Skip to main content
NCBI home page
Scientific Reports logoLink to Scientific Reports
. 2024 Oct 25;14:25338. doi: 10.1038/s41598-024-76866-0

Identification and validation of M2 macrophage-related gene signature as a novel prognostic model for head and neck squamous cell carcinoma

Shengmei He 1,2, Huarong Chen 1,2, Changya Li 1,2, Bao Feng 2,✉, Ruizhe Zhang 2, Houyu Zhao 1,2,✉, Xianlu Zhuo 1,2,✉
PMCID: PMC11512021  PMID: 39455885

Abstract

Head and neck squamous cell carcinoma (HNSCC) is a highly heterogeneous tumor. Commonly used tumor staging don’t sufficiently and accurately assess the prognosis of HNSCC patients, resulting in a lack of guidance for clinical treatment decisions. M2 macrophage infiltration has been shown to be strongly associated with the tumor prognosis. In this study, we used the Cancer Genome Atlas (TCGA) data to screen for genes co-expressed with M2 macrophages in HNSCC. We used univariate Cox regression to screen out the genes associated with HNSCC prognosis, and constructed a HNSCC prognosis model by Lasso regression analysis. The results confirmed that the model had good predictive value and accuracy for the prognosis of HNSCC patients by survival analysis, ROC curve and nomogram. We divided the HNSCC samples into high-risk and low-risk groups according to the risk score, and the results showed that patients in the high-risk group were more prone to genetic mutations and had a higher tumor mutational burden. In addition, there were significant differences between risk groups in terms of immune cell infiltration and drug sensitivity. The HNSCC prognostic model established in this study may provide guidance for clinical therapeutic decision-making and provide a theoretical foundation for the development of new immunotherapy methods.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-76866-0.

Keywords: M2 macrophage, Head and neck squamous cell carcinoma, Immune infiltration, Prognostic model

Subject terms: Cancer prevention, Head and neck cancer

Introduction

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common type of cancer worldwide, originating primarily in the squamous epithelium of the lip, oral cavity, nasal cavity, pharynx, and larynx1. Epidemiological studies suggest that HNSCC is responsible for over 310,000 deaths globally each year, and its incidence is increasing. It is predicted that the global incidence of HNSCC will increase by 30% by 2030, resulting in up to 1.08 million new cases annually2,3. The site of HNSCC is mainly concentrated in the narrow lumen of the head and neck, where muscles, blood vessels and nerves are abundant and concentrated, and organs are interconnected and anatomically complex; in addition to this, HNSCC is highly invasive and heterogeneous, and the effectiveness of standardised oncological therapies is often limited by the heterogeneity among tumors1,4. Consequently, it is difficult to achieve a complete cure with surgery, radiotherapy and chemotherapy for HNSCC. In addition to physiological and anatomical factors, the prognosis of patients with HNSCC is also closely related to the tumor, node, metastasis (TNM) staging5. In clinical practice, TNM staging is often used to guide treatment plans and assess patient prognosis. However, it has been demonstrated that patients with the same TNM staging may show disparate survival outcomes6. This indicates that there are certain limitations to the use of the clinicopathologic staging in the assessment of patient prognosis. Consequently, the identification of novel prognostic markers to assess the prognosis of HNSCC patients is of paramount importance, as this will facilitate the guidance of clinical treatment decisions.

Immune cells play a supervisory and regulatory role in tumor development and prognosis7–9. Macrophages are the main immune cells in the tumor microenvironment10 and are mainly classified into classically activated macrophages (M1 macrophages) and alternatively activated macrophages (M2 macrophages) based on their type of activation and different roles in the tumor microenvironment11. Studies have shown that M2 macrophages can lead to immune escape of tumor cells by synthesizing and releasing a large number of anti-inflammatory factors and immune factors, thereby promoting tumor progression12. M2 macrophages can also exert tumor drug resistance and promote tumor proliferation by secreting a variety of cytokines that mediate cell proliferation and migration and prevent apoptosis13,14. Furthermore, M2 macrophages are involved in the generation of tumor vasculature and lymphatic vasculature, as well as the production of extracellular matrix-degrading enzymes, which collectively alter the tumor microenvironment to regulate a variety of biological behaviors of the tumor, including tumor recurrence, metastasis, and drug resistance13,15–17. These evidences indicate that high levels of M2 macrophage infiltration are associated with a poor prognosis in patients with HNSCC. Currently, research on the role of M2 macrophage-related genes (M2RGs) in HNSCC prognosis is limited. Therefore, in-depth analyzes of the association between M2RGs and HNSCC prognosis may reveal valuable markers for prognostic assessment.

In this study, we developed a prognostic model based on M2RGs and fully evaluated the accuracy and predictive ability of this prognostic model. Furthermore, the relationship between the prognostic model and a number of other factors was evaluated, including immune characteristics, mutation characteristics, and drug sensitivity. These findings provide a theoretical foundation for guiding clinical therapeutic decision-making and the development of new immunotherapy techniques.

Results

Identification of M2RGs in HNSCC

Figure 1 presents research design for this study. In order to identify M2RGs in HNSCC, a total of 6666 co-expressed genes associated with M2 macrophage were screened based on the sequencing results of HNSCC samples in the Cancer Genome Atlas (TCGA) database and M2 macrophage infiltration in CIBERSORTx (screening conditions: P < 0.05 and |R| ≥ 0.1, Supplementary Fig. 1).

Fig. 1.

Fig. 1

Open in a new tab

Schematic layout of this study.

