Oral tumor tissues and non-cancerous contralateral oral mucosal tissue samples from 35 patients with OSCC were divided into two parts: one was fixed in formalin and embedded in paraffin for hematoxylin and eosin staining, as well as immunohistochemistry analysis, while the other samples were snap-frozen in liquid nitrogen and kept at −80°C until used for total RNA extraction. All these patients were diagnosed and treated at the Department of Head and Neck Surgery and Otorhinolaryngology, São Vicente de Paulo Hospital, Jundiaí, SP, Brazil. The initial diagnosis was based on clinical findings and confirmed by histopathological analysis of the specimens. A description of clinical–epidemiopathological data from all 35 OSCC patients is summarized in Supplementary Table S1. The HPSE1 mRNA expression levels were analyzed in 31 fresh OSCC and 35 non-cancerous oral mucosal tissues by quantitative PCR (qRT-PCR). Clinicopathological features of the 31 OSCC patients, whose samples were analyzed for HPSE1 mRNA expression, are described in Table 1. Immunohistochemical analysis was also performed to confirm protein expression levels of HPSE1 in 32 of these OSCC tissue samples. After treatment, patients were regularly followed up, and the outcomes were categorized as disease-specific survival, time from treatment initiation until death due to tumor or last known date alive, time from treatment initiation until the diagnosis of the first recurrence (local, regional, or distant), or last known date alive for those without recurrence. Also, normal oral tissues were obtained from five individuals without cancer subjected to standard third molar surgery procedures. Written informed consent was obtained from each individual, and this study was approved by the Institutional Ethics Committee of the Faculty of Medicine of Jundiaí, Jundiaí, SP, Brazil (protocol number 45091715.1.0000.5412).

Heparanase 1 (HPA, HPSE1) immunostaining was performed using the streptavidin–biotin–peroxidase complex method on 32 FFPE oral cancer specimens and five normal oral tissue specimens from non-cancerous individuals. Briefly, staining was performed using the standard method of deparaffinization and gradual rehydration, followed by heat-mediated antigen retrieval in 10 mM citric acid at pH 6.0 in a pressure cooker. The sections were further treated with 3% hydrogen peroxide to block endogenous peroxidase activity. Primary antibody incubation with polyclonal rabbit HPSE1 antibody (bs-1541R, Bioss Antibodies, United States), diluted 1:200, was carried out overnight, followed by the LSAB method (LSAB + System-HRP kit, Dako, United States). Detection was carried out using 3,3′-diaminobenzidine tetrahydrochloride (Sigma-Aldrich, United States) containing the 0.01% H₂O₂ detection system. Control reactions were performed by omission of the primary antibody. Immunoexpression of HPSE1 was analyzed semiquantitatively by the following criteria: the percentage of cytoplasm positivity was graded from 0 to 4 (0: no staining; 1: 1–25% staining; 3: 51–75% staining; 4: 76–100% staining) and the intensity, from 0 to 3 (0: no staining; 1: weak staining; 2: moderate staining; 3: strong staining). Tumor final scores were calculated as the sum of both categories, and the samples were classified as presenting low (0–4) or high (5–7) expression of HPSE1.

Human oral cancer cell lines SCC-4, SCC-9, SCC-15, SCC-25, and HSC-3, originally isolated from oral cancer (OSCC) of the human tongue, were obtained from the American Type Culture Collection (ATCC). The SCC-4, -9, -15, and -25 cells were grown in DMEM plus Ham’s F12 medium (DMEM/F-12), supplemented with 10% fetal bovine serum (FBS, Life Technologies, Inc.), penicillin (100 U/ml), streptomycin (100 μg/ml), and 0.4 μg/ml hydrocortisone. HSC3 human oral cancer cell line was placed in DMEM supplemented with 10% FBS, 100 units/ml penicillin, 100 μg/ml streptomycin, and 2 mM glutamine. The HPSE1-silenced (HPSE1-) and -upregulated (HPSE+) clones derived from the SCC-9 line were cultured in the same maintenance medium with the addition of 0.25 μg/ml puromycin and 5 μg/ml geneticin, respectively, for these cell expansion. Non-transformed human gingival keratinocyte cell line (HGK), obtained from mucosal keratinocytes and found to be spontaneously immortalized, was cultured in a serum-free, low-calcium medium (Gibco’s Keratinocyte-SFM; Invitrogen, United States) containing specific supplements and antibiotics, as previously described (Mäkelä et al., 1998; Rodrigues et al., 2017). Cells were maintained at 37°C in a humidified atmosphere of 5% CO₂.

