The data of 146 patients with OPSCC who underwent initial surgery or biopsy at the Department of Otolaryngology, Head and Neck Surgery, Keio University School of Medicine between April 2005 and September 2018 were retrospectively reviewed. A total of three patients who had received chemotherapy and one patient who had received radiotherapy for the head and neck region before the initial surgery or biopsy were excluded. A total of five patients who underwent biopsy only and were treated at other hospitals were also excluded. Consequently, 137 patients were enrolled in the present study. The cohort included 116 men and 21 women. Their median age was 64 years (range, 37–87 years). Their characteristics (age, sex, smoking and alcohol status, tumor subsite, TNM classification, pathologic characteristics, and follow-up examination findings) were extracted from their medical records. The TNM classification was determined based on the eighth edition of the American Joint Committee on Cancer staging manual (20). The histopathological diagnoses were based on the World Health Organization criteria (21). Tissue samples were obtained from the hospital tissue bank.

The treatment strategy for OPSCC in the present study was as follows. Early T-stage (T1 and T2) tumors were treated with transoral resection or radiotherapy alone. Advanced T-stage (T3 and T4) tumors were treated with concurrent chemoradiotherapy (chemotherapy: Cisplatin 80 mg/m², administered 2–3 times every 3 weeks; radiotherapy: 2.0 Gy/fraction, administered five times a week for a total dose of 60–66 Gy) or radical tumor resection with functional reconstruction. The patients with multiple LN metastases underwent neck dissection before radiotherapy or concurrently with primary surgery. The patients with resectable locoregional recurrences or neck metastases detected during follow-up underwent additional resections immediately. They received chemotherapy as palliative treatment for persistent disease or distant metastases if they were amenable to treatment according to the NCCN guidelines (4).

The paraffin sections were sliced at 5 µm each, dewaxed in xylene, rehydrated in ethanol, and washed in water. Heat-induced antigen retrieval was performed using the Decloaking Chamber NxGen^® (Biocare Medical, LLC) with Dako Target Retrieval Solution (cat. no. S1700; Dako; Agilent Technologies, Inc.) at 95°C for 1 h. Endogenous peroxidase activity was quenched by treatment with 3% hydrogen peroxide solution for 10 min. Non-specific binding of the primary antibodies was blocked by treatment with normal goat serum (cat. no. IHR-8136; ImmunoBioScience/IBC) in phosphate-buffered saline (PBS) containing TWEEN^®20 for 1 h at room temperature. The slides were stained with the primary antibody (1:500) overnight at 4°C. They were washed and incubated with SignalStain^® Boost IHC Reagent (HRP; Rabbit; cat. no. 8114; Cell Signaling Technology, Inc.) for 30 min at room temperature. Peroxidase activity was visualized using a DAB Substrate Kit (cat. no. SK-4100; Vector Laboratories, Inc.). The sections were counterstained and sealed with hematoxylin solution. The stained slides were imaged using a light microscope. The tissue images were imported, and the proportions of ANGPTL4-positive tumor cells were determined using Tissue Studio^® (Definiens). The tumorous area was selected as the region of interest for each slide. The following parameters were set: Hematoxylin threshold of 0.2, typical nucleus size of 23 µm², maximum cell growth of 5, and classification of 0.15. ANGPTL4 expression was automatically calculated as the ratio of the number of ANGPTL4-positive tumor cells to the total number of tumor cells.

FaDu, Detroit 562, OSC-19, HSC-2 and HSC-4 (human HNSCC cell lines) were obtained from the American Type Culture Collection.

The cell lines were cultured in Eagle's Minimum Essential Medium (Sigma-Aldrich; Merck KGaA) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (solution stabilized; Sigma-Aldrich; Merck KGaA). They were incubated in a humidified incubator at 37°C in a 5% carbon dioxide environment. The cells were sub-cultured continuously in accordance with the American Type Culture Collection protocol.