Correlation and functional enrichment analysis of M2RGs

To further analyze the correlation between these co-expressed genes and M2 macrophage infiltration, we performed correlation analysis on the top 100 co-expressed genes previously screened, which showed that 97 genes were positively correlated and 3 genes were negatively correlated with M2 macrophage infiltration (Fig. 2A). Then, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were performed on these genes, and the results of GO enrichment analysis showed that the M2RGs were mainly involved in biological processes such as migration and activation of myeloid leukocytes, regulation of macrophage migration, migration of monocytes, etc.; and that the molecular functions were mainly related to IgG binding, complement receptor activity, immunoreceptor activity and cargo receptor activity (Fig. 2B-C). KEGG enrichment analysis showed that M2RGs were mainly enriched in signalling pathways such as extracellular matrix receptor interactions, proteoglycans in cancer, transcriptional misregulation in cancer and complement and coagulation cascades (Fig. 2D). The results of the above enrichment analysis suggest that M2RGs in HNSCC samples may influence M2 macrophage infiltration through the biological functions and signalling pathways described above.

Fig. 2.

Fig. 2

Open in a new tab

Correlation and functional enrichment analysis of the top 100 co-expressed genes associated withM2 macrophage infiltration. (a) Correlation analysis of the top 100 co-expressed genes associated with M2 macrophage infiltration; (b) Bubble plot of GO analysis of the top 100 co-expressed genes associated with M2 macrophage infiltration; (c) Circle plot of GO analysis of the top 100 co-expressed genes associated with M2 macrophage infiltration; (d) Sankey of KEGG analysis of the top 100 co-expressed genes associated with M2 macrophage infiltration.

Construction of an HNSCC prognostic model

To improve the accuracy of predicting prognostic outcomes for HNSCC patients, we developed a prognostic model based on M2RGs. Firstly, univariate Cox regression analysis was used to explore the relationship between the expression levels of M2RGs and HNSCC overall survival (OS), and 11 potential predictive genes associated with HNSCC prognosis were screened according to the screening conditions (Cox P < 0.05, |HR| ≥ 1.75, Fig. 3A). These genes were then analyzed by Lasso regression and cross-validated to identify the nine most valuable predictor genes, involving LIMA1, RDH11, CCDC86, STC2, PRDX6, EFNB2, PCTP, AIG1, and SURF4, which were ultimately used in the construction of the prognostic model (Fig. 3B-C). Then, based on the TCGA database and Human Protein Atlas (HPA) database, the mRNA and protein expression levels of the model-constructing genes were analyzed in HNSCC tissues and normal tissues, and the results showed that the mRNA and protein expression levels of the model-constructing genes were significantly and differentially expressed in HNSCC tissues and normal tissues (P<0.05, Fig. 3D-J, Supplementary Fig. 2). Finally, we analyzed the relationship between model-constructing genes and clinicopathologic factors through the LinkedOmics database and the UALCAN database, which showed that model-constructing genes were associated with factors such as age, pathological stage and TP53 mutation status of the patients (P<0.05, Supplementary Tables 1–9, Supplementary Figs. 3–11).

Fig. 3.

Fig. 3

Open in a new tab

Construction of M2RGs prognostic model for HNSCC patients and immunohistochemical results of the genes used to construct the prognostic model. a. Univariate cox analysis identified 11 prognostic genes in the illustration of Forest plot; b-c. The results of Lasso regression algorithm. lambda curves and Cvfit showing the process through which we calculated the minimum cross-validation error point, and the corresponding nonzero coefficients for prognostic risk model construction; d-j. Immunohistochemical analysis results of model constructed gene in HPA database.

Validation of the HNSCC prognostic model

Subsequently, the samples in the TCGA-HNSCC cohort were then categorized into high-risk and low-risk groups based on the median value of the risk score (Fig. 4A). The risk heat map showed that the model-constructing genes were significantly and differentially expressed between the high- and low-risk groups (Fig. 4B), and the survival time distribution showed that the high-risk group had a worse prognosis (Fig. 4C). The results of the Kaplan-Meier (KM) curves showed that the prognosis of the high-risk group was significantly worse than that of the low-risk group (P < 0.001, Fig. 4D). Furthermore, the areas under the receiver operating characteristic (ROC) curves (AUC) of the risk scores for 1, 3, and 5 years were 0.68, 0.73, and 0.75, respectively (Fig. 4E). Similarly, the predictive value and accuracy of this prognostic model were validated in the Gene Expression Omnibus (GEO) database with similar results. (Figs. 4F-G).

Fig. 4.

Fig. 4

Open in a new tab

Validation of the M2RGs prognostic model. (a) Risk plot showing the distribution of the risk score in the training dataset; (b) Risk heat map of genes used to build model; (c) The result of survival time distribution in the training dataset; (d) Kaplan-Meier curves of OS status in the training dataset; (e) 1-year, 3-year, and 5-year ROC curves of risk model for predicting OS status in the training dataset; (f) Kaplan-Meier curves of OS status in the validation dataset; (g) 1-year, 3-year, and 5-year ROC curves of risk model for predicting OS status in the validation dataset.

Independent prognostic analysis of risk scores and clinicopathologic factors

Univariate and multivariate Cox regression analyses of risk scores and clinicopathologic factors in the TCGA-HNSCC cohort showed that age, pathological staging and risk scores were independent prognostic factors for HNSCC (Fig. 5A-B). In order to ascertain whether the independent prognostic role of risk scores was superior to that of other clinicopathologic factors, we compared the AUC for risk scores, age, and pathological stage. The results showed that the AUC was greater for the risk score than for age and pathological stage (Fig. 5C-E). The results of the above analysis suggest that the risk score can be used as an independent prognostic factor for HNSCC and that its independent prognostic role is superior to that of the patient’s age and pathological stage.

