Construction of anti-tumor immune gene signature for HNSCC and identification of 4'-Hydroxywogonin as a potential immunotherapy combination drug

Abstract

Background

Head and neck squamous cell carcinoma (HNSCC) is an aggressive and heterogeneous malignancy, presenting challenges in accurately forecasting prognosis and immunotherapy response. This study endeavors to develop a robust gene signature to augment the prognostic prediction of HNSCC, and simultaneously uncover potential immunotherapy combination drug.

Methods

Transcriptome data from clinical HNSCC patients were analyzed using LASSO regression algorithm to construct a gene signature, followed by survival curve and ROC curve analyses, immune correlation analysis, and nomogram building. Furthermore, a comprehensive virtual screening of approximately 30,000 molecules was carried out based on the key target of signature. Finally, the potential immunotherapy combination drug was identified by molecular docking, CCK-8 assay, RT-qPCR assay, and molecular dynamics simulation.

Results

An anti-tumor immune gene signature (ATIGS) comprising 14 genes was established, which had promising potential in predicting the prognosis of HNSCC patients and could serve as an independent prognostic factor. Notably, ATIGS demonstrated a significant correlation with the infiltration level of several immune cells in the tumor immune microenvironment. It also had a good performance in predicting the response to immunotherapy. Further, protein–protein interaction (PPI) network analysis identified ZAP70 as a key target of ATIGS. Virtual screening of ~ 30,000 compounds found Puerol A, 4′-Hydroxywogonin, and Cirsimaritin as candidates, with 4′-Hydroxywogonin showing the strongest inhibition of CAL-27 and SCC7 cells. It also upregulated the expression levels of immune-related genes ZAP70 and LAT in CAL-27 cells, hinting at anti-tumor effect via immune pathway regulation. Molecular dynamics simulation showed stable binding of 4′-Hydroxywogonin to ZAP70 through hydrogen bonds and hydrophobic interactions, involving residues ALA417, GLU295, and LYS369.

Conclusions

This study successfully constructed a robust gene signature ATIGS, as well as identified a potential immunotherapy combination drug 4′-Hydroxywogonin. The ATIGS could effectively predict the prognosis and immunotherapy response of HNSCC patients, which may contribute towards enriching the immunotherapy-responsive population and enhancing the clinical efficacy of immunotherapy. Meanwhile, 4′-Hydroxywogonin may provide a promising clinical treatment strategy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12935-025-03910-y.

Introduction

Head and neck squamous cell carcinoma (HNSCC) is a highly aggressive and heterogeneous malignancy with a high mortality rate and tendency for recurrence. Owing to a lack of awareness, a majority of patients were diagnosed at advanced stages, and over fifty percent of them eventually encounter recurrence or distant metastases, resulting in a 5-year survival rate of less than 40% [1]. As an emerging treatment modality, immunotherapy is considered as the most promising strategy for managing advanced or aggressive cancers, offering the potential to significantly ameliorate the prognosis of HNSCC patients [2]. Nonetheless, in the absence of patient selection, the objective response rate of immunotherapy is only about 30%. Therefore, it is imperative to identify biomarkers that can predict the clinical efficacy of immunotherapy.

Anti-tumor immune genes play a critical role in the immune system. Some genes encode proteins that act as immune checkpoints to regulate the immune response, while others produce cytokines that promote the activation and proliferation of immune cells [3]. Besides, anti-tumor immune genes involved in antigen processing and presentation, such as MHC class I and II genes, are indispensable for presenting cancer antigens to immune cells and triggering an immune response against cancer [4, 5]. Therefore, it is essential to develop a gene signature based on anti-tumor immune genes, which is helpful to enrich the immune response population and improve the overall immunotherapy effect of HNSCC patients.

In this study, based on anti-tumor immune genes and transcriptome data of HNSCC patients, we constructed an anti-tumor immune gene signature (ATIGS) and evaluated its potential in predicting the response to immune checkpoint blockers. Further, by integrating the gene signature and multiple clinicopathological factors, we established a nomogram that enables the quantitative clinical prediction of individual patient’s survival probability. In addition, we identified the key target of signature and screened potential immunotherapy combination drug through large-scale virtual screening and molecular docking. Finally, CCK-8 assay, RT-qPCR assay and molecular dynamics simulation were used for further verification.

