Comprehensive analysis illustrating the role of HOXB8 in head and neck squamous cell carcinoma: evidence from multi-omics analysis and experiments validation

Abstract

Background

HOXB8 is implicated in various cancers. However, the effect pattern of HOXB8 in head and neck squamous cell carcinoma (HNSCC) remains unclear.

Methods

Open-access transcriptional profiles, clinical information, and mutational data were downloaded from the Cancer Genome Atlas database. R software was used for all analysis based on public data through specific R packages. Western blot and real-time quantitative PCR was used to detect the protein and RNA level of HOXB8, respectively. In vivo and in vitro experiments were conducted to explore the effect of HOXB8 on HNSCC cells.

Results

Here, we discovered that HOXB8 was upregulated in HNSCC tissue and associated with worse clinical outcomes (clinical stage and prognosis). Results indicated that HOXB8 was primarily distributed in the nucleoplasm. Results of cell lines indicated that HOXB8 is upregulated in HNSCC cells. Further experiments, both in vitro and in vivo, revealed that the suppression of HOXB8 can markedly curb the proliferation, invasion, and migration capabilities of HNSCC cells. Results of biological enrichment and western blot indicated that HOXB8 can regulate the PI3K/AKT/mTOR and EMT pathways. It also came to our attention that HOXB8 could modulate the tumor microenvironment in HNSCC. We observed that patients with high HOXB8 expression had lower infiltration levels of CD8 + T cells but higher infiltration levels of M2 macrophages. Finally, we developed a prognostic model based on molecules derived from HOXB8 (ADD2, SYT1, PXYLP1, MRPL33).

Conclusions

Our study contributes to the existing knowledge on HOXB8 in HNSCC, which may inform future research directions.

Introduction

Head and neck squamous cell carcinoma (HNSCC) is a common malignant tumor worldwide, with a high degree of malignancy and susceptibility to early metastasis [1]. Head and neck cancer encompasses wide range of sites, such as the nasal cavity, oral cavity, pharynx, and throat [2]. Studies have shown that the occurrence of head and neck cancer is related to high alcohol intake, tobacco, and so on [3]. In the absence of methods for detecting and assessing risk at an early stage, more than 50% of HNSCC patients are diagnosed in advanced stages [4]. Currently, the treatment options for HNSCC are mainly surgical treatment, supplemented by radiotherapy and chemotherapy [5]. However, there are currently no extremely effective treatment measures for HNSCC, which has led to most patients still experiencing recurrence and metastasis after surgery, and their late survival rate is poor [6]. Biomarkers can be used for prognostic evaluation in clinical practice, as well as for estimating disease risk, and screening for occult primary cancer. Effective cancer markers can be used for early diagnosis and the development of clinical treatment strategies. To find effective biomarkers for HNSCC, researchers have conducted extensive research.

HOXB8, whose full name is homeobox B8, is widely involved in various pathological and physiological processes. For example, Liu et al. observed that HOXB8 could enhance the malignancy degree of ovarian cancer cells [7]. In gastric cancer, HOXB8 could facilitate tumor metastasis through the EMT process and ZEB2 [8]. Ying et al. discovered that HOXB8 could be induced by a super-enhancer, further promoting colon cancer invasion through BACH1 [9]. Zhang et al. observed that the HOXB8 could be regulated by LINC01006/miR-2682-5p axis, further promoting pancreatic cancer proliferation and metastasis [10]. However, the role of HOXB8 has not been explored in HNSCC.

Here, we illustrated the effect pattern of HOXB8 in HNSCC. We found that HOXB8 was upregulated in HNSCC tissue and associated with poor clinical performance (clinical stage and prognosis). Immunofluorescence results indicated that HOXB8 was predominantly localized in the nucleoplasm. Results of cell lines indicated that HOXB8 is upregulated in HNSCC cells. Further experiments, both in vitro and in vivo, revealed that the suppression of HOXB8 can markedly curb the proliferation, invasion, and migration capabilities of HNSCC cells. We subsequently conducted a biological enrichment analysis to shed light on HOXB8’s function in HNSCC. It also came to our attention that HOXB8 could modify the tumor microenvironment in HNSCC. We observed that patients exhibiting high HOXB8 expression had lower infiltration levels of CD8 + T cells, but higher infiltration levels of M2 macrophages. Finally, we devised a prognostic model based on molecules derived from HOXB8 (ADD2, SYT1, PXYLP1, MRPL33).