RNA was extracted using the RNeasy Mini Kit (Qiagen GmbH), and its quality was assessed using the Nanodrop 1000 (Thermo Fisher Scientific, Inc.). Complementary DNA synthesis was performed using the Superscript III First-Strand Synthesis System (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. RT-qPCR was conducted using the StepOnePlus Real-Time PCR system and software (Applied Biosystems; Thermo Fisher Scientific, Inc.) following the manufacturer's protocol. The primers and probes were procured from Applied Biosystems (TaqMan^® Gene Expression Assays) and had the following IDs: ANGPTL4 (Hs01101127_m1), ACTB (Hs01060665_g1), MKI67 (Hs04260396_g1), BAX (Hs0018269_m1), BCL2 (Hs00608023_m1), and CASP3 (Hs00234387_m1). According to the manufacturer's product information, the sequences for these pre-designed assays are not publicly disclosed. PCR amplification involved initial denaturation at 95°C for 20 sec, followed by 40 cycles of denaturation at 95°C for 3 sec and annealing at 60°C for 30 sec. The relative mRNA expression levels were determined using the 2^−ΔΔCq method and were compared with those of ACTB, which served as the endogenous control (22).

The FaDu cells were seeded in a 6-well dish at a density of 25,000 cells/ml and incubated in a medium containing 10% FBS for 24 h. The medium was changed, and the cells were transfected with 30 nM siRNA targeting ANGPTL4 (cat. no. NM_001039667; ID: SASI_Hs01_00144609; Sigma-Aldrich; Merck KGaA) and negative control (MISSION siRNA Universal Negative Control #1; Sigma-Aldrich; Merck KGaA) using 4 µl of Lipofectamine 2000 Transfection Reagent (Thermo Fisher Scientific, Inc.) in 200 µl of Opti-MEM medium (Thermo Fisher Scientific, Inc.) for 30 min at room temperature. The Opti-MEM medium was removed, and 2 ml of fresh culture medium was added. The cells were incubated for an additional 24 h, scraped, and collected for analysis. The efficiency of ANGPTL4 knockdown was evaluated using RT-qPCR. According to the manufacturer's product information, the sequences of the siRNAs used in the present study are not publicly disclosed.

In addition to FaDu cells, ANGPTL4 knockdown experiments were attempted in other HNSCC cell lines, including Detroit 562, OSC-19, HSC-2 and HSC-4, using the same siRNA sequence and transfection protocols.

The FaDu cells were seeded in a 96-well plate (3.000 cells per well) and incubated overnight after transfection. Their viability was subsequently assessed using a CellTiter-Glo 2.0 luminescence-based assay kit (cat. no. G9241; Promega Corporation). The results were normalized to those of the negative control (set as 1.0).

The FaDu cells were seeded in slide chambers (CHAMBER SLIDEII IWAKI, http://iwaki.atgc.co.jp) after ANGPTL4 knockdown for immunofluorescence analyses of ANGPTL4 and Ki-67. The cells were washed extensively with PBS and fixed with 4% paraformaldehyde for 15 min at room temperature. The samples were washed with PBS and blocked with 10% normal goat serum in PBS for 1 h. The cells were incubated with the primary antibodies overnight at 4°C. They were subsequently incubated with Alexa Fluor Plus 555- and 488-conjugated secondary antibodies (cat. nos. A32732 and A32723; Thermo Fisher Scientific, Inc.) against ANGPTL4 and Ki-67, respectively, and observed 24 h after transfection. Hoechst 33258 (Sigma-Aldrich) was used for nuclear staining. The numbers of ANGPTL4- and Ki-67-positive cells were determined based on counts in four randomly selected areas at ×10 magnification using a BZ-X710 fluorescence microscope (Keyence Corporation).

The primary antibodies used for immunostaining were anti-ANGPTL4 (rabbit; 1:500; cat. no. ab115798; Abcam) for immunofluorescence staining and IHC and anti-Ki-67 (mouse; 1:100; cat. no. ab245113; Abcam) for immunofluorescence staining.

RNA sequencing and corresponding clinical data for HNSCC were obtained from TCGA via the University of California Santa Cruz Cancer Browser (https://xenabrowser.net/). A total of 78 patients with primary OPSCC and corresponding clinical data from the cohort labeled ‘GDC TCGA Head and Neck Cancer’ were included. The patients were divided into high and low groups based on their ANGPTL4 expression levels (FPKM-UQ). The prognoses of 128 and 116 patients with tongue and laryngeal SCCs, respectively, were also evaluated. The ANGPTL4 expression data for 520 patients with HNSCC were extracted based on the subsites. Gene set enrichment analysis (GSEA) was performed based on the ANGPTL4 expression to explore its biological role in OPSCC. GSEA was performed using the Hallmark gene set from the Molecular Signatures Database (version 7.5.1, http://www.gsea-msigdb.org/gsea/msigdb/index.jsp). The enrichment score, normalized enrichment score, nominal P-value and false discovery rate (FDR) were determined using UCSC Xena (https://xenabrowser.net/). Statistical significance was set at FDR <0.05. The differentially expressed genes (DEGs) were extracted based on a fold change of >2 or <-2 and adjusted P-value of <0.05.