Materials and methods

Collection of patient clinical data and anti-tumor immune genes

The transcriptomic data and clinical information of HNSCC samples used in this study were obtained from The Cancer Genome Atlas (TCGA) data portal and Gene Expression Omnibus (GEO) [6, 7]. The TCGA cohort, consisting of 502 samples, was designated as the internal testing set and was further divided into a training set and a validation set using a random bisection method. The GSE65858 cohort, consisting of 270 samples, was selected as the external testing set. Table S1 provides a summary of the patient demographics and clinical characteristics of the included datasets. Besides, the genes related to anti-tumor immune (GRATIs) were sourced from TISIDB (http://cis.hku.hk/TISIDB/download.php), which is an integrated repository portal for tumor-immune system interactions.

Identification of differentially expressed GRATIs and functional enrichment analysis

In the TCGA cohort, a total of 502 HNSCC samples and 44 normal samples were analyzed to identify differentially expressed genes. The criteria for differential expression were established as |log2(Fold Change)|> 1 and P-value < 0.05 [8]. The intersection of the differentially expressed genes and GRATIs was captured to obtain a set of differentially expressed GRATIs. Additionally, a Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analysis was conducted on the differentially expressed GRATIs with a threshold of P-value < 0.05 [9]. Meanwhile, Gene Ontology (GO) enrichment analysis was performed, including molecular function (MF), cellular component (CC), and biological process (BP) categories [10].

Construction of anti-tumor immune gene signature (ATIGS)

Among differentially expressed GRATIs, the univariate Cox proportional hazard regression model was adopted to identify prognosis-related GRATIs with the cutoff of P < 0.05 in TCGA training set. Further, the Least Absolute Shrinkage and Selection Operator (LASSO) penalized Cox proportional hazards regression was selected to construct the most optimal gene signature ATIGS through R package “glmnet” [11]. The risk score for each patient was calculated using the following formula: risk score = [Expression level of Gene 1 * coefficient] + [Expression level of Gene 2 * coefficient] + … + [Expression level of Gene n * coefficient]. Patients were then stratified into high-risk and low-risk groups according to the median value of the risk score.

Validation and evaluation of ATIGS

To evaluate the predictive power of ATIGS, Kaplan–Meier (K-M) survival curves were generated using the R package “survival” for four gene sets: TCGA training set, TCGA validation set, TCGA internal testing set, and GSE65858 external testing set [12]. The median survival time and hazard ratio (HR) were calculated by the R package “survminer”. To determine the sensitivity and specificity of ATIGS, time-dependent receiver operating characteristic (ROC) curves were plotted using the R package “survival ROC” for the same four gene sets, including 1-, 3-, and 5-year survival [13]. In addition, we analyzed the correlation between ATIGS and multiple clinicopathological factors using independent t-tests.

Construction of prognostic nomogram

To develop a quantitative analytical tool for evaluating the survival probability of HNSCC patients, we commenced by performing univariate and multivariate analyses using the R package “survival” to assess the correlation among ATIGS, multiple clinicopathological factors, and overall survival (OS) [12]. In particular, we evaluated the efficiency of ATIGS in different pathological stages through the R packages “survival” and “survminer”. Subsequently, by integrating ATIGS and multiple clinicopathological factors, we developed a prognostic nomogram that enables a quantitative assessment of the individual patient’s probability of survival. Ultimately, we utilized calibration curves to compare the predicted and observed survival probability.

Assessment of immune infiltration and immunotherapy response

In order to compare the immune infiltration of patients in high- and low-risk groups, CIBERSORT was adopted to estimate the infiltration levels of 22 immune cell subtypes [14]. The correlation between these immune cells and risk score were assessed by the Spearman correlation analysis. Since immunotherapy involves seven-step anti-cancer immune responses, we analyzed which specific anticancer immune response is closely associated with ATIGS. Further, we evaluated the association between risk score and immunogenicity that could be characterized by immunophenoscore (IPS) obtained from The Cancer Immunome Atlas (TCIA) (https://tcia.at/home) [15]. Last but not least, relevant cohorts of patients who received PD-1 blockade were utilized for further validation.

Identification of the key target of ATIGS

To identify the key target of ATIGS, all signature genes were uploaded to the STRING database (https://string-db.org/) to construct a protein–protein interaction (PPI) network with a combined score greater than 0.4 [15]. Further, a network analyzer plug-in of Cytoscape was used to execute the topology analysis, in which degree centrality, betweenness centrality, and closeness centrality were chosen to identify the key target of ATIGS.