Methods

Collection of open-accessed data

The open-accessed data of HNSCC patients (transcriptional profile, clinical information, and genomic mutation features) were downloaded from The Cancer Genome Atlas Program (TCGA) database, HNSCC project– 522 tumor and 44 normal samples. The original expression spectrum data was in “STAR - Counts” form and then converted to “TPM” form. To obtain more stable results, all data were preprocessed before data analysis. Meanwhile, the probes with median value less than 0.5 were also deleted. Clinical data was downloaded in the “bcr - xml” form [11]. We procured the data pertaining to the subcellular localization of HOXB8 from The Human Protein Atlas (HPA) database [12]. Information regarding drug sensitivity (IC50) was obtained from the Genomics of Drug Sensitivity in Cancer (GDSC) database through an online website (https://www.home-for-researchers.com/) [13]. The acquisition of tumor stemness index mRNAsi and EREG-mRNAsi was conducted from a previous study [14].

Bioinformatic analysis

Using the median value as the cut-off for distinguishing patients with high and low expression of HOXB8. The limma package served as a tool for differentially expressed genes (DEGs) analysis with the threshold of|logFC| > 1.5 and P value < 0.05 [15]. The input data for DEGs consists of the expression profile data of patients with high and low expression of HOXB8. Enrichment analyses, including clueGO, gene set enrichment analysis (GSEA), and single-sample GSEA (ssGSEA) were utilized to assess the biological enrichment [16,17,18]. The input data for pathway enrichment is the analysis results of DEGs, including specific genes and logFC values. For clueGO analysis, a plugin of Cytoscape, the “Ontologies/Pathways” was set as “Biological Process”. The GSEA analysis was undertaken using the clusterProfiler and fgsea packages [19]. GSVA package was utilized to conduct ssGSEA analysis. Briefly, enrichment score can be quantified using ssGSEA based on a list of molecules. Estimate package was utilized to quantify the content of immune and stromal cells [20]. Meanwhile, the HNSCC microenvironment was quantified using multiple algorithms, including CIBERSORT, EPIC, MCPCOUNTER, QUANTISEQ, TIMER, and XCELL [21,22,23,24,25]. The input of tumor microenvironment analysis is the transcription profiling data of HNSCC. The HOXB8-derived molecule was identified through linear correlation analysis. Univariate Cox regression analysis was performed to identify the prognosis-related genes analysis with P < 0.05. The input of univariate Cox regression analysis is the combination of specific gene expression and survival information. Then, LASSO regression analysis was performed to reduce data dimensions and optimize variables [26]. Multivariate Cox regression analysis was used to identify the final variables for the prognosis model [27].

Cell lines

Normal human squamous epithelial cells (NOK) and three HNSCC cells (HN-4, SCC-4, and CAL-27) were bought from the Cell Bank of the Chinese Academy of Sciences and routinely stored in the laboratory. These cells were routinely stored in the laboratory and cultivated under conventional conditions.

Real-time quantitative PCR

Total RNA was extracted using a total RNA extraction kit (AC0201, SparkJade, China) according to the corresponding protocol and then reverse-transcribed into cDNA using a Synthesis Kit (R223, Vazyme, China). The 20µl of cDNA obtained through reverse transcription according to the standard protocol, was diluted to 100µl with nuclease-free water for subsequent experiments. Real-time Quantitative PCR was conducted using the Sybr Green system according to standard procedures (20µl system). The primers used were as follows: HOXB8, forward, 5’-GTCCCTGCGCCCCAATTATTA-3’; reverse; 5’-GCCCGTGGTAGAACTCCTG-3’; GAPDH, forward, 5’-GGAGCGAGATCCCTCCAAAAT-3’; reverse, 5’-GGCTGTTGTCATACTTCTCATGG-3’.

