Causal role of immune cells in head and neck squamous cell carcinoma

Abstract

Chronic inflammation, immune cell infiltration, and metabolic reprogramming play a crucial role in the development of head and neck squamous cell carcinoma (HNSCC). Based on the summary-level data from a genome-wide association study (GWAS), this study used Mendelian randomization (MR) analysis to investigate the causal relationship between immune cell phenotype and HNSCC. Two-step MR was employed to quantify the proportion of the effect of metabolite-mediated immunophenotypes on HNSCC. 16 immune phenotypes exhibited a causal relationship with HNSCC. HLA DR + CD4 + AC (OR: 1.31, 95% CI 1.07–1.59, P < 0.01) and EM CD8br AC cells (OR: 1.36, 95% CI 1.14–1.62, P < 0.01) showed the most significant P values. 13 metabolites have a causal relationship with HNSCC (P < 0.05). The proportion of the effect of Bilirubin degradation product, C16H18N2O5 (3) levels-mediated HLA DR + CD4 + AC on HNSCC was 7.17% (95% CI 1.04%-13.3%), while the proportion of the effect of Glutamine to alanine ratio-mediated EM CD8br AC on HNSCC was 8.88% (95% CI 2.2%-15.6%). Moreover, single-cell RNA-sequencing analysis confirmed that HLA DR + CD4 + AC and EM CD8br AC cells were commonly present in the HNSCC tumor microenvironment (TME). High expression of two immune phenotypes indicated a poorer prognosis. The results of differentially expressed genes (DEGs) analysis and functional enrichment indicated that Interferon-induced protein with tetratricopeptide repeats 1 (IFIT1)/Follicular dendritic cell secreted protein (FDCSP) and Keratin 19 type I (KRT19)/Kallikrein-related peptidase 5 (KLK5) may be responsible genes for two immune phenotypes, respectively. The results of quantitative real-time polymerase chain reaction (qRT-PCR) and survival analysis showed that FDCSP was significantly downregulated in HNSCC and could exert a protective effect. Collectively, our study clarified the causal relationship between immune cells, metabolites, and HNSCC, providing new insight for clinical diagnosis and treatment of HNSCC.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-02135-9.

Introduction

Head and neck squamous cell carcinoma (HNSCC) is the fifth cancer type, which is a highly invasive malignant tumor that causes over 890000 new cases and 450000 deaths annually [1]. HNSCC has multiple high-risk factors, including human papillomavirus (HPV) infection, chronic inflammatory stimulation, smoking, and chewing betel nuts [2]. HNSCC is a solid tumor, in addition to tumor cells, there are also various functional cells such as immune cells and stromal cells in the tumor microenvironment (TME). The abnormal activation and dysfunction of immune cells play a crucial role in the occurrence and development of HNSCC [3].

Immune cells, which are highly integrated and rely on a complex network of signal transduction pathways and checkpoints, play a vital role in regulating cellular processes and functions. However, the composition of immune cells is complex and subject to significant genetic differences among individuals. Further analysis of individual immune cell differences and characteristics is crucial for improving immunotherapy in cancer patients [4]. A recent Genome-wide association study (GWAS) study reported 539 immune features and 192 relative cell counts (ratios between cell levels) through flow cytometry analysis, totaling 731 immune phenotypes [5]. This provides an important reference for better understanding the composition, function, and relationship with diseases of immune cells. Despite this progress, the role of these immune phenotypes in HNSCC is still unclear.

The metabolic changes in immune cells are closely related to the inhibition of anti-tumor immune activity and the proliferation of tumor cells [6]. Changes in cellular metabolism are a hallmark of various cancers [7]. Numerous studies have highlighted the important role of abnormal cellular metabolism in the onset and progression of tumors [8, 9]. Interestingly, these metabolic abnormalities may precede tumor formation, potentially leading to mutations in oncogenes [10]. Metabolites are small molecules that are the intermediate or end products of metabolic reactions. Their levels are influenced by many factors, including genetics, diseases, and cancers. A recent study has undertaken a series of large GWASs, comprising 1,091 metabolites and 309 metabolite ratios in 8,299 individuals, which brings hope to the metabolic research and clinical treatment target exploration of HNSCC [11]. However, there is a notable lack of causal studies exploring the characteristics of immune cells, their metabolites, and HNSCC.