The five-year OS and disease-free survival (DFS) rates of the patients with OPSCC were determined using the Kaplan-Meier method, and survival curves were compared using the log-rank test. When survival curves crossed, weighted tests (Breslow and Tarone-Ware) were additionally performed as sensitivity analyses. The survival durations were calculated from the date of initial treatment to the date of the event or the latest follow-up. The variables included were age, sex, smoking and alcohol status, T classification, N classification, M classification, TNM stage, p16 status, initial definitive therapy and ANGPTL4 status. To determine the optimal cut-off value for ANGPTL4 expression, receiver operating characteristic (ROC) curve analysis was performed using the OS. The cut-off value was defined as the point that maximized Youden's index. The ROC curve is provided in Fig. S1. To evaluate the robustness of the selected cut-off value, sensitivity analyses were performed using alternative thresholds for ANGPTL4 expression. Tumors were reclassified using the median expression level, as well as fixed cut-off values of 10 and 15%. The OS and DFS of the patients in the subgroups were compared using the log-rank test during the univariate analysis. The factors that were significant in the univariate analysis were further analyzed using multivariate analysis. This was performed using a Cox proportional hazards model with backward elimination. The correlations between the variables analyzed in the multivariate analysis were examined using Pearson's correlation coefficient to avoid multicollinearity. The relationships between ANGPTL4 and the other variables were evaluated. The distributions of the categorical variables for ANGPTL4 and the other variables were compared using the chi-squared test. In addition, survival analyses were performed stratified by p16 status to evaluate the prognostic impact of ANGPTL4 within each subgroup. Furthermore, sensitivity analyses were performed using the multivariable Cox models in patients who underwent definitive therapy. Associations between continuous variables were analyzed using unpaired Student's t-test. The differences in ANGPTL4 expression across the subsites were determined using one-way analysis of variance. Post hoc tests were not performed. Statistical analyses were performed using SPSS version 27 for Mac (IBM Corp.). P<0.05 was considered to indicate a statistically significant difference.

The characteristics of the 137 patients with OPSCC included in the present study are presented in Table I. In total, 40 had locally advanced disease (T3 and T4), 101 had cervical lymph node metastases, 2 had distant metastases, and 39 had advanced TNM stages (III and IV). A total of ~80% of the patients were positive for p16 staining. The IHC staining for ANGPTL4 is shown in Fig. 1. Totally, 71 and 66 patients had positive and negative staining for ANGPTL4, respectively. The 5-year OS and DFS rates were 75.1 and 67.6%, respectively. The median follow-up duration was 52 months (range, 1–171 months). The univariate analyses of OS and DFS based on the clinicopathological factors are summarized in Table II. T classification (T3 and T4), M classification (M1), TNM stage (III and IV) and ANGPTL4 (<7.7%) were significantly associated with poorer OS (P=0.001, P<0.001, P=0.034 and P=0.002, respectively). The subsite (anterior wall), T classification (T3 and T4), M classification (M1), TNM stage (III and IV) and ANGPTL4 (<7.7%) were significant predictors of worse DFS in patients with OPSCC (P=0.007, P<0.001, P<0.001, P=0.007, and P<0.001, respectively). Multivariate analysis revealed that T classification [hazard ratio (HR)=3.209; 95% confidence interval (CI), 1.605–6.415; P<0.001], M classification (HR=94.613; 95% CI, 15.385–581.837; P<0.001) and ANGPTL4 (HR=3.676; 95% CI, 1.678–8.056; P=0.001) were independent prognostic factors for OS. The subsite (HR=2.161; 95% CI, 1.183–3.947; P=0.012), T classification (HR=3.029; 95% CI, 1.652–5.550; P<0.001), M classification (HR=15.081; 95% CI, 3.083–73.771; P<0.001) and ANGPTL4 (HR=2.959; 95% CI, 1.533–5.713; P=0.001) were independent prognostic factors for DFS in patients with OPSCC (Table III). The Kaplan-Meier curves for OS and DFS stratified by ANGPTL4 expression are shown in Fig. 2. The 5-year OS rates for patients who tested positive and negative for ANGPTL4 expression were 88.4 and 61.6%, respectively (P=0.002). The corresponding 5-year DFS rates were 82.7 and 52.5%, respectively (P<0.001). Sensitivity analyses using alternative cut-off values including median ANGPTL4 expression level, 10%, and 15% demonstrated consistent results (Table SI). In addition, baseline characteristics stratified by p16 status are summarized in Table SII. High ANGPTL4 expression was associated with improved OS and DFS in the p16-positive subgroup, whereas no significant prognostic impact was observed in the p16-negative subgroup (Table SIII). In sensitivity analyses restricted to patients who received definitive therapy, additional adjustment for p16 status (model 1) and for initial definitive therapy (model 2) did not alter the prognostic association of ANGPTL4 with the OS and DFS (Table SIV).