Cell transfection

Cell transfection was conducted using Lipofectamine 3000 (L3000001, ThermoFisher) according to standard procedures. Specifically, cell transfection was conducted using Lipofectamine 3000. Detailed, cells were initially cultured in a 6-wells-plate to 70–80% confluency, ensuring optimal conditions for transfection. Before transfection, the Lipofectamine 3000 reagent and the plasmid DNA of interest were prepared separately in serum-free medium to form complexes. Specifically, for each transfection, Lipofectamine 3000 reagent was gently mixed with a specified volume of Opti-MEM without any antibiotics. Simultaneously, the genetic material intended for delivery into the cells was also diluted in a separate tube with Opti-MEM. After 10 minutes at room temperature, the diluted Lipofectamine 3000 and the nucleic acid solutions were combined and incubated for an additional 20 minutes to allow complex formation. Following complex formation, this mixture was added directly to the cells. After 6 hours, the transfection medium was replaced with fresh culture medium containing serum and antibiotics. Construction of stably transduced cell lines using a three-plasmid system to package lentivirus (pspax2, pMD2G and pLKO-1). The target sequence of HOXB8 was as follows: sh-NC: 5’-GGTTCTCCGAACGTGTCACGT-3’; sh#1: 5’- GCTCTTATTTCGTCAACCTCACTGTTCTCC-3’; sh#2: 5’-GGGCAATTGTTACAAGTGT-3’; sh#3: 5’-CGCAAATCCAGGAGTTCTA-3’.

Cell proliferation assays

Cell Counting Kit-8 (CCK8), colony formation, and 5-ethynyl-2’-deoxyuridine (EdU) assay were used to evaluate the cell proliferation ability of cancer cells, which were conducted according to the standard procedures. For the CCK-8 assay, cancer cells were seeded in 96-well plates at a precise density of 5,000 cells per well and allowed to adhere overnight. At designated time points post-treatment (24 h, 48 h and 72 h), CCK-8 solution was added to each well, followed by incubation for 2 h at 37 °C (C0075S, Beyotime, China). The absorbance at 450 nm was measured using a microplate reader. In the colony formation assay, 500 cancer cells were plated in each well of a 6-well plate and cultured for 14 days to allow colony formation, with the medium refreshed every three days. After the incubation period, colonies were fixed with methanol and stained with 0.5% crystal violet. For the 5-Ethynyl-2ʹ-Deoxyuridine (EdU) assay, cells were incubated with 10 µM EdU for 2 h to integrate EdU into the DNA of dividing cells. Following incubation, cells were fixed and subjected to a click chemistry reaction, which attaches a fluorescent dye to the incorporated EdU (C0078S, Beyotime, China).

Transwell assay

Transwell assay and wound-healing assays were utilized to evaluate the invasion and migration ability of HNSCC cells, which were performed according to standard procedures. The Transwell migration assay assesses the ability of cells to move through a porous membrane. For this assay, HNSCC cells were seeded (SCC-4: 5 × 10^4 cells; CAL-27: 2 × 10^5 cells) into the upper chamber of Transwell inserts without a Matrigel coating, to avoid impeding non-invasive cell movement. The upper chamber was filled with serum-free medium to induce cell migration, while the lower chamber contained medium with 20% fetal bovine serum (FBS) as a chemoattractant. After specific hours (SCC-4: 24 h; CAL-27: 48 h) of incubation at 37 °C, cells that remained on the upper surface of the membrane were removed, and those that migrated to the lower surface were fixed with methanol, stained with 0.1% crystal violet, and quantified under a microscope. This migration assay quantitatively measures the migratory capacity of HNSCC cells. To specifically assess cell invasion, a Transwell invasion assay was conducted by seeding cells (SCC-4: 5 × 10^4 cells; CAL-27: 2 × 10^5 cells) into the upper chamber of inserts coated with Matrigel. After 48 h, non-invading cells were wiped from the upper side of the membrane, whereas invading cells on the lower side were fixed, stained, and counted.

Wound-healing assay

For the wound-healing assay, HNSCC cells were cultured in 6-well plates until they formed a confluent monolayer. A sterile pipette tip was then used to scratch a “wound” across the cell layer. After washing away detached cells, the culture was continued in medium with reduced serum (1% FBS) to focus on migration over proliferation. Images taken at 0 and 24 h provided data on the rate of wound closure.