Mendelian randomization (MR) is a powerful causal inference method that uses genetic variation as an instrumental variable [12]. This study aims to use Mendelian randomization to investigate the causal relationship between immunophenotype and HNSCC, as well as assess the proportion of metabolites mediate the effect of immune cells on HNSCC. The two immune phenotypes are most likely to have a causal relationship with HNSCC. We extracted two types of cells through single-cell RNA-sequencing analysis and confirmed their widespread presence in HNSCC. In addition, the prognostic significance, differentially expressed gene identification and functional analysis of two potential immune cell subtypes were performed by analyzing bulk RNA-sequencing data (Fig. 1).

Fig. 1.

The research framework diagram of this study

Materials and methods

Dataset

The GWAS summary statistical data were publicly obtained from the GWAS directory (GCST0001391 to GCST0002121) [5]. A total of 731 immune phenotypes were included, including absolute cell (AC) counts (n = 118), median fluorescence intensities (MFI) reflecting surface antigen levels (n = 389), morphological parameters (MP) (n = 32), and relative cell (RC) counts (n = 192). GWAS data for HNSCC was obtained from the IEU database (https://gwas.mrcieu.ac.uk/) finn-b-C3-LIP-ORAL-PHARYNX dataset. This study conducted GWAS on 218792 European individuals. 1400 GWAS data on individual disease-related metabolites was publicly obtained from the GWAS directory (GCST90199621-90201020) [11]. The single-cell sequence data for HNSCC included 24 samples, retrieved from the GEO database (GSE139324, GSE173647, and GSE173964) [13, 14]. The TPM expression data for HNSCC was downloaded using R package TCGAbiolinks (n = 501).

Selection of instrumental variable

R package TwoSampleMR was performed [15]. The significance level of the instrumental variable for each immune trait was set to 10^–5. The function trims single nucleotide polymorphisms (SNPs) according to the linkage disequilibrium r² threshold < 0.001 within 10000 kb distance standard. Then calculate the F-statistic for each instrumental variable and remove instrumental variables with F < 10 (weak instrumental variables) [16–18].

Inference of causal relationships

Five methods, including Inverse variance weighted (IVW, main method), MR Egger, Weighted median, Simple mode, and Weighted mode were used to evaluate the causal relationships between 731 immune phenotypes and HNSCC, as well as 1400 metabolites and HNSCC. The P-value calculated by the IVW was less than 0.05, indicating a statistically significant causal role. Only 5 methods with consistent Odds Ratio (OR) values and causal relationships with P > 0.05 in the horizontal pleiotropy test were retained.

Mediation analysis

The TwoSampleMR package was used to determine the causal relationship between immunophenotypes and metabolites [15]. b1 was the effect of immunophenotype on metabolites. b2 was the effect of metabolites on HNSCC. b3 was the total effect of immunophenotype on HNSCC. The mediating effect was calculated as b1*b2, while the direct effect was derived by subtracting the mediating effect from the total effect, i.e., b3-b1*b2.

Quality control, cell type clustering of single-cell sequence data

We filtered out cells with mitochondrial gene content exceeding 20%, resulting in 73,660 cells for downstream analysis. Due to the insignificant batch effects between different samples, the batch effects of the samples were not removed. Standardize and normalize single-cell data by using the SCTransform function in Seurat. The RunPCA function was employed to reduce the dimensionality of the data, and the parameter was set to dim = 1:30 to cluster and t-Distributed Stochastic Neighbor Embedding (t-SNE) to reduce dimensionality following the cell characteristics provided by CellTypist (https://www.celltypist.org) (randomly selected 1/5 of the cells) [19].

Annotations for HLA-DR + CD4 + T cell and EM CD8 + T cell

Repeat the dimensionality reduction clustering operation, removing the non-CD3 expressing cell population (NK cell population). Based on the expression levels of CD4 and CD8A genes, divide the T cell population into CD4 + T-cells and CD8 + T-cells. According to the expression of HLA-DR series genes and gene markers for EM CD8T [20], HLA DR + CD4 + AC and EM CD8br AC were identified by using the AddModuleScore.