Virtual screening and molecular docking

To identify potential compounds for the key target of ATIGS, we selected ~ 30,000 compounds from six compound libraries for virtual screening, including the FDA-Approved Drug Library of Selleck, Natural Compound Library of TargetMol, Natural Product Library of MedChemExpress, Natural Product Library of Pharmacodia, Natural Product Library of BioBioPha BBP, and Anti-cancer-Compound Library Plus of APExBIO. Firstly, the crystal structure of the key target was predicted by AlphaFold2 and preprocessed using the Protein Preparation Wizard module of Schrodinger. Simultaneously, the LigPrep module of Schrodinger was used to process the 2D structures of ~ 30,000 compounds and generate all 3D chiral conformations. Following this, the SiteMap module of Schrodinger was used to predict the optimal binding site, and then the Receptor Grid Generation module was used to obtain the active pocket of the protein. Further, the preprocessed compounds were docked with the active site of the target protein at the levels of HTVS, SP, and XP in turn. Finally, the bindings of compounds with low-docking scores to the key target were analyzed by MM-GBSA [16].

Cell culture and viability assay

CAL-27 and SCC7 cells were cultured in DMEM basic medium supplemented with 10% FBS and 100 U/mL penicillin–streptomycin, which was placed in a humidified incubator containing 5% CO₂ at 37 °C. Cells in good condition in the logarithmic growth phase were seeded into 96-well plates at a certain cell density at 100 μL per well. After overnight, the cells were treated with Top3 potential compound solutions at different gradient concentrations for 24 h. Subsequently, 10 μL CCK-8 reagent was added to each well and incubated for another 2 h. Finally, the optical density (OD) of the culture plate was measured at 450 nm wavelength on the microplate reader.

Real-time quantitative q-PCR

CAL-27 cells in logarithmic growth phase were spread into 6-well plates and treated with the most promising candidate compound. After 24 h of incubation, the cells were lysed with Trizol to extract total RNA, and the concentration and purity were tested. RNA was reverse transcribed into cDNA using PrimeScript RT Reverse Transcription Kit (Takara, Japan), and then PCR amplification was performed using primers and SYBR Green Master Mix reagent (Takara, Japan). The forward and reverse primers were shown in Table S2. The GAPDH gene was used as the reference gene, and the relative expression levels of other genes were calculated by the 2^−ΔΔCT method.

Molecular dynamics simulation

To further investigate the binding affinities of the key target and the most promising candidate compound, we conducted molecular dynamics (MD) simulations using the Desmond program [17]. The OPLS4 force field was employed to parameterize the protein and small molecules, while the SPCE model was used for the water solvent. The protein-small molecule complex was placed in a cubic water box and solvated. The system’s charge was neutralized by adding 0.150 M chloride and sodium ions. The energy of the system was initially minimized using the steepest descent minimization method for 50,000 steps. Subsequently, the positions of heavy atoms were restrained for NVT and NPT equilibration for an additional 50,000 steps. The system temperature was maintained at 300 K, and the system pressure was maintained at 1 bar. After completing the two equilibration stages, an unrestricted simulation was performed for 100 ns. The interactions were analyzed and dynamic trajectory data were generated using Maestro 13.5 [18].

Statistical analysis

The volcano plot was plotted by R package “ggplot2” [19], while the Venn diagram was created using the Venn diagram web tool (http://bioinformatics.psb.ugent.be/webtools/Venn/). The KEGG analysis was visualized by R package “ggplot2”. The nomogram and calibration curves were analyzed via R package “rms” [20]. Univariate and multivariate Cox regression analyses were performed via R package “survival” [12]. A P-value < 0.05 indicated statistical significance.

Results

Construction of anti-tumor immune gene signature (ATIGS)