Fig. 5.

Fig. 5

Open in a new tab

Independent prognostic analysis of risk score and clinicopathological factors. (a) The results of univariate analysis of risk score; (b) The results of multivariate analysis of risk score; (c) The area under the 1-year ROC curve for the risk score in relation to age and pathological stage; (d) The area under the 3-year ROC curve for the risk score in relation to age and pathological stage; (e) The area under the 5-year ROC curve for the risk score in relation to age and pathological stage.

The predictive role of risk scores in the prognosis of HNSCC

To assess the role of risk scores in predicting the prognosis of HNSCC patients, we constructed a nomogram for predicting OS in HNSCC patients by combining clinical information and risk scores from the TCGA-HNSCC cohort (Fig. 6A), the results showed that age, pathological stage, and risk scores were significant prognostic factors. To further validate the accuracy of the constructed nomogram, we plotted the KM curve, time-dependent ROC and the nomogram calibration curve for different risk groups according to the nomogram risk score. The results showed significant differences in survival outcomes between different risk groups (P < 0.001, Fig. 6B). Additionally, the AUC predicting the patients’ prognosis for 1 year, 3 years and 5 years were 0.70, 0.72 and 0.73, respectively (Fig. 6C). And the calibration curve of the nomogram was very close to the ideal diagonal (Fig. 6D), these results show that the nomogram has good accuracy in predicting prognosis.

Fig. 6.

Fig. 6

Open in a new tab

Predictive role of risk score on HNSCC prognosis. (a) Nomogram for predicting 1-year, 3-year, and 5-year probabilities of OS outcome; (b) The Kaplan-Meier curves for predicting the nomogram of 1-year, 3-year, and 5-year OS in the TCGA-HNSCC cohort; (c) The ROC curves for predicting the nomogram of 1-year, 3-year, and 5-year OS in the TCGA-HNSCC cohort; (d) The nomogram calibration curves for predicting the 1-year, 3-year, and 5-year OS of the TCGA-HNSCC cohort; (e) Nomogram for predicting 1-year, 3-year, and 5-year probabilities of PFS outcome; (f) The nomogram calibration curves for predicting the 1-year, 3-year, and 5-year PFS of the TCGA-HNSCC cohort; (g) The Kaplan-Meier curves for predicting the nomogram of 1-year, 3-year, and 5-year PFS in the TCGA-HNSCC cohort; (h) Results of a joint survival analysis of risk score and age; (i) Results of a joint survival analysis of risk score and stage. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

In addition, we also constructed a nomogram of progression-free survival (PFS) and plotted the KM curve and the nomogram calibration curve of this nomogram, the results showed that the nomogram of PFS also had a good predictive value for the prognosis of patients with HNSCC (Fig. 6E-G). Finally, a joint survival analysis was performed of risk scores and clinical factors with a prognostic role. This analysis showed that patients in the high-risk group with an age greater than 65 years and high-risk patients with later pathological staging had the worst prognosis, whereas low-risk patients with an age of less than or equal to 65 years and low-risk patients with earlier pathological staging had the best prognosis (Fig. 6H-I). The results of the survival analysis described above suggest that the combination of patients’ age, pathological stage and risk score provides a more precise estimation of patients’ prognosis.

Correlation analysis of risk score and immunity

Based on the risk score, we further compared the infiltration of immune cells in the high-risk and low-risk groups. The results demonstrated statistically significant differences in the infiltration of B cells naïve, plasma cells, T cells CD8, T cells CD4 naive, T cells CD4 memory resting, T cells CD4 memory activated, T cells follicular helper, T cells regulatory, macrophage M0, macrophage M2, dendritic cells resting, mast cells resting, mast cells activated and neutrophils between the two groups (Fig. 7A). Next, we conducted a correlation analysis between the risk score and immune cell infiltration and immune checkpoint molecules. The results demonstrated that macrophage M0, macrophage M2, mast cells activated and T cells CD4 memory resting were positively correlated with the risk score, whereas T cells regulatory, T cells follicular helper, T cells CD8, plasma cells, mast cells resting and B cells naïve were negatively associated with the risk score (Fig. 7B). Figure 7C illustrates the correlation between immune checkpoint and the risk score. It can be observed that ATIC and OLA1 were positively correlated with the risk score, whereas CTLA4, PDCD1 and IDO1 were negatively correlated with the risk score. Subsequently, gene set enrichment analysis (GSEA) enrichment analysis of different risk groups showed that Kegg glycosaminoglycan biosyn thesis chondroitin sulfate, Kegg basal cell carcinoma, Kegg focal adhesion, Kegg pathways in cancer and Kegg pentose and glucuronate interconversions were mainly enriched in high-risk group (Fig. 7D), Kegg intestinal immune network for iga production, Kegg allograft rejection, Kegg asthma, Kegg autoimmune thyroid disease and Kegg primary immunodeficiency were mainly enriched in the low-risk group (Fig. 7E). Overall, there was a correlation between the risk score and immune responses in HNSCC patients, and the correlation between them may be influenced by the signalling pathways described above.

Fig. 7.

Fig. 7

Open in a new tab

Correlation analysis of risk score and immunity. (a) Violin plots concerning the differences in abundance of immune cells between the two risk groups; (b) Correlation between expression of immune cells and risk scores in the TCGA-HNSCC cohort. The color of each unit represents the correlation degree. *, P < 0.05; **, P < 0.01; ***, P < 0.001. (c) Correlation between expression of checkpoints and risk scores in the TCGA-HNSCC cohort. (d) GSEA analysis in high-risk group; (e) GSEA analysis in low-risk group.