Immunofluorescence

Immunofluorescence on cells was performed following a protocol to ensure visualization of protein expression and localization. Initially, cells were seeded at a density of 1 × 10^5 per well on glass coverslips in 6-well plates and allowed to adhere overnight at 37 °C in a 5% CO2 atmosphere. Subsequent steps included fixing the cells with 4% paraformaldehyde in PBS for 15 min at room temperature, followed by permeabilization with 0.1% Triton X-100 in PBS for 10 min. Blocking was conducted with 5% BSA in PBS for 1 h to minimize non-specific antibody binding. The cells were then incubated with a primary antibody at a dilution of 1:200 in blocking solution overnight at 4 °C (HOXB8, ab125727, Abcam) followed by washing with PBS and incubation with a fluorescently labeled secondary antibody at a dilution of 1:500 for 1 h in the dark. Nuclei were stained with DAPI at 1 µg/mL for 5 min, and coverslips were mounted on slides using anti-fade mounting medium. Fluorescence microscopy was employed to capture the detailed images of the stained cells.

Western blot

Western blot analysis was performed with a detailed protocol using a PVDF membrane. Starting with cell lysate preparation, cells were lysed using RIPA, with protease and phosphatase inhibitors, followed by sonication and centrifugation at 14,000 x g for 15 min at 4 °C. Protein concentrations were determined via BCA Protein Assay, aiming for 30 µg per lane on an SDS-PAGE gel. After electrophoresis at 80 V for the stacking gel and 120 V for the resolving gel, proteins were transferred to a PVDF membrane in a cold transfer buffer. The membrane was blocked with 5% non-fat milk in TBST, incubated with primary antibody (AKT, 1:5000, 60203-2-Ig, proteintech; p-AKT, 1:2000; 66444-1-Ig, proteintech; mTOR, 1:5000, 66888-1-Ig, proteintech; p-mTOR, 1:2000, 67778-1-Ig, proteintech; PI3K, 1:2000; ab151549; Abcam; p-PI3K, 1:1000; ab138364; Abcam; cyclinD1, 1:5000, 60186-1-Ig, proteintech; N-cadherin, 1:5000, 66219-1-Ig, proteintech; E-cadherin, 1:5000, 60335-1-Ig, proteintech; GAPDH, 1:10000, 60004-1-Ig, proteintech; HOXB8, 1:1000, ab125727, Abcam) overnight at 4 °C, followed by a horseradish peroxidase-conjugated secondary antibody (1:5000 in 5% milk in TBST) for 1 h at room temperature. Detection was achieved using an enhanced chemiluminescence (ECL) system.

In vivo experiments

Mice were purchased from SPF (Beijing) Biotechnology Co., Ltd. Tumor formation in nude mice was conducted using male BALB/c nude mice that were five weeks old (18.0–20.0 g). Each mouse was subcutaneously inoculated with 10 × 10^5 cells in their back. Afterward, nude mice were raised under conventional conditions. After 24 days, all nude mice were euthanized by carbon dioxide overdose, followed by cervical dislocation, and the tumor bodies were removed to be weighed. Based on standard procedures, immunohistochemistry (IHC) was used to detect Ki67 levels in the xenograft tissue (anti-Ki67 antibody, 1:200 dilution). The colonization ability in the lung of cancer cells was realized through tail vein injection (3 × 10^6 cells). Four weeks later, the mice were euthanized, and their lung tissue was stained with hematoxylin-eosin (HE).

Statistical analysis

Statistical evaluations were performed utilizing SPSS and R software. For data exhibiting normal distribution, T-tests were employed for comparisons, whereas data with non-normal distribution were analyzed using Mann-Whitney-U tests.

Results

Figure 1 illustrated the brief flow chart of this study.

Fig. 1

The flow chart of the whole study

The clinical role of HOXB8 in HNSCC and its expression pattern

We noticed that HOXB8 was significantly overexpressed in HNSCC tissue (Fig. 2A). Furthermore, we performed a clinical correlation analysis of HOXB8 in HNSCC. Results showed that HOXB8 was upregulated in patients with worse clinical stage (Fig. 2B: T3-4 compares with T1-2; Fig. 2C: N1-3 compares with N0). Kaplan-Meier survival curves showed that HOXB8 expression might be associated with poorer outcomes in HNSCC patients (Fig. 2D, HR = 1.49, P = 0.004; Fig. 2E, HR = 1.27, P = 0.174; Fig. 2F, HR = 1.34, P = 0.041). Meanwhile, the immunofluorescence results of SCC-4 and CAL-27 cells showed that HOXB8 was mainly localized in the cell nucleoplasm (Fig. 2G).