Considering that SCC-9 cell line had an intermediate expression level of HPSE1 within all four oral cell lines analyzed, we selected this cell line to carry out both gain- and loss-of-function in vitro experiments. Then, transduction of SCC-9 cells with control (scrRNA control cells) or shRNA HPSE1 sequences was performed using HuSH shRNA plasmid panel (short-hairpin RNA) following the manufacturer’s instructions (OriGene Technologies, United States/HPSE1, human, cat. No. TR307138). Briefly, after transduction, the silenced cells were selected in the presence of 1 μg/ml of puromycin (Invitrogen, United States) for 2 weeks. For the gain-of-function strategy, the transduction of SCC-9 was performed using TrueORF cDNA clones and PrecisionShuttle vector system containing the human HPSE1 cDNA clone (ORF with C-terminal GFP tag) (cat. No. RG219659) along with control SSC-9 Mock transduced cells following the manufacturer’s instructions (OriGene Technologies, United States). Then, after transduction, these cells were selected in the presence of 500 μg/ml of geneticin (ThermoFisher, United States) for 2 weeks. The insertion of the plasmids in the clones of transduced cells under- and overexpressing HPSE1 was confirmed by PCR with sequences 5′ primer and 3′ primer of the Origen Kit components before the expansion of the transduced clones.

Real-time quantitative PCR was carried out to determine the efficiency of HPSE 1 silencing and upregulation in oral cancer cell lines and to investigate HPSE1 expression in oral cancer and non-cancerous oral tissue samples. Total RNA was isolated and purified using the TRIzol reagent method in accordance with the manufacturer’s protocol (Invitrogen, United States). Reverse transcription reactions were conducted with a single step with 1 ug of RNA (each sample) using the commercial kit Luna OneStep qPCR (Biolabs) in accordance with the manufacturer’s protocol. All qRT-PCRs were prepared and run in triplicates using the quantitative thermal cycler 7500 Real-Time PCR-System (Applied Biosystem, United States). Gene expressions of HPSE1 were determined by the relative quantification using the 2^−ΔΔCt method, and the housekeeping gene PPIA (cyclophilin A) was used as a reference gene for data normalization. For the HPSE1 expression analysis in the clinical samples, a pool of five normal oral tissues from individuals without cancer was used as the reference for the relative quantification. The amplification conditions were standardized following the manufacturer’s protocol as follows: 10 min at 55°C for enzyme activation, 1 min at 95°C for initial denaturation, followed by 45 cycles of denaturation at 95°C for 10 s, and 60 s at 60°C for pairing and extending the primers. At the end of amplification, an additional dissociation step was included (45 cycles with a decrease of 1°C every 15 s, starting at 95°C) to generate a dissociation curve (melting curve) necessary to confirm the specificity of the amplified product. Additionally, the expression analysis of markers relevant to invasion (i.e., EMT markers—E-cadherin, N-cadherin and vimentin; and ECM remodeling markers—MMP2 and MMP9) and angiogenesis (VEGFA) was carried out in cell lines and HPSE1-silenced and -upregulated clones, using qPCR as previously described (Rodrigues et al., 2017). The sequence of primers used is shown in Supplementary Table S2.

Western blotting was performed to determine the HPSE1 protein expression levels of parental OSCC cell line (WT), and the silenced and upregulated clones. The cells were washed with cold PBS and lysed with a buffer containing 50 mM NaCl, 0.5% Triton X-100, 0.5 mg EDTA/mL, and 10 μL/ml cocktail protease inhibitor (cat. P-8340—Sigma Chemical Co., St. Louis, MO, United States). After centrifugation (10,000×g at 4°C, 10 min), protein concentrations were determined using the Bradford method following the manufacturer’s instructions (Bio-Rad Protein Assay, Bio-Rad, United States). A total of 30 ug of protein was subjected to SDS-PAGE electrophoresis and transferred to nitrocellulose membranes (Bio-Rad, United States), which were blocked with non-fat dry milk (0.1% of PBS with Tween 20) and incubated overnight (4°C) with specific primary antibodies (HPSE1/1:1000, Bioss-Bs-1541R). The membranes were washed and incubated with horseradish peroxidase-conjugated secondary anti-rabbit antibody (1:2000/Cell Signaling Technology), followed by detection with enhanced chemiluminescence (GE Healthcare, United States). Anti-β-actin was employed as the loading control (AC-74, Sigma Aldrich, St. Louis, MO). GAPDH (1:1000/Santa Cruz Biotechnology- Sc-25778) was used as the loading control. Immunoreactive bands were detected using enhanced chemiluminescence (ECL) Western Blotting System (GE Healthcare, United States) by G-Box system (Syngene).