By comparing HNSCC samples and normal samples, we identified 4788 differentially expressed genes, comprising 3604 up-regulated genes and 1184 down-regulated genes (Fig. 1a). Meanwhile, a total of 988 genes related to anti-tumor immune (GRATIs) were collected from TISIDB. By taking the intersection, 307 differentially expressed GRATIs were extracted and subjected to KEGG/GO functional enrichment analysis (Fig. 1b). As shown in Fig. 1c, d, the most related signaling pathway was determined to be “cytokine-cytokine receptor interaction”, while “immune response” emerged as the most related biological process. Among these differentially expressed GRATIs, a subset of 49 genes were deemed to be associated with HNSCC prognosis and underwent LASSO regression analysis. Ultimately, an optimal signature comprising 14 GRATIs was constructed, including HOXA1, FCGR2A, PTX3, GAST, AURKA, ULBP2, HSP90B1, PALU, HSPB8, CLEC2D, DNMT1, TCL1A, IL-34, and ZAP70 (Fig. 1e, g). The risk score of patients was defined as multiplying the expression levels of these 14 genes by their respective regression coefficients as follows: risk score = [Expression level of HOXA1 * 0.093147] + [Expression level of FCGR2A * 0.0407079)] + [Expression level of PTX3 * 0.0274496] + [Expression level of GAST * 0.0081514] + [Expression level of AURKA * 0.0072085] + [Expression level of ULBP2 * 0.0008793] + [Expression level of HSP90B1 * 0.0008104] + [Expression level of PALU * 0.0005161] + [Expression level of HSPB8 * (− 0.00278)] + [Expression level of CLEC2D * (− 0.006798)] + [Expression level of DNMT1 * (− 0.014939)] + [Expression level of TCL1A * (− 0.017532)] + [Expression level of IL-34 * (− 0.046477)] + [Expression level of ZAP70 * (− 0.144047)].

Fig. 1.

The optimal signature consisted of 14 GRATIs. a Volcano plot showing differentially expressed genes between head and neck squamous cell carcinoma and normal tissues. b Venn diagram visualizing differentially expressed GRATIs. c The top 10 most significantly enriched KEGG signaling pathways. d The top 10 most significantly enriched GO terms. e LASSO coefficient profiles in LASSO regression analysis. f Selection of the tuning parameter (lambda) in LASSO analysis through tenfold cross-validation. g Coefficients of 14 GRATIs used to construct the signature

Evaluation of the predictive ability of ATIGS

To determine the predictive capability of ATIGS, we computed the risk score for each patient and classified them into high- and low-risk groups across the training set, the validation set, and two testing sets (Fig. 2a). The Kaplan–Meier survival curves demonstrated that low-risk group had better outcomes across all four datasets (Fig. 2b, P < 0.05). Details of median survival and HR for each dataset were presented in Table S3. Furthermore, the time-dependent ROC curves indicated that ATIGS exhibited promising performance in monitoring survival in multiple datasets, with the area under curves (AUCs) of 0.736, 0.705, and 0.694 for 1-year, 3-year, and 5-year survival, respectively, in the TCGA training set (Fig. 2c). Additionally, correlation analysis between ATIGS and multiple clinicopathologic variables was performed via independent t-tests. As depicted in Fig. 2d, the high-risk group was associated with increased mortality rates, larger tumor burdens, an increased proportion in advanced stages, and was mostly HPV-negative, which was consistent with previous literature.

Fig. 2.

Predictive performance of ATIGS in TCGA training set, TCGA validation set, entire TCGA cohort, and GSE65858 cohort. a Risk score distribution, survival status, and expression patterns of 14 GRATIs in both high- and low-risk groups. b Kaplan–Meier survival curves of OS for HNSCC patients. c Time-dependent ROC curves of ATIGS. d Correlation analysis between ATIGS and multiple clinicopathological factors

Establishment of ATIGS-based prognostic nomogram

To better predict the prognosis of HNSCC patients, we integrated ATIGS with multiple clinicopathological factors to construct a nomogram, providing a quantitative analysis tool for clinical practice. Our results revealed that ATIGS is an independent prognostic factor, significantly associated with overall survival (OS) in both univariate and multivariate analyses, along with tumor burden and N stage (Fig. 3a, P < 0.05). Additionally, pathological stage was significantly associated with OS in univariate analysis, so we also analyzed the efficiency of ATIGS against the stage (Fig. S1 and Table S4). Ultimately, based on ATIGS and two clinicopathological factors, we constructed a prognostic nomogram that accurately evaluates the probability of survival of HNSCC patients at 1-, 3-, and 5- years (Fig. 3b). Further, the calibration curves of the nomogram showed good consistency between the predicted and actual survival probability of HNSCC patients at 1-, 3-, and 5- years (Fig. 3c). Importantly, the nomogram showed the highest net benefit at most risk thresholds (Fig. 3d). The above results demonstrated that the nomogram performed well in predicting the survival probability of 1-, 3-, and 5-year of HNSCC patients.

Fig. 3.