BayesPrism infers the cellular composition of bulk samples based on single-cell data

BayesPrism package was performed to firstly visualize abnormal genes in single-cell sequence data, filtering out ribosomal proteins, chrM, and MALAT1 genes. The consistency of gene expression across different data was examined. The highest consistency of protein-coding genes in bulk and single-cell sequence data was observed. To reduce batch effects, only the protein-coding genes were deconvoluted. The relative proportion of different types of cell populations was obtained in each bulk sample by BayesPrism [21].

Limma calculates intergroup differentially expressed genes

Based on the relative proportions of different types of cell populations output by BayesPrism in each bulk sample, the median relative proportions of HLA DR + CD4 + AC and EM CD8br AC were used as the cutoff points to divide the samples into high and low expression groups. After log2 (x + 1) processing of bulk data, the differential expressed genes (DEGs) were determined by using limma package [22]. The Kaplan–Meier plot was drawn using the Sangerbox platform and R software [23]. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were further performed to reveal the function of DEGs.

Quantitative reverse transcription polymerase chain reaction (qRT-PCR) and Prognostic significance analysis

A total of 30 HNSCC tissues and paired normal tissue specimens were collected postoperatively from Taizhou Central Hospital between January and December 2018. All specimens were frozen in liquid nitrogen at − 80℃. Total RNA was extracted with Trizol Reagent (Invitrogen) according to the official protocol. RNA was reverse transcribed into cDNA using a PrimeScript RT reagent kit (Takara Bio). qRT-PCR was conducted on SYBRH Select Master Mix (Invitrogen). The relative expression of four selected genes were normalized by glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Data were quantified using 2^−∆∆Ct method. The primer was synthesized by Genepharma (Table S1). The study was approved by the Ethics Committee of Taizhou Central Hospital (E-2020KT01). All procedures were conducted strictly in compliance with the Declaration of Helsinki 1964 or equivalent ethical principles. Informed consent was obtained from all participants prior to their enrollment in the study. According to the median relative expression levels of the four genes, patients were divided into high-expression group and low-expression group. The gene expression levels in the cancer tissues and normal tissues, as well as the survival curves, were plotted using Graphpad Prism (version 7.0) software. Statistical significance was calculated using paired t-test and Log-rank test. P-value less than 0.05 is considered statistically significant.

Results

Association of immunophenotypes with HNSCC

The IVW method was used as the main analysis. 17 immunophenotypes (Unsw Mem %lymphocyte, IgD- CD38dim AC, HLA DR +  + monocyte %leukocyte, EM CD4 + AC, EM CD8br AC, CD8br %leukocyte, HLA DR + CD4 + AC, CD3- lymphocyte %leukocyte, NK AC, CD24 on IgD + CD38br, CD27 on CD20-, CD27 on CD20- CD38-, CD27 on CD24 + CD27 + , HVEM on TD CD8br, CD11c on myeloid DC, CD45RA on resting Treg, HLA DR on plasmacytoid DC) showed causal effects on HNSCC (P < 0.05, Fig. 2, Table S2). The results of MR analysis showed only EM CD4 + AC exhibited reverse causality on HNSCC (P < 0.05), and subsequently filtered.

Fig. 2.

Mendelian Randomization Analysis showed the causal effect of immunophenotypes on HNSCC

Association of metabolite with HNSCC

13 metabolites (Docosapentaenoate n3 DPA; 22:5n3 levels, 2-palmitoyl-GPC (16:0) levels, Cortolone glucuronide (1) levels, Bilirubin degradation product, C16H18N2O5 (3) levels, X-11315 levels, X-11843 levels, X-24531 levels, X-24970 levels, Biliverdin levels, Adenosine 5'-diphosphate (ADP) to N-palmitoyl-sphingosine (d18:1–16:0) ratio, Phenylpyruvate to 4-hydroxyphenylpyruvate ratio, Alpha-ketoglutarate to proline ratio, Glutamine to alanine ratio) showed causal effects on HNSCC (P < 0.05, Table S3).

