ns, not statistically significant. HNSCC, head and neck squamous cell carcinoma; NK, natural killer; TCGA, The Cancer Genome Atlas.

Furthermore, our findings reveal that the low-risk group shows enhanced expression of immune checkpoint genes (ICGs) (Figure 6B). Immune cells such as B cells, plasma cells, CD8⁺ T cells, CD4⁺ memory T cells, follicular helper T cells, regulatory T cells (Tregs), resting NK cells, and M1 macrophages are more abundant in the low-risk group. In contrast, M0 macrophages and M2 macrophages are more prevalent in the high-risk group (Figure 6C).

GSEA

We performed GSEA on hallmark gene sets to compare pathway enrichment between the high- and low-risk groups in the TCGA-HNSCC dataset. The results indicate that pathways highly expressed in the high-risk group are associated with cancer aggressiveness and proliferation. These include angiogenesis, epithelial-mesenchymal transition (EMT), hedgehog signaling, hypoxia, myogenesis, and UV-response-DN (Figure 7A). Conversely, the low-risk group exhibited enrichment in pathways related to immune activation and cell cycle regulation, such as allograft rejection, E2F targets, G2M checkpoint, IL6-JAK-STAT3 signaling, interferon-alpha response, and interferon-gamma response (Figure 7B).

Figure 7 GSEA of two risk groups from the TCGA-HNSCC cohort. (A) Pathways enriched in the high-risk group. (B) Pathways enriched in the low-risk group. HNSCC, head and neck squamous cell carcinoma; GSEA, gene set enrichment analysis; TCGA, The Cancer Genome Atlas.

Risk score predicts therapeutic benefit in HNSCC

Palmitoyl acyltransferases (PATs) catalyze the transfer of palmitate to target proteins and are considered promising targets for drug development (32). Currently, no therapeutic drugs have been successfully developed to modulate the palmitoylation status of specific proteins. However, Fraser et al. have outlined potential strategies for selectively manipulating palmitoylation, which represents a critical initial step toward the development of clinically applicable molecules for disease treatment (33). We comprehensively compared the estimated IC₅₀ score of 198 chemotherapeutic agents or inhibitors in the GDSC database between the two risk groups. Among these, the IC₅₀ values of 16 drugs were significantly different and positively correlated with the risk score, indicating that the high-risk group was potentially insensitive to chemotherapy (Figure 8).

Figure 8 Risk score predicts drug therapeutic benefits in HNSCC. The high-risk group was potentially insensitive to chemotherapy. HNSCC, head and neck squamous cell carcinoma.

Supplementary materials

We provide instructions, input files and R code to reproduce all analyses in this work on GitHub at https://github.com/ji-group/HNSCC_and_Pam_supplementary.

Discussion

HNSCC is a highly aggressive malignancy driven by complex molecular mechanisms. Despite advances in surgery and chemotherapy, the prognosis remains poor, especially for patients with advanced stages (34,35). PTMs of proteins play a crucial role in cancer biology, with palmitoylation, a reversible lipid modification, emerging as a key regulatory mechanism in various cellular processes (36). Investigating the role of PRGs in HNSCC could improve prognostic accuracy and reveal novel therapeutic targets. Moreover, the rapid advancement of bioinformatics technologies now enables the systematic analysis of large-scale transcriptomic datasets, facilitating the identification of PRGs associated with cancer progression.

In this study, a PRG signature was developed using the TCGA dataset to predict the prognosis of HNSCC and was further validated in the independent external GEO dataset. Patients were stratified into high-risk and low-risk groups based on the median risk score. Kaplan-Meier curve analysis revealed that the low-risk group had significantly higher survival probabilities in both the training and validation datasets. Additionally, the palmitoylation-related risk model was confirmed as an independent prognostic factor after adjusting for other clinical variables, including age, gender, stage, N-stage, and T-stage. In conclusion, our findings demonstrate that the palmitoylation-related prognostic model effectively predicts the pathogenesis and progression of HNSCC and is strongly correlated with OS. However, the predictive performance of the 8-gene PRG-based model in HPV-positive versus HPV-negative HNSCC remains untested due to incomplete HPV annotation; future studies in well-characterized cohorts will be necessary to assess potential subtype-specific differences.