Construction and calibration of prognostic nomogram. a Univariate and multivariate Cox regression analyses of ATIGS and other clinicopathological variables for OS. b Nomogram for predicting the survival probability of 1-, 3-, and 5-years. c Calibration curves of prognostic nomogram. d Net benefits at different risk thresholds

ATIGS was closely related to the immune microenvironment of HNSCC

The immune microenvironment is closely associated with tumor development and patient prognosis, so we evaluated the infiltration of 22 immune cells in HNSCC patient samples using the CIBERSORT algorithm. Our analysis revealed significant differences in the infiltrating abundance of most immune cell types between the high- and low-risk groups, including six adaptive immune cell types and six innate immune cell types (Fig. 4a). Moreover, correlation analysis indicated that the risk score was positively correlated with six types of immune cells and negatively correlated with nine types of immune cells (Fig. 4b). Based on the results of differential analysis and correlation analysis, we identified 11 types of immune cells, including six adaptive immune cells (native B cells, plasma cells, CD8⁺ T cells, resting memory CD4⁺ T cells, follicular helper T cells, and regulatory T cells) and five innate immune cells (resting NK cells, M0 macrophages, M2 macrophages, resting mast cells, and activated mast cells), which were thought to contribute to the stratification of HNSCC patients by ATIGS (Fig. 4c). These results suggested that ATIGS are closely related to the immune microenvironment of HNSCC patients, and the abundance of immune cell infiltration is critical for patient stratification.

Fig. 4.

Overview of immune cell infiltration. a Differential analysis of immune cell infiltration abundance in high- and low-risk groups. b Correlation analysis between immune cell types and risk score. c Venn diagram revealing the 11 intersected cell types between differential analysis and correlation analysis

ATIGS have potential to predict the response to immune checkpoint blockers

Immune checkpoints refer to a series of molecules that are expressed on immune cells and are responsible for regulating the activation of the immune system. Usually, tumor cells induce abnormal expression of immune checkpoints to escape from the immune system, so immune checkpoint blockers (ICBs) are currently one of the most widely used immunotherapies for cancer treatment. To this end, we initially assessed the expression of 50 immune checkpoints in both high- and low-risk groups. As shown in Fig. 5a, our finding demonstrated a significant correlation between the expression of most immune checkpoints and risk scores, highlighting the ability of ATIGS to broadly predict immune checkpoint expression levels. Given that normal cancer-immune cycles is the prerequisite for successful clinical response to ICBs, we also analyzed the activity status of the seven-step anticancer immune responses. As shown in Fig. 5b, ATIGS were negatively correlated with most anticancer immune responses. Additionally, patients in the low-risk group exhibited higher immunophenotypic scores, suggesting greater immunogenicity and a heightened likelihood of responding to ICBs (Fig. 5c). Furthermore, the predictive performance of ATIGS was validated on datasets of cancer patients who were previously treated with PD-1 blockade. Due to the small sample size of HNSCC patients who had been treated with PD-1 blockade, we selected other cancer cohorts with larger sample sizes for supplemental studies. The results showed that patients who responded to PD-1 blockade received a lower risk score in our scoring system, indicating that ATIGS possess promising potential in forecasting the response to immune checkpoint blockers (Fig. 5d).

Fig. 5.

The potential of ATIGS in predicting immune checkpoint blockers’ response. a The expression of 50 immune checkpoints in high- and low-risk groups. b Correlation matrix of ATIGS and anticancer immune responses. c Immunophenotypic scores of patients in high- and low-risk groups. The correlation of risk score with PD-1 blockade response over three cohorts, represented separately d and merged e. *P < 0.05; **P < 0.01; ***P < 0.001; ns, no significance

Identification of the key target ZAP70 and screening of candidate compounds

To identify the key targets of ATIGS, a protein–protein interaction (PPI) network was constructed, and topology analysis was performed. The results revealed that ZAP70 not only belongs to the core node of the PPI network, but also ranks high in all three centrality algorithms, which indicated that ZAP70 may be the key target of ATIGS (Fig. S2 & Table S5). For this key target, large-scale compound screening was conducted to identify potential combination compounds for immunotherapy. Therefore, ~ 30,000 compounds from six databases were selected for virtual screening and molecular docking (Fig. 6a). A lower docking score and lower binding free energy imply a more stable binding of the ligand compound to the target protein. Among the ~ 30,000 compounds, the ones with higher affinity for ZAP70 included Puerol A, 4′-Hydroxywogonin, and Cirsimaritin (Fig. 6b). As shown in Fig. 6c, d, Puerol A formed one hydrogen bond with the key residues TYR292 and GLU415, as well as two hydrogen bonds with ALA417 of ZAP70. The 3D interaction diagram of 4′-Hydroxywogonin with ZAP70 active site showed that it formed a hydrogen bond with residues ASP479, ALA417, GLU295, and ALA297, respectively (Fig. 6e, f). Besides, Cirsimaritin relied on forming a hydrogen bond with residues ALA417 and GLU295 to maintain its good binding to ZAP70 (Fig. 6g, h).