To determine the localization of HPSE1, we used the immunofluorescence assay. Briefly, the parental oral cancer cell line (W)T, stable inhibition clone shRNA HPSE1, and upregulated clone orfRNA HPSE1+ were plated/well at a density of 5 × 10⁴ (24-well plate) containing sterile glass coverslips. The immunofluorescence (IF) samples were fixed in paraformaldehyde (4%), and the primary antibodies were diluted and applied to the cells (1: 250 −1 ml 3% BSA in TBS-T and 4 µL of HPA antibody −1). The secondary antibody solution (1: 1000 −1 ml of 3% BSA in TBS-T, 1 µL of Rabbit Alexa Fluor 488) was added to each cell/well, and then the cells were mounted onto glass slides with Vectashield mounting medium (Vector Labs, United States). The slides were visualized under a Leica fluorescence microscope (LEIKA, Germany) with a 40 × or ×60 × oil immersion objectives.

The distribution of cells in the cell cycle was evaluated by flow cytometry. Cells were incubated in DMEM/F-12 without serum for 24 h and then trypsinized, resuspended in 70% ethanol, treated with 10 μg/ml of RNAse (Sigma-Aldrich, United States), and stained with 50 μg/ml of propidium iodide (Sigma-Aldrich, United States). DNA contents were then analyzed using an FACSCalibur flow cytometer (BD Biosciences, United States).

For the apoptosis assay, cells were resuspended in binding buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 5 mM KCl, 1 mM MgCl₂, and 1.8 mM CaCl₂) containing Annexin V-Alexa Fluor 647 and 7-AAD (BD Biosciences, United States). Cell apoptosis, characterized by positivity for annexin V, was detected using an FACSCalibur flow cytometer (BD Biosciences, United States). A minimum of 10,000 events were analyzed for each sample.

Cell viability was determined using MTT assay (MTT, Sigma), as defined earlier by Alam et al., 2011. Briefly, 20 ml of MTT (5 mg/ml) was added to each well containing a density of 1 × 10⁴ cells [parental OSCC cells, HPSE1-inhibited (shRNA HPSE1-) and -upregulated (orfRNA HPSE1+) clones]. After 3 h of incubation at 37°C, the cell supernatants were discarded, the MTT crystals were dissolved with DMSO, and the absorbance was measured at 570 nm at 12, 24, and 48 h. The percentage of viability was defined as the relative absorbance of transfected versus non-transfected control cells. The experiment was performed three independent times in quintuplicates.

To analyze the movement of the cell toward the artificial space, a scratch was created in a monolayer confluent cell, the oral cancer cell lines, and their respective HPSE1-inhibited and -upregulated clones were plated at 5 × 10⁴ in 12-well plates. When the cell monolayer reached 90–95% confluence, a scratch was made with a tip of 200 µL (time 0). Following 24 and 48 h, the images were recorded by a camera coupled to the inverted microscope (Nikon, Eclipse TS100) using Motic Image Plus 2.0 software.

For the assay, 96-well plates were coated with 10 ug/cm² of type I collagen (Corning, Cat # 354236) in 100 ul of PBS for 24 h at 4°C. A quantity of 200 ul of PBS was used to wash the wells each three times, followed by covering with 200 ul of 3% BSA in PBS for 2 h at 37°C. Control wells were coated with only 3% BSA solution. Then, 3 × 10³ cells were harvested and resuspended in 100 ul DMEM/F12 medium supplemented with 3% BSA, added to each well, and incubated at 37^oC/5% CO₂ for 1 h. After washing the non-adherent cells, the remaining cells were fixed with 4% paraformaldehyde for 15 min and stained with 1% toluidine blue in 1% borax solution. The absorbance was measured at 650 nm by using an Epoch plate reader (Biotek).