Previous studies have confirmed that some PRGs in our risk model play a role in HNSCC progression. UNC13C, which has the highest hazard ratio (1.125), is suggested to be a risk gene. However, its expression is significantly down-regulated in oral squamous cell carcinoma (OSCC, a subtype of HNSCC) tissues compared to adjacent normal tissues, and its reduced expression correlates with poor survival in OSCC patients (37). This apparent contradiction suggests that tissue-specific palmitoylation effects, tumor heterogeneity, or cohort differences may underlie these conflicting observations and warrant further experimental investigation. TIMP4 is also significantly reduced in both OSCC and HNSCC and is associated with poor prognosis (38). The remaining genes have not yet been explored in HNSCC, highlighting the need for further research to elucidate their precise molecular mechanisms. Although the 8 PRGs were selected from curated PRG sets, their palmitoylation status and functional impact in HNSCC remain to be experimentally validated; future studies using mass spectrometry and functional assays will be needed to confirm these mechanisms.

Palmitoylation plays a crucial role in regulating both innate and adaptive immunity, as well as modulating immune checkpoint pathways. This lipid modification influences the stability, localization, and function of key immune signaling molecules, thereby shaping immune responses and potentially impacting the effectiveness of immunotherapy (39). One of the key findings of this study is the distinct immune infiltration patterns observed between the high- and low-risk groups. Low-risk patients exhibited higher immune scores, greater infiltration of cytotoxic T cells, and elevated expression of immune checkpoint molecules, suggesting a more active immune surveillance in these individuals. This finding aligns with growing evidence that the tumor immune microenvironment plays a crucial role in shaping tumor progression and patient outcomes. Given these results, low-risk patients may be more responsive to immune checkpoint inhibitors, making immunotherapy a promising treatment strategy for this subgroup. However, these computational estimates remain subject to inherent limitations of the algorithm and should be interpreted cautiously.

Additionally, we utilized GSEA to explore enriched pathways and characteristics between the two risk groups. Notably, in the high-risk group, the enriched pathways strongly suggest that palmitoylation plays a key role in tumorigenesis, as evidenced by the following findings. First, it is well established that angiogenesis, EMT, and Hedgehog signaling are crucial for tumorigenesis (40). Recent studies have reported that enhancing the palmitoylation of Gpx1 inhibits angiogenesis (41). Furthermore, palmitoylated SCP1 is targeted to the plasma membrane, where it negatively regulates angiogenesis (42). Additionally, ZDHHC5-mediated palmitoylation of FAK promotes its membrane localization and drives EMT in glioma (43). Moreover, palmitoylation is essential for Hedgehog signaling, further reinforcing its role in cancer progression (44).

Although no Food and Drug Administration (FDA)-approved drugs currently target palmitoylation directly, the field is rapidly evolving, particularly in the development of agents targeting palmitoylation-dependent oncogenic pathways. In this study, we demonstrated that patients in the high-risk group exhibited overall reduced sensitivity to chemotherapeutic agents compared to those in the low-risk group. This finding is particularly intriguing and may provide valuable insights for the design of future preclinical trials aimed at exploring palmitoylation-related therapeutic strategies (33). However, drug sensitivity predictions were based on cell line IC₅₀ data using “oncoPredict” and may not fully reflect clinical responses; experimental validation in patient-derived models or clinical cohorts is warranted.

Several limitations must be acknowledged. First, this study relied on publicly available transcriptomic and clinical datasets (TCGA and GEO) without experimental validation; in vitro and in vivo assays will be essential to confirm the causal role and palmitoylation status of the identified PRGs. Second, key clinical variables such as HPV status and tumor grade were not incorporated due to incomplete annotation, underscoring the need for validation in larger, well-annotated cohorts. Third, although the TCGA cohort includes survival data extending to ~6,000 days, the number of patients with long-term follow-up is limited, which may restrict the model’s predictive accuracy for very late events. Finally, although the predictive accuracy of our model was moderate (AUC 0.6–0.7), this is comparable to other HNSCC prognostic signatures (45-47) and highlights the potential value of incorporating palmitoylation-related biology; integration with additional molecular and clinical features may further enhance predictive power.