Fig. 6.

Screening and docking of candidate compounds. a Virtual screening flowchart. b The XP docking score and binding free energy of the top 10 compounds with ZAP70. c, d 2D and 3D diagrams of Puerol A binding to the active site of ZAP70. e, f 2D and 3D diagrams of 4′-Hydroxywogonin binding to the active site of ZAP70. g, h 2D and 3D diagrams of Cirsimaritin binding to the active site of ZAP70

Inhibitory effects of candidate compounds on tumor cell proliferation

Based on the candidate compounds obtained from the virtual screening, we first selected Top3 compounds for validation. CAL-27 and SCC7 cells were pretreated with different concentrations of candidate compounds, and then CCK-8 assay was used to detect the effects of these three compounds on tumor cell proliferation. As shown in Fig. 7a, d, Cirsimaritin produced a certain inhibitory effect on the proliferation of CAL-27 cells at 40 μM, but had no effect on SCC7 cells. In contrast, 4′-Hydroxywogonin could inhibit the proliferation of CAL-27 and SCC7 cells in a dose-dependent manner (Fig. 7b, e). Besides, Puerol A inhibited the proliferation of CAL-27 cells at different concentrations, while lacked inhibitory effect on SCC7 cells (Fig. 7c, f). These results suggested that 4′-Hydroxywogonin had the best inhibitory effect on tongue squamous cell carcinoma cells.

Fig. 7.

The effects of Top3 candidate compounds on CAL-27 and SCC7 cell viability. CAL-27 cells were incubated with various concentrations of Cirsimaritin a, 4′-Hydroxywogonin b, and Puerol A c. SCC7 cells were incubated with various concentrations of Cirsimaritin d, 4′-Hydroxywogonin e, and Puerol A f. Values are expressed as mean ± SEM. *P < 0.05, **P < 0.01, ****P < 0.0001 versus control group

Effect of 4′-Hydroxywogonin on the expression levels of tumor immune-related genes

Based on the established gene signature ATIGS, we identified the key target ZAP70 by PPI network analysis. ZAP-70 is an essential kinase in the T-cell signaling pathway and a master regulator of adaptive immunity. LAT, a downstream gene of ZAP70, is a T-cell activation adaptor factor. When phosphorylated, LAT can activate T cells and recruit multiple adaptor factors and effector molecules. To further evaluate the potential effect of 4′-Hydroxywogonin on tumor immunity, we examined the expression levels of immune-related genes ZAP70 and LAT in CAL-27 cells treated with different concentrations of 4′-Hydroxywogonin. As shown in Fig. 8a, the expression level of ZAP70 was significantly upregulated by 4′-Hydroxywogonin at 15 μM. Meanwhile, 4′-Hydroxywogonin dose-dependently upregulated LAT gene expression (Fig. 8b). These results suggested that 4′-Hydroxywogonin may exert its anti-tumor effect by regulating tumor immune-related pathways.

Fig. 8.

Effect of 4′-hydroxywogonin on gene expression levels in CAL-27 cells. a Gene expression levels of ZAP70. b Gene expression levels of the downstream gene LAT. Values are expressed as mean ± SEM. **P < 0.01, ***P < 0.001 versus control group