Transwell (Corning, United States) assay was carried out to examine OSCC cell invasion. Briefly, the membranes were coated with 50 ul of gelatinous matrix Miogel (2.4 mg/ml) with type I collagen (0.8 mg/ml) (Corning, Cat # 354236) (Salo et al., 2015) and incubated for 12–16 h. Then, the cells were plated in 200 ul of DMEM/F12 medium without FBS (8 × 10⁴ cells per well). As chemo-attractive, 500 ul of complete DMEM/F12 medium was added to the bottom layer of the inserts inside the wells, so the plates were incubated at 37°C/5% CO2 for 72 h. The invasion profile was assessed by carefully removing the formed membrane with cells from the inside of the insert.

The invading cells on the lower side of the insert were fixed in 4% paraformaldehyde for 15 min and stained with 1% toluidine blue in 1% borax solution. Then the cells were eluted in a 1% SDS solution for 5 min. Dye absorbance was measured at 650 nm in the Epoch plate reader (Biotek).

To verify the potential HPSE1 modulation in the capillary structure formation, the oral cancer cells, the lines, and their respective inhibited and upregulated clones were grown in a culture medium of 10% FBS. After 70% confluence, the cells were incubated with SFB 0% for more than 24 h, and the preconditioned medium from each cell was collected. A measure of 50 µl of gelatinous matrix Miogel (2.4 mg/ml) with type I collagen (0.8 mg/ml) (Corning, Cat # 354236) (Salo et al., 2015) was polymerized in 96-well plates. HUVECs were incubated without FBS in RPMI1640 medium for 24 h. The cells were suspended in RPMI-1640 and added to wells coated with Miogel at a density of 1 × 10⁴ cell medium plus 80% the preconditioned medium and incubated at 37°C for 48 h. During the experimental period, cell morphology of capillary structures was observed and photo-documented at 12 and 24 h. The formation of the tube was examined and photographed under an inverted microscope. The quantification of antiangiogenic activity was done using Motic Image Plus 2.0 software and calculated by measuring the length of the tubule wall perimeter of endothelial cells.

The expression of HPSE1 among different tissues was analyzed and compared using the Kruskal–Wallis test with Dunn’s post hoc test, and the receiver operating characteristic (ROC) curve with area under the curve (AUC) was applied to compare the discriminatory ability of the HPSE1 between normal and tumor tissues. A chi-squared test was used to analyze the association of HPSE1 expression and clinicopathological features. The Kaplan–Meier method was used to construct survival curves that were compared using log-rank tests. The Cox proportional hazard model was used for univariate and multivariate survival analysis. For the multivariate analysis, the stepwise method, including the clinicopathological parameters verified in the univariate analysis as potential predictor variables, was performed to characterize the independent survival factors of OSCC patients. In this method, variables are sequentially included, and those that become non-significant are then removed. The final model is built only with variables displaying a p value ≤ 0.05. These statistical analyses were performed using MedCalc® statistical software, version 19.5.3 (MedCalc Software Ltd., Ostend, Belgium). All in vitro assays were performed at least three times, and data were analyzed using one- or two-way analysis of variance with repeated measures (RM-ANOVA) with post hoc comparisons based on Tukey’s multiple comparisons test. These analyses were performed using GraphPad Prism software version 5.0. For all statistical analyses, the significance level was set at 5% (p ≤ 0.05).

The expression of HPSE1 was analyzed by qPCR, and the levels of HPSE1 mRNA were significantly higher in the OSCC cell lines SCC-15 (p < 0.005), SCC-25 (p < 0.005), SCC-9 (p < 0.001), and HSC-3 (p < 0.0001) compared with the immortalized, but not transformed, epithelial cell line HGK (Figure 1A). Western bloting analysis was carried to confirm HPSE1 expression in these OSCC cell lines (Figure 1B). The cell line SCC-9 was selected to carry out functional in vitro assays to investigate the HPSE1 role in oral cancer tumorigenesis. Of note, we checked the nuclear and cytoplasmic HPSE1 protein expression in the SCC-9 cell line; these OSCC cells showed higher HPSE1 expression in the cytoplasm compartment than in the non-malignant HGK cells, where the HPSE1 expression was more restricted to the nuclear fraction (Supplementary Figure S1A).

We also sought to determine HPSE1 mRNA expression levels in 35 fresh samples from OSCC and normal oral mucosa (n = 40). For that, equivalent concentrations of total RNA of these normal tissues were pooled and used as a reference for mRNA relative quantification. A significantly higher level of HPSE1 mRNA was observed in the group of tumors than the controls (p < 0.001, Figure 1B), and the variation on HPSE1 expression levels was very small among control samples. For instance, we also verified the HPSE1 expression pattern in a subset of 33 OSCCs, and in normal oral tissue (Figure 1D), by immunohistochemistry analysis. Interestingly, the immunostaining for tumor cells revealed that HPSE1 upregulation had marked staining intensity in the cytoplasm of OSCC tissues (Figures 1E,F). Immunoreactivity for HPSE1 was also detected in some inflammatory, endothelial, and stromal cells.