Conclusions

We developed a PRGs-based prognostic model using the TCGA-HNSCC cohort and validated its predictive performance in an independent external cohort, GSE41613. Furthermore, our study provides new insights into the role of palmitoylation in the progression of HNSCC. Future research aimed at developing palmitoylation-targeting therapies and integrating them with existing treatment strategies holds great promise for improving clinical outcomes in HNSCC patients.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1228/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1228/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1228/coif). The authors have no conflicts of interest to declare.

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.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Jin Q, Qi D, Zhang M, et al. CLDN6 inhibits breast cancer growth and metastasis through SREBP1-mediated RAS palmitoylation. Cell Mol Biol Lett 2024;29:112. [Crossref] [PubMed]
2. Tang J, Peng W, Feng Y, et al. Cancer cells escape p53's tumor suppression through ablation of ZDHHC1-mediated p53 palmitoylation. Oncogene 2021;40:5416-26. [Crossref] [PubMed]
3. Li M, Zhang L, Chen CW. Diverse Roles of Protein Palmitoylation in Cancer Progression, Immunity, Stemness, and Beyond. Cells 2023;12:2209. [Crossref] [PubMed]
4. Zhou B, Hao Q, Liang Y, et al. Protein palmitoylation in cancer: molecular functions and therapeutic potential. Mol Oncol 2023;17:3-26. [Crossref] [PubMed]
5. Lin J, Lyu Z, Feng H, et al. CircPDIA3/miR-449a/XBP1 feedback loop curbs pyroptosis by inhibiting palmitoylation of the GSDME-C domain to induce chemoresistance of colorectal cancer. Drug Resist Updat 2024;76:101097. [Crossref] [PubMed]
6. Wang WK, Lin HY, Lin CH, et al. GRK6 palmitoylation dictates triple-negative breast cancer metastasis via recruiting the β-Arrestin 2/MAPKs/NF-κB signaling axis. Breast Cancer Res 2024;26:193. [Crossref] [PubMed]
7. Zhu Z, Feng S, Zeng A, et al. Advances in Palmitoylation: A key Regulator of liver cancer development and therapeutic targets. Biochem Pharmacol 2025;234:116810. [Crossref] [PubMed]
8. Zhu H, Hu H, Hao B, et al. Insights into a Machine Learning-Based Palmitoylation-Related Gene Model for Predicting the Prognosis and Treatment Response of Breast Cancer Patients. Technol Cancer Res Treat 2024;23:15330338241263434. [Crossref] [PubMed]
9. Zhang Q, Du Z, Zhou W, et al. ZDHHC1 downregulates LIPG and inhibits colorectal cancer growth via IGF2BP1 Palmitoylation. Cancer Gene Ther 2024;31:1427-37. [Crossref] [PubMed]
10. Zhu J, Cao X, Chen Z, et al. Inhibiting S-palmitoylation arrests metastasis by relocating Rap2b from plasma membrane in colorectal cancer. Cell Death Dis 2024;15:675. [Crossref] [PubMed]
11. Li W, Liu J, Yu T, et al. ZDHHC9-mediated Bip/GRP78 S-palmitoylation inhibits unfolded protein response and promotes bladder cancer progression. Cancer Lett 2024;598:217118. [Crossref] [PubMed]
12. Zhang H, Sun Y, Wang Z, et al. ZDHHC20-mediated S-palmitoylation of YTHDF3 stabilizes MYC mRNA to promote pancreatic cancer progression. Nat Commun 2024;15:4642. [Crossref] [PubMed]
13. Zeng S, Zeng L, Xie X, et al. Palmitoylation-related gene expression and its prognostic value in ovarian cancer: insights into immune infiltration and therapeutic potential. Discov Oncol 2024;15:802. [Crossref] [PubMed]