Validating analysis of 4′-Hydroxywogonin binding to ZAP70 using MD simulation

Proteins may undergo significant conformational changes during interactions with drug molecules, while molecular dynamics (MD) simulations may help to understand the conformational changes and the stability of protein–ligand complexes. Therefore, the binding mode of 4′-Hydroxywogonin with ZAP70 was analyzed by MD simulations for 100 ns and its molecular dynamics trajectory was analyzed. Figure 9a showed the root-mean-square deviation (RMSD), where small fluctuations indicate that the complex has obtained a stable conformation. The result suggested that 4′-Hydroxywogonin and ZAP70 protein are relatively stable after 65 ns, from which time the system is in equilibrium. Root-mean square fluctuation (RMSF) characterized the local changes of the protein chain, and the high peak value indicated that the protein region fluctuates greatly during the simulation. Figure 9b revealed that after binding to 4′-Hydroxywogonin, ZAP70 protein exhibited high structural flexibility in the residue regions of 260-270AA, 280-290AA, and 480-500AA. In addition, the interaction between 4′-Hydroxywogonin and ZAP70 mainly depended on hydrogen bonding and hydrophobic interaction, and the amino acids important for their binding included TYR292, GLU295, LEU344, LYS369, GLU415, ALA417 and LEU468 (Fig. 9c). The binding pattern diagram in Fig. 9d shows that, 4′-Hydroxywogonin directly formed hydrogen bonds with residues ALA417 (98%), GLU295 (88%), GLU415 (64%), LYS369 (73%), ASP479 (22%), and TYR292 (22%) of ZAP70 protein, as well as indirectly formed hydrogen bond with residue PRO296 (23%) through a water bridge. What’s more, the conformational evolution of each rotatable bond (RB) of the ligand in the whole simulation trajectory was summarized by the ligand torsion diagram (Fig. 9e, f). Taken together, these results further indicated a stable binding of 4′-Hydroxywogonin to ZAP70.

Fig. 9.

Binding stability of the 4′-Hydroxywogonin-ZAP70 complex assessed by MD simulations for 100 ns. a RMSD plot and b RMSF plot of 4′-Hydroxywogonin-ZAP70 complex as function of time. c Amino acid residues that contribute to the binding of ZAP70 protein to 4′-Hydroxywogonin during MD simulation. d The binding pattern diagram of 4′-Hydroxywogonin to ZAP70 protein generated from the last frame of the MD simulation. e, f Graphics of the torsional conformation of each rotatable bond of 4′-Hydroxywogonin during MD simulation

Discussion

HNSCC is a highly invasive malignant tumor with special sites, that seriously affects the basic physiological function and quality of life of patients. Most patients lost the opportunity for surgery when they were diagnosed in the middle and late stages, which led to the limited efficacy of traditional treatment methods. As a subversive achievement, immunotherapy has the potential to improve the survival and quality of life of patients. In order to predict the prognosis of HNSCC patients, some studies have established models based on genes related to autophagy [21], hypoxia [22], senescence [23], pyroptosis [24], or ferroptosis [25]. However, these factors have a weak correlation with tumor-related immunity. In order to better predict the prognosis of HNSCC patients and find a model that can simultaneously predict the response to immunotherapy, a new gene signature closely related to anti-tumor immunity is urgently needed.

As is known to all, in the process of tumor occurrence and development, the immune system controls the growth of tumor cells through immune monitoring, while tumor cells escape from the attack of the immune system through immune escape. In this interaction between tumor and immune system, anti-tumor immune-related genes are considered to be very important in regulating the immune microenvironment. Therefore, we have developed a robust prediction model, ATIGS, to predict the prognosis and immunotherapy response of HNSCC patients using anti-tumor immune-related genes. The gene signature ATIGS constructed in this study contained 14 GRATIs, of which 8 genes were considered risk factors and 4 genes were identified as protective factors. Interestingly, most of the genes have been recognized to be associated with various immune cells. For example, HOXA1 is significantly associated with immune cell infiltration and immune checkpoints, and knockdown of HOXA1 can enhance CD8⁺ T cell responses [26]. PLAU expression has also been reported to be associated with infiltration of immune cells, such as a positive correlation with M1 macrophages and a negative correlation with CD4⁺ T cells and Tregs cells [27]. Noteworthily, IL-34 has a powerful immunomodulatory effect in pathological states such as cancer and inflammatory diseases, and exerts effects on a variety of immune cells, including monocytes, macrophages, and regulatory T cells that shape the immune microenvironment [28]. The expression level of HSBP8 is closely related to the infiltration level of B cells, CD4⁺ T cells, and CD8⁺ T cells [29]. Furthermore, AURKA reshapes the immunosuppressive tumor microenvironment by regulating the infiltration level of multiple immune cells [30]. In addition, several genes have been reported to be associated with the regulation of certain types of immune cells. For example, LLT1 is expressed in NK cells, and its overexpression can inhibit NK cell-mediated cytotoxicity [31]. As an important regulator of the immune system, PTX3 is mainly involved in macrophage migration and inflammation resolution [32]. Similarly, FCGR2A encodes an immunoglobulin Fc receptor on phagocytes and is involved in the process of phagocytosis and clearance of immune complexes. The inhibition of DNMT1 can improve the tumor immune microenvironment and inhibit tumor growth by reducing MDSCs and increasing tumor-infiltrating T cells [33]. As a key target, ZAP70 is an essential kinase in T cell signaling and plays a role in T cell development and activation [34, 35]. More importantly, the GSEA results revealed that these 14 genes were enriched in the immune-related pathways and biological processes, suggesting that ATIGS may have potential predictive value of tumor immunotherapy response (Fig. S3).