Analysis of gene mutations and drug susceptibility in different risk groups

Gene-targeted therapy of cancer represents a novel approach to cancer treatment that may prove to be a valuable clinical tool in the future18. Therefore, we compared the frequency of gene mutations between different risk groups. The results showed that TP53 (76%), TTN (38%), FAT1 (24%), CDKN2A (21%) and MUC16C (18%) were the five genes most frequently mutated in the high-risk group (Fig. 8A), while TP53 (54%), TTN (35%), FAT1 (18%), CSMD3 (18%) and SYNE1 (18%) were the five most frequently mutated genes in the low-risk group (Fig. 8B). We also analyzed the tumor mutation burden (TMB) between different risk groups. It was found that there was a statistically significant difference in TMB between different risk groups (P < 0.01, Fig. 8C). Subsequently, we conducted drug sensitivity analyses of anticancer drugs and found statistically significant differences in the sensitivity of 138 drugs, including cisplatin, cytarabine, gefitinib, paclitaxel and gemcitabine, among different risk groups. Among them, 36 anticancer drugs were more sensitive in the high-risk group and 102 anticancer drugs were more sensitive in the low-risk group. Detailed results are presented in Supplementary Table 10. The analysis of drug sensitivity in different risk groups can facilitate the development of personalised HNSCC treatment regimens in the clinic.

Fig. 8.

Fig. 8

Open in a new tab

Analysis of gene mutations and drug susceptibility in different risk groups. (a) Waterfall plot of gene mutation frequency in the high-risk group; (b) Waterfall plot of gene mutation frequency in the low-risk group; (c) Violin plot showing the difference in TMB between the two risk groups; d-h. Box plots showing the difference in drug sensitivity between the two risk groups(d. cisplatin; e. cytarabine; f. gefitinib; g. gemcitabine; h. paclitaxel).

Discussion

In recent years, more and more studies have revealed that immune cell infiltration in the tumor microenvironment plays a key role in tumor treatment response and patients’ prognosis19,20. M2 macrophages represent a crucial component of the tumor microenvironment, their high infiltration is closely associated with a poor prognosis in a variety of malignant tumors21,22. It has been reported that M2 macrophages are involved in regulating a variety of biological behaviors, including the occurrence, development, recurrence, metastasis and drug resistance of HNSCC23–25, and this suggests that M2 macrophages may be a potential tumor therapeutic target and prognostic indicator. In this study, a novel HNSCC prognostic model was developed based on M2RGs. Following comprehensive analysis and verification, it was demonstrated that the model exhibited excellent predictive value, accuracy, superiority, and clinical guidance.

In this study, based on the infiltration of M2 macrophages in HNSCC and the sequencing results of HNSCC in the TCGA database, we screened the genes with a co-expression relationship with M2 macrophage infiltration. Subsequently, we analyzed the effect of these co-expression genes on the prognosis of HNSCC patients and the nine most predictive genes (LIMA1, RDH11, CCDC86, STC2, PRDX6, EFNB2, PCTP, AIG1, and SURF4) were identified for the construction of prognostic model. In the constructed prognostic model, the results of survival analysis and time-dependent ROC curves of TCGA and GEO data confirmed that there was a significant difference in the prognosis of different risk groups, and the model exhibited good predictive value and accuracy for the prognosis of HNSCC patients. Similarly, this finding was fully assessed and supported in the nomogram model constructed on the basis of risk scores and clinicopathologic factors. In addition, the risk score was confirmed to be an independent prognostic factor for HNSCC prognosis by univariate and multivariate Cox regression analysis. Its independent prognostic role was found to be superior to that of the patient’s age and pathological stage. These results indicate that our newly constructed HNSCC prognostic model is reliable. Based on the risk score can effectively distinguish high-risk patients from low-risk patients. However, it is well known that human papilloma virus (HPV) infection status is an important factor in the pathogenesis of HNSCC and also plays a crucial role in the prognosis of HNSCC, but we did not stratify the HPV infection status in the OS and PFS outcomes, mainly because the HPV status was not counted in the training set used to construct the model, and previous studies have shown that some patients with HPV-associated HNSCC have a better prognosis. Therefore, it is necessary to refine the effect of different HPV infection status on OS and PFS outcomes in the future.

Among the M2RGs used to construct the prognostic model. LIMA1 has been documented as a key gene in the occurrence and metastasis of some solid tumors. It promotes tumor progression and metastasis by regulating actin dynamics and influencing cell migration26,27. Not only that, LIMA1 has been identified as a highly expressed gene in HNSCC, and its elevated expression level is closely associated with a poor prognosis for HNSCC27. CCDC86 has the function of promoting epithelial mesenchymal transition and up-regulating matrix metalloproteinase expression. It is an important gene affecting the proliferation, invasion and migration of nasopharyngeal carcinoma28. STC2 has been shown to have pro-oncogenic effects in numerous cancers. Furthermore, studies pertaining to HNSCC have indicated that STC2 can regulate the distant metastasis of HNSCC through the PI3K/AKT/Snail signalling axis29. PRDX6 has been demonstrated to be highly expressed in tongue squamous cell carcinoma tissues, and the gene is strongly associated with cancer development30. Similarly, EFNB2 is found to be abnormally highly expressed in HNSCC and is associated with malignant progression of the tumor. Blocking EFNB2 can inhibit tumor proliferation and improve patient survival rate31,32. As a transporter, the cargo protein SURF4 regulates the binding of SHH to the macrophage surface receptor PTCH1 through an extracellular transport mechanism, and then mediates the polarization of macrophages towards M2 macrophages, thus promoting tumor progression33–35. However, there is a paucity of literature reporting on the association between PCTP, AIG1, and RDH11 and the biological behavior of HNSCC. In this study, our univariate Cox regression analysis showed that these model-constructing genes were associated with OS and PFS of HNSCC patients. In view of this, it is reasonable to speculate that M2RGs utilized in model construction are associated with the biological behavior and prognosis of HNSCC. This hypothesis may provide a promising research direction for the development and prognosis of HNSCC.