Mediation analysis of metabolite

Whether immune cells could cause HNSCC by regulating metabolites was further investigated by employing Mediation Mendelian randomization. Adjust the P-value to < 0.01 and screen out two immunophenotypes associated with HNSCC for downstream analysis. HLA DR + CD4 + AC may indirectly lead to the HNSCC mediated Bilirubin degradation product, C16H18N2O5 (3) levels (proportion mediated: 7.17%; 95% CI 1.04–13.3%) (Fig. 3, Table 1). EM CD8br AC may indirectly lead to HNSCC by affecting the glutamine-to-alanine ratio (proportion mediated: 8.88%; 95% CI 2.2–15.6%) (Fig. 4, Table 1).

Fig. 3.

Forest plot to visualize the metabolite mediation effect (HLA DR + CD4 + AC)

Table 1.

The metabolite mediation effect

Immune cell	Metabolite	Outcome	Mediated effect	Mediated proportion	P value
HLA DR + CD4 + AC	Bilirubin degradation product, C16H18N2O5 (3) levels	Head and neck cancer	0.0193 (0.00281, 0.0358)	7.17% (1.04%, 13.3%)	0.022
EM CD8br AC	Glutamine to alanine ratio	Head and neck cancer	0.0275 (0.00682, 0.0482)	8.88%(2.2%, 15.6%)	0.009

Bold values signify P <0 .05

Fig. 4.

Forest plot to visualize the metabolite mediation effect (EM CD8br AC)

Identification of two immune subpopulations

By reducing the dimension and clustering the single-cell data, six cell population were identified (Fig. 5A). After excluding the NK cell population, multiple dimensionality reduction annotations were performed. Based on the HLA-DR-related genes and EM CD8T-related gene markers in the single-cell expression matrix, HLA DR + CD4 + AC and EM CD8br AC were identified (Fig. 5B, C). By deconvoluting protein-coding genes, the relative proportions of different cell populations in bulk samples were predicted. As shown in Fig. 5D, HLA DR + CD4 + AC and EM CD8br AC were commonly present in HNSCC microenvironment. Moreover, based on the expression levels of the two cells, we further evaluated their impact on the prognosis of HNSCC patients by using Kaplan–Meier method. The results showed that both HLA DR + CD4 + AC and EM CD8br AC played a pro-tumorigenic role, suggesting a poorer prognosis (Fig. 5E).

Fig. 5.

Single-cell and Bulk Sequence Analysis. A Dimensionality reduction plot for single-cell sequence data. B, C Dimensionality reduction plots of HLA DR + CD4 + AC and EM CD8br AC cells. D Bar chart of cell type proportion in bulk samples. E The Kaplan–Meier plots showed the prognostic significance of the two cell types

The DEGs according to the expression of two immune subpopulations

The DEGs were identified based on the proportion of HLA DR + CD4 + AC. Among these, interferon-induced protein with tetratricopeptide repeats 1 (IFIT1) showed the most significant up-regulation, while follicular dendritic cell secreted protein (FDCSP) was significantly down-regulated (Fig. 6A, Table S4). The GO enrichment analysis indicated that these DEGs were primarily enriched in categories such as defense response to virus, response to virus, defense response to other organisms, and immune effector process (Fig. 6B). The DEGs according to the proportion of EM CD8br AC were also identified. Keratin 19 type I (KRT19) showed the most significant up-regulation, while kallikrein-related peptidase 5 (KLK5) showed the most significant down-regulation (Fig. 6C, Table S5). The results of GO enrichment analysis showed that the DEGs were mainly enriched in Adaptive immune response, Humoral immune response, and Immune response regulating signaling pathway (Fig. 6D). KEGG functional enrichment analysis was further employed to infer the functions of two types of cells. The results showed that DEGs of HLA DR + CD4 + AC were mainly enriched in antiviral immunity, such as Epstein-Barr virus infection and Human papillomavirus infection. The DEGs of EM CD8br AC were mainly enriched in some metabolic pathways, such as Drug Metabolism cytochrome P450, Chemical carcinogenesis, and Glutathione metabolism (Fig. 6E). Notably, a large number of antibody-related genes were different expressed accroding to the EM CD8br AC expression. As is well known, B cells are the main group of cells that produce antibodies. We further compared the correlation between the expression of EM CD8br AC and B cells, and the results showed a weak positive correlation between them (Fig. 6F).