Immunogenicity refers to the ability of tumor antigens to be recognized by the immune system and trigger an immune response, which can be determined by four factors, including effector cells, immunosuppressed cells, MHC molecules, and immunomodulators [15]. High immunogenicity means stronger immune response and easier attack by the immune system. In present study, ATIGS can divide patients into high-risk group and low-risk group, in which the low-risk group has lower immunogenicity and is more likely to respond to immunotherapy. Additionally, the infiltration of immune cells in tumor immune microenvironment also contributes to immune response. Our results show that ATIGS is significantly related to the infiltration of most immune cells, indicating that it has a promising potential in predicting immune response.

Except for constructing gene signature that can predict the prognosis and immunotherapy response of HNSCC patients, we are also committed to exploring potential therapeutic drugs. Virtual screening is an efficient method for drug screening, which not only helps to improve the discovery efficiency of candidate compounds, but also reduces the experimental cost. Therefore, based on ZAP70, the key target of the gene signature, we performed a virtual screening consisting of six compound libraries with over 30,000 molecules. ZAP-70 is a cytoplasmic protein tyrosine kinase that is required for T-cell signaling and adaptive immunity. The inhibition of ZAP-70 kinase activity contributes to the blockade of T-cell receptor and CD28 superagonist signaling [36]. By virtual screening, we identified three compounds with high affinity for ZAP70, including Puerol A, 4′-Hydroxywogonin, and Cirsimaritin. Among them, 4′-hydroxywogonin has been reported to exert anti-inflammatory effects and reduce angiogenesis in colorectal cancer [37, 38]. Cirsimaritin can not only regulate inflammation by inhibiting the phosphorylation of c-fos and STAT3, but also inhibit the proliferation of lung squamous cell lines by inducing apoptosis [39, 40]. Furthermore, we experimentally proved that 4′-Hydroxywogonin could dose-dependently inhibit the growth of tongue squamous cancer cells, and up-regulate the expression of immune-related gene LAT, suggesting that 4′-Hydroxywogonin may exert its anti-tumor effect by regulating tumor immune-related pathways.

The present study has some advantages worthy of attention. Firstly, our gene signature ATIGS is based on the genes closely related to anti-tumor immunity, which is more in line with the tumor-immune interaction in the tumor immune microenvironment. Secondly, unlike other studies that only predict the prognosis of patients, ATIGS constructed through a more in-depth and comprehensive analysis can predict the response to immunotherapy while predicting the prognosis of HNSCC patients. Finally, this study conducted a large-scale virtual screening based on gene signature to find potential therapeutic drugs, and also carried out experimental validation. Nevertheless, this study lacks clinical sample validation of gene signature and in-depth mechanistic investigation of the candidate compound, which will be further investigated in the future.

In conclusion, we constructed a gene signature ATIGS based on anti-tumor immunity-related genes, which can predict the prognosis and immunotherapy response of HNSCC patients. Based on the key target ZAP70 of this signature, combined with virtual screening, molecular docking, experimental validation and molecular dynamics simulation, 4′-Hydroxywogonin was considered as a potential immunotherapy combination drug. Notably, ATIGS will provide a solid foundation for clinical individualized treatment of HNSCC patients, and the potential immunotherapy combination compound 4′-Hydroxywogonin will also provide a promising therapeutic strategy for clinical treatment.

Author contributions

W.Q., Y.D., and W.G. conceived the study; W.Q. and Y.D. designed and performed the experiments; W.Q. wrote the manuscript; Y.D., L.W., J.W., and W.G. edited the manuscript. All authors read and gave final approval to submit the manuscript.

Funding

This study was supported by the National Natural Science Foundation of China (No. 82304986), Jiangsu Funding Program for Excellent Postdoctoral Talent (No. 2023ZB803) and the China Postdoctoral Science Foundation (No. 2023M741653).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval and consent to participate

Not applicable.

Consent for publication

Authors are in agreement with the submission of their article for publication.

Competing interests

The authors declare no competing interests.