14. Ni Q, Pan C, Han G. Modification-specific Proteomic Analysis Reveals Cysteine S-Palmitoylation Involved in Esophageal Cancer Cell Radiation. ACS Omega 2025;10:1541-50. [Crossref] [PubMed]
15. Lin SC, Cheng YS, Lin YS, et al. The long noncoding RNA lncZBTB10 facilitates AR function via S-palmitoylation to promote prostate cancer progression and abiraterone resistance. Br J Cancer 2025;132:587-98. [Crossref] [PubMed]
16. Xin L, Xu HS, Fan LJ, et al. Methionine Restriction Exerts Anti-Tumor Immunity via Joint Intervention of T-Bet Palmitoylation in Gastric Cancer. Biotechnol J 2025;20:e202400574. [Crossref] [PubMed]
17. Barsouk A, Aluru JS, Rawla P, et al. Epidemiology, Risk Factors, and Prevention of Head and Neck Squamous Cell Carcinoma. Med Sci (Basel) 2023;11:42. [Crossref] [PubMed]
18. Sun Z, Sun X, Chen Z, et al. Head and Neck Squamous Cell Carcinoma: Risk Factors, Molecular Alterations, Immunology and Peptide Vaccines. Int J Pept Res Ther 2022;28:19. [Crossref] [PubMed]
19. de Bakker T, Journe F, Descamps G, et al. Restoring p53 Function in Head and Neck Squamous Cell Carcinoma to Improve Treatments. Front Oncol 2021;11:799993. [Crossref] [PubMed]
20. Jiang Q, Xiao J, Hsieh YC, et al. The Role of the PI3K/Akt/mTOR Axis in Head and Neck Squamous Cell Carcinoma. Biomedicines 2024;12:1610. [Crossref] [PubMed]
21. Shah PA, Huang C, Li Q, et al. NOTCH1 Signaling in Head and Neck Squamous Cell Carcinoma. Cells 2020;9:2677. [Crossref] [PubMed]
22. Nair S, Bonner JA, Bredel M. EGFR Mutations in Head and Neck Squamous Cell Carcinoma. Int J Mol Sci 2022;23:3818. [Crossref] [PubMed]
23. Huang Q, Xiao Y, Lan T, et al. WNT7A promotes tumorigenesis of head and neck squamous cell carcinoma via activating FZD7/JAK1/STAT3 signaling. Int J Oral Sci 2024;16:7. [Crossref] [PubMed]
24. Uzelac M, Barakchi A, Beldona V, et al. Effects of Human Papilloma Virus E6/E7 Oncoproteins on Genomic Structure in Head and Neck Squamous Cell Carcinoma. Cancers (Basel) 2022;14:6190. [Crossref] [PubMed]
25. Li W, Yu F, Wang M, et al. Development of a quantitative genomic instability scoring system and a related competing endogenous RNA network in head and neck squamous cell carcinoma. Transl Cancer Res 2025;14:4115-41. [Crossref] [PubMed]
26. Qin G, Pang G, Wu S, et al. Construction and validation of a novel prognostic model with palmitoylation-related genes for glioblastoma. Transl Cancer Res 2024;13:6117-35. [Crossref] [PubMed]
27. Feng R, Cheng D, Chen X, et al. Identification and validation of palmitoylation metabolism-related signature for liver hepatocellular carcinoma. Biochem Biophys Res Commun 2024;692:149325. [Crossref] [PubMed]
28. Wei Z, Wang G, Hu Y, et al. Development and validation of an immune-related gene signature for the prognostic and immune landscape prediction in head and neck squamous cell carcinoma by integrated analysis of machine learning and Mendelian randomization. Transl Cancer Res 2025;14:4520-38. [Crossref] [PubMed]
29. Yuan S, Zhai Z, Wang Y, et al. Construction and validation of an anoikis-related long non-coding RNA-based prognostic model for head and neck squamous cell carcinoma. Transl Cancer Res 2025;14:4160-78. [Crossref] [PubMed]