Fig. 6.

The DEGs according to the expression of HLA DR + CD4 + AC and EM CD8br AC. A) Volcano plot of DEGs for HLA DR + CD4 + AC. B The GO enrichment analysis for HLA DR + CD4 + AC. C Volcano plot of DEGs for EM CD8br AC. D The GO enrichment analysis for EM CD8br AC. E The KEGG enrichment analysis for HLA DR + CD4 + AC (left) and EM CD8br AC (right). F Correlation analysis between EM CD8br AC and B cells

High expression of FDCSP correlated with better prognosis in HNSCC

In order to preliminarily explore the molecular targets and mechanisms underlying the improved the prognosis of HNSCC patients, we perform qRT-PCR and survival analysis focusing on two highly differential expressed genes in two cell types: IFIT1 and FDCSP, KRT19 and KLK5. IFIT1 was significantly upregulated in HNSCC patients. Although high expression of IFIT1 suggests a poorer prognosis, the correlation was not statistically significant (P > 0.05, Fig. 7A). FDCSP was significantly downregulated in HNSCC. HNSCC patients with high expression of FDCSP have a better prognosis (P < 0.05, Fig. 7B). There was no statistically significant difference in the expression and prognostic significance of KRT19 and KLK5 (P > 0.05, Fig. 7C, D).

Fig. 7.

The results of qRT-PCR and Kaplan–Meier. A The expression level of IFIT1 and the prognostic significance for HNSCC. B The expression level of FDCSP and the prognostic significance for HNSCC. C The expression level of KRT19 and the prognostic significance for HNSCC. D The expression level of KLK5 and the prognostic significance for HNSCC

Discussion

HNSCC poses a significant health burden due to its multifaceted pathogenesis, often triggered by persistent local chronic stimulation and inflammation [24]. Inflammation plays an important role in the initiation, progression, invasion, and metastasis of tumors. Chronic inflammation may induce the occurrence of HNSCC through various effects such as inducing gene mutations, promoting tumor cell formation, and promoting angiogenesis [25]. Key inflammatory cells, such as immune cells, profoundly influence the tumor microenvironment, modulating the expression of oncogenes and tumor suppressors through intricate cytokine networks [26, 27]. For example, the nuclear factor-kappa B (NF-kB) signaling pathway can activate and control cell proliferation, neovascularization, and chemokine regulation, becoming a major mechanism for promoting tumor formation [28]. This study investigates the specific immune cell types that potentially contribute to HNSCC pathogenesis. The results showed a causal relationship between 16 immunophenotypes and HNSCC. The P values of HLA DR + CD4 + AC and EM CD8br AC cells were the most significant.

Numerous studies have confirmed significant metabolic reprogramming in tumors, including not only glycolysis and the tricarboxylic acid cycle but also involving alterations in fatty acid and glutamine metabolism [29, 30]. The metabolic changes of immune cells may significantly alter their functional activity and anti-tumor ability. This study aimed to investigate the potential causal relationship between metabolites and HNSCC, as well as the role of immune cell metabolism in mediating HNSCC. The results showed a causal relationship between 13 metabolites and HNSCC. However, the direct influence of metabolites on HNSCC progression was found to be relatively modest. This suggests that immune cells likely promote HNSCC through mechanisms such as modulation of gene expression, chemokine signaling, and other pathways, rather than primarily through metabolite-mediated effects.