We then assessed the association between mRNA expression of HPSE1 and the clinical prognosis of OSCC patients. Patients with HPSE1 overexpression had a significantly poorer disease-specific survival rate than those with lower expression. The probability of death was approximately 70% among patients expressing higher levels of HPSE1, significantly “higher” than the ones expressing lower HPSE1 (37%) (p < 0.05, Figure 2A). The receiver operating characteristic (ROC) curve showed the ability of the HPSE1 higher expression in discriminate OSCC tumors and normal oral samples (AUC = 0.876, p < 0.001) (Figure 2B). Also, recurrence was diagnosed in 30% of patients with higher expression of HPSE1 after up to a 55-month follow-up, compared with 6.25% for those with low HPSE1 expression.

To investigate the association between mRNA expression levels of HPSE1 and clinicopathological features, 35 OSCC patients were divided into HPSE1 lower and higher expressing subgroups, with the median value as the cutoff. Considering the clinicopathological characteristics of tumors (Table 1), the overexpression of HPSE1 in OSCC tissues was significantly associated with moderately and poorly differentiated histological grades of tumors (p = 0.02). Our results also demonstrate a significant correlation between the expression of HPSE1 in tumor cells and cervical lymph node metastasis (pN stage), in that patients with metastasis (N+) had significantly higher HPSE1 expression (p = 0.02). Approximately 81% of the advanced stage tumors (pTNM stages III and IV) demonstrated high levels of HPSE1, while only 18% of pTNM stages I and II tumors presented high HPSE1 expression (p = 0.04). Univariate analysis for disease-specific survival of patients with the OSCC cohort is also provided in Table 2. This analysis revealed that tumor size (pT3+pT4) [7.81 (95% CI 1.78–34.12, p = 0.006)], pTNM advanced stages (III and IV) [3.48 (95% 1.09–11.14, p = 0.035)], and barely lymphovascular invasion and HPSE1 expression were prognostic factors of the presented OSCC cohort. For HPSE1, a hazard ratio (HR) of 3.21 (95% CI 1.00–10.25, p = 0.047) was found for higher HPSE1 mRNA expression about low HPSE1 levels. Similarly, the presence of invasion in lymph vasculature had an HR of 3.88 (95% CI 1.02–14.82, p = 0.047).

It was also observed that the size of the primary tumor (pT stage) was a risk factor for disease-specific survival [4.48 (95% CI 1.38–14.6, p = 0.0126)], but neither the local regional metastasis at diagnosis (pN stage) nor the expression of the HPSE1 mRNA itself (Supplementary Figure S2). To strengthen the prognostic data of the independent factors, we performed univariate and multivariate survival analyses by combining HPSE1 expression levels with the pT stage. For that, we categorized three different scores: score 1: HPSE1 lower expressing samples in association with pT1/2, score 2: HPSE1 lower expressing samples in association with pT3/4 or higher HPSE1 with pT1/2, and score 3: HPSE1 higher expressing samples in association with pT3/4. The multivariate Cox regression analyses of the combination of HPSE1 expression and tumor size (pT) discriminated well and confirmed that overexpression of this enzyme is significantly associated with more aggressive tumors of high rank [HR: 3.22 (95% CI: 1.42–7.32, p = 0.005)] (Figure 2C). Together, these findings suggest that the expression levels of HPSE1 could be used in association with the clinicopathological parameters to better predict the poor prognosis of OSCCs. Considering the closer p-value for HPSE1 expression as an independent factor for OSCC prognosis, future studies should consider increasing the patient sample cohort.