Immunotherapy is currently a hot spot in the field of tumor research, which mainly plays an anti-tumor role by activating or enhancing the function of the immune cells36. Among these, blockade of immune checkpoints is regarded as one of the most efficacious immunotherapeutic approaches37. In this study, differences in immune cell infiltration and correlation with immune checkpoints were analyzed to obtain differentially infiltrated immune cells between risk groups, and the immune checkpoint molecules ATIC and OLA1 were found to be positively correlated with the risk score, while CTLA4, PDCD1, and IDO1 were found to be negatively correlated with the risk score. This is expected to provide molecular targets for immunotherapy. On the other hand, genetic mutation is considered to be one of the major drivers of cancer development38. The number of mutations occurring in tumor cells is known as the TMB, which to some extent predicts the outcome of immunotherapy in patients39. The results of our analysis showed that there were statistically significant differences in both mutation frequency and TMB between the high and low risk groups. This suggests that the risk score may be useful in the initial assessment and judgement of a patient’s mutation frequency and TMB, and thus in predicting the efficacy of a patient’s immunotherapy. Although immunotherapy has demonstrated efficacy in the treatment of tumors, surgery in conjunction with radiotherapy and chemotherapy remains the primary treatment modality for patients with HNSCC in clinical practice1. We analyzed the sensitivity of cancer treatment drugs in different risk groups, the results showed that the sensitivity of multiple cancer treatment drugs exhibited significant differences between different risk groups. This included the commonly used chemotherapy drugs in the clinical treatment of HNSCC, namely cisplatin, docetaxel and paclitaxel40–42. Sensitivity analyses of cancer therapeutics in different risk groups can help clinicians to make an initial determination of which patients are more likely to benefit from cancer therapeutics, which in turn can help to develop personalised HNSCC treatment regimens in the clinic.

In summary, we developed a novel prognostic model for HNSCC based on M2RGs. The model has good predictive value and accuracy for the prognosis of HNSCC patients. The correlation analysis of immune checkpoints and the sensitivity analysis of cancer therapeutic drugs have demonstrated that the model is expected to be used to guide clinical therapeutic decisions. Furthermore, it provides a theoretical foundation for the development of new immunotherapeutic technologies.

Methods

Data collection and processing

The HNSCC gene expression data and clinical information were obtained from TCGA database (https://portal.gdc.cancer.gov), which included 522 HNSCC samples and 44 normal samples. After removing patients without clinical information, the final 451 patients were included in the analysis as a training set (TCGA-HNSCC). The GSE41613 dataset was obtained from the GEO database (https://www.ncbi.nlm.nih.gov/geo/), which contains 97 patients with HNSCC, and after removing patients without clinical information, the final 76 patients were included in the analysis as a validation set (GSE41613).

Identification of M2RGs

The proportion of M2 macrophage infiltration in each HNSCC sample was analyzed using the R package “limma” based on the gene expression data of HNSCC in the TCGA database and the information about M2 macrophage infiltration in the CIBERSORTx43. According to the screening conditions (P < 0.05, |R|≥0.1), the R packages “tidyverse”, “ggplot2”, “ggpubr” and “ggExgra” were used to screen the genes co-expressed with M2 macrophage infiltration in HNSCC, these genes were regarded as M2RGs in HNSCC.

Functional enrichment and signaling pathway analysis of M2RGs

The biological processes of M2RGs and possible signalling pathways influencing M2 macrophage infiltration were explored by GO, KEGG and GSEA analysis. This procedure was done using the R packages “clusterProfiler”, “org.Hs.eg.db”, “enrichplot”, “ggplot2”, “circlize”, “RColorBrewer”, “dplyr”, “ggplot2”, “circlize”, “RColorBrewer”, “dplyr " and “ComplexHeatmap”.

Construction and validation of the prognostic model

A univariate Cox regression analysis of the genes in the training set was conducted using the R packages “survival” and “survminer,” which yielded M2RGs that were associated with the prognosis of HNSCC (P < 0.05, HR ≥ 1.75). Lasso regression analysis were conducted on the obtained genes using the R package “glmnet”, with cross-validation employed to identify the parameters of the regression model with the lowest error. The risk score for each HNSCC sample was calculated based on the Lasso regression coefficient of each gene and its corresponding gene expression level. This score was calculated by summing the product of the expression level of each gene and its regression coefficient44. The HNSCC samples were classified into high-risk and low-risk groups based on the median risk score. The predictive performance and accuracy of this prognostic model were validated in the TCGA and GEO databases, respectively.

Gene expression analysis and its relationship with clinicopathologic factors

The mRNA and protein expression of the model-constructing genes were analyzed in HNSCC tissues and normal tissues, based on the TCGA database and the HPA database (https://www.proteinatlas.org). The LinkedOmics database (https://www.linkedomics.org/login.php) and the UALCAN database (https://ualcan.path.uab.edu/analysis.html) were used to analyze the relationship between model-constructing genes and clinicopathologic factors.