Fig. 2

Expression pattern and clinical role of HOXB8 in HNSCC. Notes: A: Expression level of HOXB8 in paired HNSCC tissue, *** = P < 0.001; B: Expression level of HOXB8 in T1-2 and T3-4 patients, ** = P < 0.01; C: Expression level of HOXB8 in N0 and N1-3 patients, ** = P < 0.01; D: Kaplan-Meier survival curves of HOXB8 in HNSCC (overall survival); E: Kaplan-Meier survival curves of HOXB8 in HNSCC (disease-specific survival); F: Kaplan-Meier survival curves of HOXB8 in HNSCC (progression-free survival); G: Subcellular localization of HOXB8 in SCC-4 and CAL-27 cells detected by immunofluorescence

HOXB8 promotes the proliferation, invasion and migration of HNSCC cells

Through in vitro experiments, we explored the role of HOXB8 in HNSCC cells. Results of real-time quantitative PCR showed that HOXB8 has a higher expression level in HNSCC cells (Fig. 3A, HN-4, SCC-4, CAL-27 vs. NOK cell). Results of western blot also indicated the same trend (Figure S1). Among these three HNSCC cells, SCC-4 and CAL-27 cells have the highest HOXB8 expression. Therefore, we selected SCC-4 and CAL-27 to perform HOXB8 knockdown. Results of real-time quantitative PCR and western blot indicated that sh#HOXB8-2 has the best knockdown efficiency (Fig. 3B-C and Figure S2). Cell proliferation of SCC-4 and CAL-27 was significantly suppressed by inhibiting HOXB8 according to CCK8 and colony formation assays (Fig. 3D-F). The knockdown of HOXB8 resulted in a significant reduction in EdU-positive cells in EdU assays (Fig. 3G). Transwell assay indicated that the HOXB8 knockdown group has remarkably reduced invasion and migration cells (Fig. 4A). The result of wound healing indicated that SCC-4 and CAL-27 cells with HOXB8 knockdown have a slower movement rate (Fig. 4B).

Fig. 3

HOXB8 promotes the proliferation ability of HNSCC cells. Notes: A: Expression level of HOXB8 in normal and HNSCC cell lines, *** = P < 0.001; B-C: Real-time quantitative PCR was used to detect the knockdown efficiency of HOXB8 in SCC-4 and CAL-27 cells, * = P < 0.05, *** = P < 0.001; D-E: CCK8 assay were performed in sh#ctl and sh#HOXB8 cells (SCC-4 and CAL-27 cell lines), ** = P < 0.01, *** = P < 0.001; F: Colony formation assay was performed in sh#ctl and sh#HOXB8 cells (SCC-4 and CAL-27 cell lines), *** = P < 0.001; G: EdU assay was performed in sh#ctl and sh#HOXB8 cells (SCC-4 and CAL-27 cell lines), *** = P < 0.001

Fig. 4

HOXB8 promotes the invasion and migration ability of HNSCC cells. Notes: A: Transwell assay was performed in sh#ctl and sh#HOXB8 cells (SCC-4 and CAL-27 cell lines), *** = P < 0.001; B: Wound-healing assay was performed in sh#ctl and sh#HOXB8 cells (SCC-4 and CAL-27 cell lines)

HOXB8 facilitates HNSCC progression in vivo.

We have preliminarily demonstrated through in vitro experiments that HOXB8 can promote the malignant biological behavior of HNSCC cells. Then, we tried to validate our results in vivo. We selected SCC-4 cells for further experiments. The tumor-forming experiment in nude mice indicated that the knockdown of HOXB8 could significantly suppress the tumor growth rate. We found that weight of extracted tumor in the sh#HOXB8 group was remarkably lower than that in the sh#ctl group (Fig. 5A-B). Interestingly, we found that the cells in the mock group had better proliferation ability than the cells in the sh#ctl group. This may be because the process of constructing the stable transfection strain affected the cell status of the sh#ctl cells. Ki67 staining results showed that the IHC of sh#HOXB8 had a significantly lower proportion of Ki67-positive cells than the control group (Fig. 5C-D). Results of the tail vein injection model showed that the sh#HOXB8 group had a lower number of tumor foci in the lungs, indicating that the knockdown of HOXB8 can hamper the colonization ability in the lungs of cancer cells (Fig. 5E).