The correlations between ANGPTL4 expression and clinicopathological factors in OPSCC are provided in Table IV. No significant correlations were observed between ANGPTL4 expression and age, sex, subsite, stage, or p16 expression.

The FaDu cells were transfected with siRNA targeting ANGPTL4, and their proliferation was evaluated compared with the negative controls. Their relative mRNA expression levels were determined using RT-qPCR. The relative mRNA expression level of ANGPTL4 decreased to 38%. The results of immunofluorescence staining are provided in Fig. 3. The knockdown cells had significantly reduced ANGPTL4 expression and significantly increased Ki-67 expression compared with the negative controls (P=0.021 and 0.021, respectively).

The FaDu cells with ANGPTL4 knockdown showed significantly increased proliferation compared with the negative controls (P=0.010) (Fig. 4).

ANGPTL4 knockdown was attempted in additional HNSCC cell lines (Detroit 562, OSC-19, HSC-2 and HSC-4); however, robust and reproducible knockdown could not be achieved. Therefore, functional analyses were performed only in FaDu cells.

The relative gene expression levels of various mRNAs in the FaDu cells transfected with ANGPTL4 siRNA were determined via RT-qPCR (Fig. 5). The relative mRNA expression levels of ANGPTL4 decreased to 38%, and those of MKI67 increased significantly. The expression of apoptosis-related genes was also evaluated. The expression of BAX decreased, and that of BCL2 increased. In addition, the expression of CASP3 increased in the ANGPTL4 knockdown cells.

The correlation between ANGPTL4 expression and prognosis was analyzed using TCGA data. High ANGPTL4 expression was associated with a tendency towards improved prognosis in 78 patients with OPSCC, but the difference was not statistically significant (P=0.09; Fig. 6A). The group with high ANGPTL4 expression had significantly worse OS than the group with low expression in tongue SCC (log-rank, P=0.02; Breslow, P=0.003; Tarone-Ware, P=0.005; Fig. 6B). No significant differences in OS were detected between the patients with laryngeal SCC with high and low ANGPTL4 expressions (log-rank, P=0.45; Breslow, P=0.47; Tarone-Ware, P=0.49; Fig. 6C). The mean ANGPTL4 expression differed significantly across HNSCC subsites (P=0.042; Fig. 6D). The OPSCCs, including those of the base of the tongue, oropharynx and tonsils, had lower ANGPTL4 expression levels than those of other subsites.

GSEA revealed significant enrichment of the HALLMARK_E2F_TARGETS and HALLMARK_G2M_CHECKPOINT gene sets in the ANGPTL4-low expression group (NES=6.86 and 4.42, respectively; FDR <0.001 for both) (Table V). The HALLMARK_TNFA_SIGNALING_VIA_NFKB, HALLMARK_INTERFERON_GAMMA_RESPONSE, and HALLMARK_INFLAMMATORY_RESPONSE gene sets were significantly depleted in the ANGPTL4-low expression group. NDRG1 and CDH3 were identified as DEGs based on their fold changes (>2 or <-2) and adjusted P-values (<0.05) (Table VI).