Independent prognostic analysis

Univariate and multivariate Cox regression analyses of risk scores and clinicopathologic factors were performed using the R package “survival”. Time-dependent ROC curve analysis of risk scores and clinical factors with independent prognostic roles were performed using the R package “timeROC” to comprehensively assess the accuracy of risk scores in predicting the prognosis of HNSCC patients.

Construction and evaluation of nomogram

OS and PFS nomograms were constructed for patients with HNSCC based on the TCGA-HNSCC cohort using the R packages “survival”, “regplot” and “rms”. The ROC curve, KM curve, and calibration curve of the nomogram were plotted by the risk scores of the nomogram, and the accuracy of this nomogram in predicting the survival of HNSCC patients was comprehensively assessed.

Immunological correlation analysis

A comparative analysis of immune cell infiltration between groups based on different risk groups was conducted using the R packages “limma”, “vioplot”, “reshape2”, “tidyverse”, “ggplot2”, “ggpubr”, “corrplot”, and “ggExtra”, as well as an assessment of the correlation of risk scores with immune cell infiltration and immune checkpoints.

Analysis of gene mutation, TMB and drug sensitivity

Gene mutation data from HNSCC were downloaded from the TCGA database and combined with the risk scores of the samples in the TCGA-HNSCC cohort. The R packages “maftools”, “limma” and “ggpubr” were used to compute gene mutation frequencies and TMB between different risk groups. Combined with drug sensitivity files from the Genomics of Drug Sensitivity in Cancer (GDSC) database (https://www.cancerrxgene.org/). Using the R packages “limma”, “car”, “ridge”, “preprocessCore “, “genefilter”, “sva”, “biomaRt”, " GenomicFeatures”, “maftools”, “stringr”, “Org. Hs. Eg.Db “, “TxDb Hsapiens.UCSC.hg19.konwnGene”, and “oncoPredict” analyzed the sensitivity of drugs used for cancer treatment in different risk groups.

Statistical analysis

The data analysis and graphical visualisation in this study were conducted using the R software (version 4.3.2, R: The R Project for Statistical Computing, https://www.r-project.org). Univariate and multivariate Cox regression analyses were employed to identify prognostic genes for HNSCC and to assess the independent prognostic role of risk scores and clinicopathologic factors. Lasso regression was used to construct HNSCC prognostic model of M2RGs. A comparison of OS and PFS between risk groups was conducted using KM survival analysis. Time-dependent ROC curve was used to evaluate the accuracy of the prognostic model and the nomogram. P < 0.05 was considered statistically significant.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1 (164.8KB, pdf)
Supplementary Material 2 (9.9MB, pdf)

Acknowledgements

We acknowledge TCGA, GEO, CIBERSORTx, HPA, LinkedOmics, UALCAN and GDSC database for providing their platforms and contributors for uploading their meaningful datasets.

Author contributions

Designed the studies: S.M.H., H.Y.Z., X.L.Z.; Analyzed the data: S.M.H., H.R.C., C.Y.L., B.F., R.Z.Z.; Wrote the manuscript: S.M.H.; Supervised all studies: H.Y.Z. and X.L.Z., all authors critically reviewed the report and approved the final version.

Funding

This study was supported by the Science and Technology Foundation of Guizhou Provincial Health Committee (gzwkj2025-317, gzwkj2024-190).

Data availability

Data is provided within the manuscript or supplementary information files. The data sets analysed during the current study are all available in public databases. All the data corresponding to the HNSC series used in this study are available in GEO (https://www.ncbi.nlm.nih.gov/geo/) and TCGA (https://portal.gdc.cancer.gov), which are public functional genomics data repositories. The M2 macrophage-related genes data were downloaded from CIBERSORTx (https://cibersortx.stanford.edu/). The immunohistochemical data of M2 macrophage-related genes in HNSC and normal tissues were obtained from the HPA database (https://www.proteinatlas.org/). The clinicopathologic factors of HNSC were analyzed from LinkedOmics (https://www.linkedomics.org/login.php) and UALCAN databases (https://ualcan.path.uab.edu/analysis.html) . Drug sensitivity analysis data were obtained from GDSC database (https://www.cancerrxgene.org/).

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The study has been conducted in accordance with the ethical standards, according to the Declaration of Helsinki, and according to national and international guidelines. TCGA, GEO, CIBERSORTx, HPA, LinkedOmics, UALCAN and GDSC belong to public databases. The patients involved in the database have obtained ethical approval. Users can download relevant data for free for research and publish relevant articles. Our study is based on open source data, so there are no ethical issues and other conflicts of interest.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Contributor Information

Houyu Zhao, Email: zhy230322@163.com.

Xianlu Zhuo, Email: zxl221122@163.com.