Fig. 5

HOXB8 accelerates tumor growth and metastasis. Notes: A-B: Tumor-forming experiment in nude mice was performed based on the mock、sh#ctl and sh#HOXB8 SCC-4 cells, ** = P < 0.01; C-D: Ki67 staining of the tumor body, the part outlined in red is the 5X field of view; E: Lung metastasis model was performed based on the sh#ctl and sh#HOXB8 SCC-4 cells

Biological enrichment of HOXB8 in HNSCC

In the subsequent investigation, we aimed to elucidate the biological impact of HOXB8 on HNSCC. Subsequently, we conducted a comprehensive analysis of DEGs between HNSCC patients exhibiting high and low HOXB8 expression, as illustrated in Fig. 6A. Following this, a series of biological enrichment analyses were performed to unravel the functional implications. Results of clueGO analysis indicated that these DEGs were mainly enriched in embryonic skeletal system development, thyroid gland development, establishment of the skin barrier, positive regulation of striated muscle cell differentiation, positive regulation of synaptic vesicle exocytosis, regulation of action potential, and keratinization (Fig. 6B). Furthermore, results from ssGSEA indicated a positive correlation between HOXB8 and biological processes like fat digestion and absorption, synthesis and degradation of ketone bodies, and spliceosome (Fig. 6C). GSEA analysis (Hallmark) indicated that HOXB8 could activate the pathways of epithelial-mesenchymal transition (EMT), PI3K/AKT/mTOR signaling, but inhibit the pathways of interferon-alpha response, cholesterol homeostasis (Fig. 6D). Validation through Western blot analysis demonstrated a significant suppression of the PI3K/AKT/mTOR and EMT pathways upon HOXB8 inhibition (Fig. 6E). Additionally, GSEA analysis encompassing GO and KEGG pathways unveiled positive correlations between HOXB8 and processes such as steroid hormone biosynthesis, cell fate specification, regionalization, and pattern specification process (FigureS3A-F).

Fig. 6

Biological enrichment analysis. Notes: A: DEGs analysis was performed in patients with high and low HOXB8 expression, 77 downregulated and 431 upregulated; B: ClueGO analysis of these DEGs; C: ssGSEA analysis was used to identify the correlation between specific pathways and HOXB8; D: GSEA analysis based on Hallmark gene set; E-G: GSEA analysis based on KEGG gene set; E: Western blot was used to detect the key protein involved in PI3K/AKT/mTOR and EMT signaling in the sh#ctl and sh#HOXB8 cells (SCC-4 cells)

HOXB8 reshapes the tumor microenvironment of HNSCC

The tumor microenvironment is also important for cancer progression. Based on the results quantified by the estimate package, we noticed that HOXB8 was negatively correlated with immune score, but not the stromal and estimate score (Fig. 7A-C). Then, we used multiple algorithms to further quantify the tumor microenvironment of HNSCC (Fig. 7D). Our observations indicated that patients exhibiting elevated HOXB8 expression demonstrated a diminished infiltration of CD8 + T cells, but conversely, a heightened infiltration of M2 macrophages (Fig. 7E-H).