30. Xue L, Jahng WJ, Gollapalli D, et al. Palmitoyl transferase activity of lecithin retinol acyl transferase. Biochemistry 2006;45:10710-8. [Crossref] [PubMed]
31. Marino M, Ascenzi P. Steroid hormone rapid signaling: the pivotal role of S-palmitoylation. IUBMB Life 2006;58:716-9. [Crossref] [PubMed]
32. Chavda B, Arnott JA, Planey SL. Targeting protein palmitoylation: selective inhibitors and implications in disease. Expert Opin Drug Discov 2014;9:1005-19. [Crossref] [PubMed]
33. Fraser NJ, Howie J, Wypijewski KJ, et al. Therapeutic targeting of protein S-acylation for the treatment of disease. Biochem Soc Trans 2020;48:281-90. [Crossref] [PubMed]
34. Hoesseini A, Sewnaik A, van den Besselaar BN, et al. Prognostic model for overall survival of head and neck cancer patients in the palliative phase. BMC Palliat Care 2024;23:54. [Crossref] [PubMed]
35. Qiu X, Lin Y, Li Q, et al. Integrated single-cell and bulk RNA sequencing reveals prognostic and immunotherapy-associated myofibroblastic cancer-associated fibroblast subtypes in head and neck squamous cell carcinoma. Transl Cancer Res 2025;14:3730-45. [Crossref] [PubMed]
36. Kong Y, Liu Y, Li X, et al. Palmitoylation landscapes across human cancers reveal a role of palmitoylation in tumorigenesis. J Transl Med 2023;21:826. [Crossref] [PubMed]
37. Velmurugan BK, Yeh KT, Hsieh MJ, et al. UNC13C suppress tumor progression via inhibiting EMT pathway and improves survival in oral squamous cell carcinoma. Front Oncol 2019;9:728. [Crossref] [PubMed]
38. Zhou L, Tang C, Shuai R, et al. TIMP4 serves as a novel potential prognostic biomarker for oral squamous cell carcinoma. Sci Rep 2025;15:6313. [Crossref] [PubMed]
39. Chen Y, Li Y, Wu L. Protein S-palmitoylation modification: implications in tumor and tumor immune microenvironment. Front Immunol 2024;15:1337478. [Crossref] [PubMed]
40. Liu ZL, Chen HH, Zheng LL, et al. Angiogenic signaling pathways and anti-angiogenic therapy for cancer. Signal Transduct Target Ther 2023;8:198. [Crossref] [PubMed]
41. Ma Y, Yuan X, Wei A, et al. Enhancing Gpx1 palmitoylation to inhibit angiogenesis by targeting PPT1. Redox Biol 2024;77:103376. [Crossref] [PubMed]
42. Liao P, Wang W, Li Y, et al. Palmitoylated SCP1 is targeted to the plasma membrane and negatively regulates angiogenesis. Elife 2017;6:e22058. [Crossref] [PubMed]
43. Wang Y, Shen N, Yang Y, et al. ZDHHC5-mediated S-palmitoylation of FAK promotes its membrane localization and epithelial-mesenchymal transition in glioma. Cell Commun Signal 2024;22:46. [Crossref] [PubMed]
44. Resh MD. Palmitoylation of Hedgehog proteins by Hedgehog acyltransferase: roles in signalling and disease. Open Biol 2021;11:200414. [Crossref] [PubMed]
45. Wu ZH, Tang Y, Zhou Y. A Metabolic Gene Signature to Predict Overall Survival in Head and Neck Squamous Cell Carcinoma. Mediators Inflamm 2020;2020:6716908. [Crossref] [PubMed]
46. Shen S, Bai J, Wei Y, et al. A seven-gene prognostic signature for rapid determination of head and neck squamous cell carcinoma survival. Oncol Rep 2017;38:3403-11. [Crossref] [PubMed]
47. Wang X, Wang Z. A novel basement membrane-related gene signature for predicting prognosis of HNSCC. Medicine (Baltimore) 2025;104:e41316. [Crossref] [PubMed]