References

  • 1.Johnson, D. E. et al. Head and neck squamous cell carcinoma. Nat. Rev. Dis. Primers. 6, 92. 10.1038/s41572-020-00224-3 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Ferlay, J. et al. Estimating the global cancer incidence and mortality in 2018: GLOBOCAN sources and methods. Int. J. Cancer. 144, 1941–1953. 10.1002/ijc.31937 (2019). [DOI] [PubMed] [Google Scholar]
  • 3.Sung, H. et al. Global cancer statistics 2020: globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin.71, 209–249. 10.3322/caac.21660 (2021). [DOI] [PubMed] [Google Scholar]
  • 4.Iacobuzio-Donahue, C. A. et al. Intratumor heterogeneity reflects clinical disease course. Nat. cancer. 1, 3–6. 10.1038/s43018-019-0002-1 (2020). [DOI] [PubMed] [Google Scholar]
  • 5.Abrahao, R. et al. The influence of smoking, age and stage at diagnosis on the survival after larynx, hypopharynx and oral cavity cancers in Europe: the arcage study. Int. J. Cancer. 143, 32–44. 10.1002/ijc.31294 (2018). [DOI] [PubMed] [Google Scholar]
  • 6.Takes, R. P. et al. Future of the TNM classification and staging system in head and neck cancer. Head Neck. 32, 1693–1711. 10.1002/hed.21361 (2010). [DOI] [PubMed] [Google Scholar]
  • 7.Ferris, R. L. Immunology and immunotherapy of head and neck cancer. J. Clin. Oncol.33, 3293–3304. 10.1200/JCO.2015.61.1509 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Dunn, G. P., Bruce, A. T., Ikeda, H., Old, L. J. & Schreiber, R. D. Cancer immunoediting: from immunosurveillance to tumor escape. Nat. Immunol.3, 991–998. 10.1038/ni1102-991 (2002). [DOI] [PubMed] [Google Scholar]
  • 9.Mittal, D., Gubin, M. M., Schreiber, R. D. & Smyth, M. J. New insights into cancer immunoediting and its three component phases–elimination, equilibrium and escape. Curr. Opin. Immunol.27, 16–25. 10.1016/j.coi.2014.01.004 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Noy, R. & Pollard, J. W. Tumor-associated macrophages: from mechanisms to therapy. Immunity. 41, 49–61. 10.1016/j.immuni.2014.06.010 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Pan, Y., Yu, Y., Wang, X. & Zhang, T. Tumor-associated macrophages in tumor immunity. Front. Immunol.11, 583084. 10.3389/fimmu.2020.583084 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Ferrari, S. M. et al. Immune and inflammatory cells in thyroid cancer microenvironment. Int. J. Mol. Sci.2010.3390/ijms20184413 (2019). [DOI] [PMC free article] [PubMed]
  • 13.Ruffell, B. & Coussens, L. M. Macrophages and therapeutic resistance in cancer. Cancer Cell.27, 462–472. 10.1016/j.ccell.2015.02.015 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Ngambenjawong, C., Gustafson, H. H. & Pun, S. H. Progress in tumor-associated macrophage (TAM)-targeted therapeutics. Adv. Drug Deliv Rev.114, 206–221. 10.1016/j.addr.2017.04.010 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Lewis, C. E., Harney, A. S. & Pollard, J. W. The multifaceted role of perivascular macrophages in tumors. Cancer Cell.30, 365. 10.1016/j.ccell.2016.07.009 (2016). [DOI] [PubMed] [Google Scholar]
  • 16.Li, B., Ren, M., Zhou, X., Han, Q. & Cheng, L. Targeting tumor-associated macrophages in head and neck squamous cell carcinoma. Oral Oncol.106, 104723. 10.1016/j.oraloncology.2020.104723 (2020). [DOI] [PubMed] [Google Scholar]
  • 17.Starska-Kowarska, K. The role of different immunocompetent cell populations in the pathogenesis of head and neck cancer-regulatory mechanisms of pro- and anti-cancer activity and their impact on immunotherapy. Cancers (Basel). 15. 10.3390/cancers15061642 (2023). [DOI] [PMC free article] [PubMed]
  • 18.Wan, P. K., Ryan, A. J. & Seymour, L. W. Beyond cancer cells: targeting the tumor microenvironment with gene therapy and armed oncolytic virus. Mol. Ther.29, 1668–1682. 10.1016/j.ymthe.2021.04.015 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Fridman, W. H., Pages, F., Sautes-Fridman, C. & Galon, J. The immune contexture in human tumours: impact on clinical outcome. Nat. Rev. Cancer. 12, 298–306. 10.1038/nrc3245 (2012). [DOI] [PubMed] [Google Scholar]
  • 20.Gajewski, T. F., Schreiber, H. & Fu, Y. X. Innate and adaptive immune cells in the tumor microenvironment. Nat. Immunol.14, 1014–1022. 10.1038/ni.2703 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Mantovani, A., Marchesi, F., Malesci, A., Laghi, L. & Allavena, P. Tumour-associated macrophages as treatment targets in oncology. Nat. Rev. Clin. Oncol.14, 399–416. 10.1038/nrclinonc.2016.217 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Zhang, Q. & Sioud, M. Tumor-associated macrophage subsets: shaping polarization and targeting. Int. J. Mol. Sci.2410.3390/ijms24087493 (2023). [DOI] [PMC free article] [PubMed]
  • 23.Zhang, Q. W. et al. Prognostic significance of tumor-associated macrophages in solid tumor: a meta-analysis of the literature. PLoS One. 7, e50946. 10.1371/journal.pone.0050946 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Liu, Z. et al. Tumor-associated macrophages promote metastasis of oral squamous cell carcinoma via CCL13 regulated by stress granule. Cancers (Basel). 14. 10.3390/cancers14205081 (2022). [DOI] [PMC free article] [PubMed]
  • 25.Hsieh, C. Y. et al. Macrophage secretory IL-1beta promotes docetaxel resistance in head and neck squamous carcinoma via SOD2/CAT-ICAM1 signaling. JCI Insight. 710.1172/jci.insight.157285 (2022). [DOI] [PMC free article] [PubMed]
  • 26.Zeng, J., Jiang, W. G. & Sanders, A. J. Epithelial protein lost in neoplasm, EPLIN, the cellular and molecular prospects in cancers. Biomolecules. 1110.3390/biom11071038 (2021). [DOI] [PMC free article] [PubMed]