Fig. 7

HOXB8 can reshape the HNSCC microenvironment. Notes: A: Correlation between HOXB8 and immune score; B: Correlation between HOXB8 and stromal score; C: Correlation between HOXB8 and estimate score; D: TIMER, XCELL, EPIC, MCPCOUNTER and CIBERSORT algorithms were used to quantify the tumor microenvironment of HNSCC tissue, the * with red color = positive correlation; E: Level of CD8 + T cells_CIBERSORT in patients with high and low HOXB8 expression, *** = P < 0.001; F: Level of M2 macrophages_CIBERSORT in patients with high and low HOXB8 expression, *** = P < 0.001; G: Level of CD8 + T cells_EPIC in patients with high and low HOXB8 expression, *** = P < 0.001; H: Level of M2 macrophages_EPIC in patients with high and low HOXB8 expression, *** = P < 0.001

A prognosis model based on the HOXB8-derived molecules

Moreover, to bolster the prognostic significance of HOXB8 in HNSCC, we attempted to establish a prognostic model based on HOXB8-associated molecules. To begin with, we identified the molecules that held significant correlations with HOXB8 through the correlation analysis (Figure S4A-B; Supplementary file 1). Subsequently, in the training cohort, a univariate Cox regression analysis was performed to identify molecules with significant correlations to HNSCC patient survival at a threshold of P < 0.1 (Supplementary file 2). This was followed by the application of LASSO regression to minimize data dimensions and enhance variable optimization (Figure S4C). Multivariate Cox regression analysis eventually pinpointed four genes (ADD2, SYT1, PXYLP1, and MRPL33) for the construction of the prognostic model, which was formulated as “Riskscore = ADD2 * 0.1271 + SYT1 * 0.2456 + PXYLP1 * -0.4335 + MRPL33 * 0.4222” (Figure S4D). A broad view of our prognostic model in the training cohort is depicted in Figure S5A, which shows a higher frequency of fatal cases among patients with high-risk scores. Kaplan-Meier survival curves indicate that high-risk patients potentially demonstrate poorer survival outcomes compared to those at lower risk (Figure S5B, HR = 3.99, P < 0.001). The ROC curves reveal our model’s satisfactory performance in predicting patient survival (Figure S5C-E, with AUC values for 1, 3, and 5-year periods being 0.774, 0.771, and 0.787, respectively). Furthermore, our prognostic model also displayed strong performance in the validation cohort (Figure S5F-J, with AUC values for 1, 3, and 5-year periods being 0.732, 0.731, and 0.680, respectively).

Discussion

HNSCC is the sixth largest malignant tumor worldwide, causing approximately 350,000 deaths annually [28]. HNSCC mostly occurs in the mucosa of the mouth and throat. The location of the disease is relatively hidden [29]. The early symptoms are not typical clinically, and there is a lack of a biomarker that can detect the disease early. Therefore, it is not easy to be detected early. More than half of the patients with HNSCC are found in the middle and late stages of the disease [30]. At this time, most patients have lymph nodes and/or distant metastasis, which poses great difficulties to the clinical treatment of the disease. Although comprehensive treatments including surgery, radiotherapy, and chemotherapy have been used in recent years to improve the disease-free survival rate of HNSCC, the survival rate of advanced HNSCC patients hovers around 30% [31]. Therefore, it is necessary to further understand the biological mechanism of HNSCC and identify predictive molecular markers, which will help in the early diagnosis of diseases and individualized treatment.

In this study, we discovered that HOXB8 was upregulated in HNSCC tissue and associated with worse clinical performance (clinical stage and prognosis). Results indicated that HOXB8 was primarily distributed in the nucleoplasm. Results of cell lines indicated that HOXB8 is upregulated in HNSCC cell. Further experiments, both in vitro and in vivo, revealed that the suppression of HOXB8 can markedly curb the proliferation, invasion, and migration capabilities of HNSCC cells. We subsequently conducted a biological enrichment analysis to shed light on HOXB8’s function in HNSCC. It also came to our attention that HOXB8 could modify the tumor microenvironment in HNSCC. We observed that patients exhibiting high HOXB8 expression had lower infiltration levels of CD8 + T cells, but higher infiltration levels of M2 macrophages. Finally, we devised a prognostic model based on molecules derived from HOXB8 (ADD2, SYT1, PXYLP1, MRPL33).

As a member of the antp homeobox family, HOXB8 is widely involved in various pathological and physiological processes. Salmanidis et al. found that in the presence of IL-3, the downregulation of HOXB8 can cause cell cycle arrest and apoptosis in most cells, indicating its role in carcinogenesis [32]. Wilmerding et al. found that HOXB8 could exert a tumor suppressor molecule to counteract tumor formation induced by ERK [33]. However, Li et al. found that the knockdown of HOXB8 could remarkably suppress the proliferation and migration of colon cancer cells [34]. Interestingly, Stavnes et al. noticed that the expression level of HOXB8 in ovarian serous carcinoma effusions could indicate the prognosis of patients [35]. Chu et al. found that HOXB8 could affect neutrophil signaling and functions [36]. These results indicated that the effect of HOXB8 on cancers might be affected by different factors, rather than solely depending on the type of cancer.