We further screened HLA DR + CD4 + AC and EM CD8br AC immune cells by incorporating data from single-cell RNA and bulk data. These two types of cells were commonly present in the HNSCC TME. By incorporating Bulk data, we confirmed that these two immune cell phenotypes play a pro-tumorigenic role in HNSCC. The DEGs and functional enrichment were performed to further understand the potential role of two immune phenotypes in HNSCC. IFIT1 and FDCSP were the gene with the most differential expression based on HLA DR + CD4 + AC. Previous study has reported that IFIT1 is highly expressed and promotes the progression of HNSCC [31]. FDCSP has been proven to be at the level of bioinformatics and may serve as a potential chemokine-associated prognostic biomarker in HPV + HNSCC [32]. KRT19 and KLK5 were the most differentially expressed gene based on EM CD8br AC. The protein encoded by KRT19 is a member of the keratin family. This gene has been observed to be abnormally expressed in thyroid cancer [33]. Previous studies have found that KLK5 in saliva as a molecular marker for malignant transformation of HNSCC [34]. In addition, abnormal expression of KLK5 has been found in cervical cancer, with significant implications for tumor prognosis [35]. Here, we perform qRT-PCR and prognostic analysis on a cohort of HNSCC patients. We confirm that FDCSP is significantly down-regulated in HNSCC and is closely associated with patient prognosis. This suggests that FDCSP may be a key molecular target for HLA DR + CD4 + AC to exert pro-tumor effects.

In summary, our study utilized Mendelian randomization's causal inference ability to clarify the roles of immune cells and metabolites in the occurrence of HNSCC. Simultaneously, it provided new insights into the causal role of chronic inflammation and HNSCC. We have identified two immune cells that are most likely causally related to HNSCC and confirmed FDCSP as a molecular target for anti-cancer treatment. This may provide experimental and theoretical basis for the development of HNSCC immunotherapy and chemotherapy targets in the future. The limitation of this study was that GWAS data is based on the European, and further evidence is needed to determine whether the conclusions can be extended. Secondly, the mediating effect of metabolites was weak, and it is necessary to explore more accurate mediators. In addition, we found that the high EM CD8br AC group also showed upregulation of several antibody genes. However, the correlation between EM CD8br AC and B cells was weak. The clinical significance of these antibody genes needs further exploration.

Abbreviations

HNSCC

Head and neck squamous cell carcinoma

GWAS

Genome-wide association study

MR

Mendelian randomization

DEGs

Differentially expressed genes

TME

Tumor microenvironment

IFIT1

Tetratricopeptide repeats 1

FDCSP

Follicular dendritic cell secreted protein

KRT19

Keratin 19 type I

KLK5

Kallikrein-related peptidase 5

qRT-PCR

Quantitative real-time polymerase chain reaction

HPV

Human papillomavirus

AC

Absolute cell

MFI

Median fluorescence intensities

MP

Morphological parameters

RC

Relative cell

SNPs

Single nucleotide polymorphisms

IVW

Inverse variance weighted

OR

Odds Ratio

t-SNE

T-distributed stochastic neighbor embedding

GO

Gene ontology

KEGG

Kyoto encyclopedia of genes and genomes

GAPDH

Glyceraldehyde-3-phosphate dehydrogenase

ADP

Adenosine 5′-diphosphate

IFI6

Interferon alpha-inducible protein 6

NF-kB

Nuclear factor-kappa B

Author contributions

Conceptualization, Panshu Xu and Zhenxing Zhang; Formal analysis, Haoran Zhu, Panshu Xu and Zhenxing Zhang; Funding acquisition, Zhenxing Zhang; Project administration, Yukang Ying; Supervision, Yukang Ying; Writing—original draft, Panshu Xu and Zhenxing Zhang; Writing—review & editing, Conglin Hu and Yukang Ying.

Funding

This study was supported by the Taizhou Social Development Science and Technology Project (No.23ywa20).

Data availability

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Code availability

R software was used in this study. The relevant R package was explained in the Materials and Methods.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of Taizhou Central Hospital (E-2020KT01). All procedures were conducted strictly in compliance with the Declaration of Helsinki 1964 or equivalent ethical principles. Informed consent was obtained from all participants prior to their enrollment in the study.

Consent for publication

All authors approved the final manuscript and the submission to this journal.

Competing interests

The authors declare no competing interests.