  • 27.Huang, H., Du, Y., Zhao, D. & Chen, K. The relationship between the prognostic marker LIMA1 in head and neck squamous cell carcinoma and immune infiltration. J. Oncol.2022 (1040116). 10.1155/2022/1040116 (2022). [DOI] [PMC free article] [PubMed]
  • 28.Wang, Z. et al. CCDC86 promotes the aggressive behavior of nasopharyngeal carcinoma by positively regulating EGFR and activating the PI3K/Akt signaling. Neoplasma. 70, 761–776. 10.4149/neo_2023_221021N1039 (2023). [DOI] [PubMed] [Google Scholar]
  • 29.Yang, S. et al. STC2 promotes head and neck squamous cell carcinoma metastasis through modulating the PI3K/AKT/Snail signaling. Oncotarget. 8, 5976–5991. 10.18632/oncotarget.13355 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Huang, C. F. et al. Increased expression of peroxiredoxin 6 and cyclophilin A in squamous cell carcinoma of the tongue. Oral Dis.17, 328–334. 10.1111/j.1601-0825.2010.01730.x (2011). [DOI] [PubMed] [Google Scholar]
  • 31.Oweida, A. et al. Ephrin-B2 overexpression predicts for poor prognosis and response to therapy in solid tumors. Mol. Carcinog.56, 1189–1196. 10.1002/mc.22574 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Li, D. et al. Elevated ITGA5 facilitates hyperactivated mTORC1-mediated progression of laryngeal squamous cell carcinoma via upregulation of EFNB2. Theranostics 12, 7431–7449, doi: (2022). 10.7150/thno.76232 [DOI] [PMC free article] [PubMed]
  • 33.Shen, Y., Gu, H. M., Qin, S. & Zhang, D. W. Surf4, cargo trafficking, lipid metabolism, and therapeutic implications. J. Mol. Cell. Biol.1410.1093/jmcb/mjac063 (2023). [DOI] [PMC free article] [PubMed]
  • 34.Tang, X. et al. A SURF4-to-proteoglycan relay mechanism that mediates the sorting and secretion of a tagged variant of sonic hedgehog. Proc. Natl. Acad. Sci. U S A. 119 (e2113991119). 10.1073/pnas.2113991119 (2022). [DOI] [PMC free article] [PubMed]
  • 35.Petty, A. J. et al. Hedgehog signaling promotes tumor-associated macrophage polarization to suppress intratumoral CD8 + T cell recruitment. J. Clin. Invest.129, 5151–5162. 10.1172/JCI128644 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Zhang, Y. & Zhang, Z. The history and advances in cancer immunotherapy: understanding the characteristics of tumor-infiltrating immune cells and their therapeutic implications. Cell. Mol. Immunol.17, 807–821. 10.1038/s41423-020-0488-6 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Liu, Y. et al. Emerging phagocytosis checkpoints in cancer immunotherapy. Signal. Transduct. Target. Ther.8, 104. 10.1038/s41392-023-01365-z (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Liu, Y., Gusev, A. & Kraft, P. Germline Cancer gene expression quantitative trait loci are Associated with local and global tumor mutations. Cancer Res.83, 1191–1202. 10.1158/0008-5472.CAN-22-2624 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Samstein, R. M. et al. Tumor mutational load predicts survival after immunotherapy across multiple cancer types. Nat. Genet.51, 202–206. 10.1038/s41588-018-0312-8 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Cohen, E. E. W. et al. Pembrolizumab versus methotrexate, docetaxel, or cetuximab for recurrent or metastatic head-and-neck squamous cell carcinoma (KEYNOTE-040): a randomised, open-label, phase 3 study. Lancet. 393, 156–167. 10.1016/S0140-6736(18)31999-8 (2019). [DOI] [PubMed] [Google Scholar]
  • 41.Kiyota, N. et al. Weekly cisplatin plus radiation for postoperative head and neck cancer (JCOG1008): a multicenter, noninferiority, phase II/III randomized controlled trial. J. Clin. Oncol.40, 1980–1990. 10.1200/JCO.21.01293 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Han, J. et al. Concurrent carboplatin and paclitaxel definitive radiation therapy for locally advanced head and neck cancer. Head Neck. 45, 2207–2216. 10.1002/hed.27456 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Newman, A. M. et al. Robust enumeration of cell subsets from tissue expression profiles. Nat. Methods. 12, 453–457. 10.1038/nmeth.3337 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Xu, S., Wang, Z., Ye, J., Mei, S. & Zhang, J. Identification of iron metabolism-related genes as prognostic indicators for lower-grade glioma. Front. Oncol.11, 729103. 10.3389/fonc.2021.729103 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1 (164.8KB, pdf)
Supplementary Material 2 (9.9MB, pdf)

Data Availability Statement

Data is provided within the manuscript or supplementary information files. The data sets analysed during the current study are all available in public databases. All the data corresponding to the HNSC series used in this study are available in GEO (https://www.ncbi.nlm.nih.gov/geo/) and TCGA (https://portal.gdc.cancer.gov), which are public functional genomics data repositories. The M2 macrophage-related genes data were downloaded from CIBERSORTx (https://cibersortx.stanford.edu/). The immunohistochemical data of M2 macrophage-related genes in HNSC and normal tissues were obtained from the HPA database (https://www.proteinatlas.org/). The clinicopathologic factors of HNSC were analyzed from LinkedOmics (https://www.linkedomics.org/login.php) and UALCAN databases (https://ualcan.path.uab.edu/analysis.html) . Drug sensitivity analysis data were obtained from GDSC database (https://www.cancerrxgene.org/).

Articles from Scientific Reports are provided here courtesy of Nature Publishing Group

RESOURCES