In our study, we performed biological enrichment of HOXB8 and its association with key genomic features through comprehensive GSEA analysis. Our findings highlight the pivotal role of HOXB8 in modulating several critical pathways implicated in cancer biology. Specifically, the analysis revealed that HOXB8 is capable of activating pathways associated with the EMT process and PI3K/AKT/mTOR signaling, while simultaneously inhibiting pathways related to interferon-alpha response and cholesterol homeostasis. The activation of EMT by HOXB8 underscores its potential role in promoting cancer metastasis, tumor stemness, and drug resistance. This process is characterized by the loss of epithelial characteristics and the acquisition of mesenchymal traits, facilitating tumor cell migration and invasion [37, 38]. For instance, Luo et al. demonstrated the capacity of melatonin to counteract tumor proliferation by impeding the EMT process and downregulating PD-L1 expression, highlighting the EMT’s significance in immune evasion mechanisms [39]. Similarly, Jiang et al. uncovered that the EMT could be modulated by the LINC00460/PRDX axis, linking it directly to cancer progression [40]. Furthermore, Chang et al. illustrated that CLDN1 could influence EMT via the AMPK/TGF-β signaling pathway, thereby impacting the development of HNSCC [41]. Parallelly, our analysis pointed to the activation of PI3K/AKT/mTOR signaling by HOXB8. This pathway is a cornerstone in cancer biology, involved in cell growth, proliferation, survival, and metabolism. The involvement of HOXB8 in enhancing this signaling pathway suggests a mechanism by which HOXB8 may promote oncogenic processes. In the context of HNSCC, the significance of PI3K/AKT/mTOR signaling cannot be overstated, with numerous studies, including a comprehensive review by Su et al., identifying its blockade as a strategy to augment radiosensitivity in HNSCC cells. This underscores the potential therapeutic value of targeting HOXB8-mediated pathways [42]. This underscores the potential therapeutic value of targeting HOXB8-mediated pathways.

Moreover, our observations indicated that patients exhibiting elevated HOXB8 expression demonstrated a diminished infiltration of CD8 + T cells, but conversely, a heightened infiltration of M2 macrophages. It is generally believed that CD8 + T cells have an inhibitory effect on killing cancer cells in solid tumors. Zhang et al. noticed that the exhaustion of CD8 + T cells leads to the poor prognosis of HNSCC patients [43]. Also, M2 macrophages were regarded as a tumor-promoting factor in solid tumors [44]. Dan et al. found that RACK1 could increase the ratio of M2/M1 macrophages through the NF-κB signaling, further promoting cancer progression [45]. Moreover, Wang et al. found that lncRNA HOTAIR could promote the EMT process and cancer metastasis, as well as M2 macrophage polarization [46]. These results indicated that the reshaping of HOXB8 on HNSCC might be partly responsible for the cancer-promoting role of HOXB8.

Some limitations still need to be noted. Firstly, the reliance on bioinformatics tools and specific cell lines may not fully capture the biological complexities or represent all HNSCC subtypes, potentially introducing biases and limiting generalizability. Secondly, our experimental design assesses short-term effects, and longer-term studies are necessary to evaluate the enduring impact of HOXB8 on clinical outcomes and therapy resistance. Additionally, while our in vivo models offer a glimpse into the potential clinical behavior, they cannot fully replicate the intricate interactions of the human immune system with tumors, nor do they account for confounding factors such as patient demographics or genetic background, which can significantly affect outcomes. The scarcity of advanced-stage clinical data also restricts our ability to directly correlate experimental findings with clinical metastasis stages. Lastly, our analysis, reliant on retrospective data, limits proactive conclusions about the therapeutic targeting of HOXB8, highlighting the need for prospective clinical trials to validate our findings and further detailed mechanistic studies to clarify the molecular pathways involved.