To gain insights into the role of HPSE1 in OSCC progression, HPSE1-silenced and -upregulated cells were achieved from the SCC-9 cell line by lentivirus-mediated shRNA and human GFP-tagged ORF plasmid, respectively. Conventionally, cells transduced with lentivirus carrying an shRNA-targeted sequence to HPSE1 exhibited a significant reduction in both HPSE1 protein (p < 0.001, Figures 3A,B) and mRNA (p < 0.05, Figure 3C) levels in comparison to parental SCC-9 cells (control) or cells transduced with the vector control (scrRNA), non-targeting sequence. Accordingly, cells transduced with the plasmid carrying a specific ORF sequence to enhance HPSE1 expression demonstrated a significant upregulation in both HPSE1 proteins (p < 0.01, Figures 4A,B) and mRNA (p < 0.01, Figure 3C) levels in comparison with parental and mock-transduced control cells. Of note, the cellular fraction analysis of the SCC-9 nuclear and cytoplasm compartments by WB analysis revealed that abrogation of HPSE1 protein expression was markedly reduced in the cytoplasmic fraction of shHPSE1-silenced cells and that ORF-mediated upregulation of HPSE1 enhanced cytoplasmatic expression of HPSE1 (Supplementary Figures 1B,C).

The specific shRNA against HPSE1 drastically impaired the in vitro proliferation of SCC-9 cells (p < 0.01, Figure 5A). Accordingly, HPSE1 downregulation stimulates a significant increase in the cell amount at G0–G1 phases (p < 0.001) and an important reduction at the S phase (p < 0.0001) as compared to controls (Figure 4C). Conversely, orfHPSE1-upregulated cells showed enhancement in proliferation (p < 0.0001, Figure 4A), with marked reduction of cells at G0-G1 phases (p < 0.001) and an increase at the S phase (p < 0.0001, Figure 4C), compared to control cells. Also, HPSE1 upregulation was able to substantially decrease the number of cells in apoptosis (p < 0.0001, Figure 4B). Contrariwise, the number of apoptotic cells was significantly increased in SCC-9 shHPSE1-transfected cells (p < 0.0001, Figure 4B).

During the process of invasion and metastasis, the epithelial aspect of tumor cells is lost as the mesenchymal phenotype is evoked, which characterizes the process known as EMT. To acquire an understanding of the role of HPSE1 in the regulation of EMT and invasion of oral cancer, we examined several features such as motility, invasiveness, adhesive capacity, expression of the epithelial marker E-cadherin, mesenchymal markers N-cadherin and vimentin, and ECM remodeling markers. The adhesive properties of HPSE1 in SCC9 cells were first investigated, and neither upregulation nor downregulation of HPSE1 affected SCC9 cell adhesion on surfaces coated with type 1 collagen (Figure 4D). Next, the scratch wound migration assay showed that upregulation of HPSE1 significantly enhanced the migration (p < 0.001, Figure 5A) and invasiveness (p < 0.001, Figure 5B) of SCC9-orfHPSE1 cells, whereas shHPSE1-silenced cells revealed significantly lower migration and invasion than the controls.

To further characterize the effects of HPSE1 on the expression of EMT and ECM remodeling makers, we examined the modulation of mRNA levels of E-cadherin, vimentin, N-cadherin, Snail, Twist, MMP9, and MMP2 in HPS1-downregulated and -upregulated SCC9 cells by qRT-PCR. Significantly reduced levels of E-cadherin mRNA were observed in HPSE1-upregulated SCC9 cells (p < 0.05, Figure 5C). Concomitantly, HPSE1 increased the mRNA expression of the mesenchymal marker vimentin (p < 0.05), but not N-cadherin, and augmented the metalloproteinases MMP2 (p < 0.01) and MMP9 (p < 0.0001) (Figure 5D). Conversely, shHPSE1-silenced cells had upregulation of E-cadherin mRNA levels (p < 0.01), whereas diminishing expression of N-cadherin and vimentin was observed in these ShRNA SCC-9-transfected cells (Figure 5C) but not significatively compared to control cells. The mRNA expression of Snail, but not Twist, was higher in both down- and upregulated HPSE1-transfected SCC-9 cells, although it was much higher in the cells overexpressing HPSE1 (Figure 5C).

We next assembled an in vitro tube formation assay, using human umbilical vein endothelial cell (HUVEC), to directly assess the possible effects of HPSE1 on oral cancer angiogenesis. HUVECs composed interlaced tubes in Miogel (Salo et al., 2015) (Figure 6). HPSE1-depleted cells demonstrated a significantly lower number of tubes than control cells (p < 0.05, Figures 6A–D). Conversely, HPSE1 upregulation drastically promoted HUVEC tube formation (p < 0.05, Figures 6A–D). Afterward, we assessed whether endogenous HPSE1 modulation could interfere with the expression of the vascular endothelial growth factor A (VEGFA) gene, often associated with angiogenesis, vasculogenesis, and endothelial cell growth induction. As shown in Figure 6E, significantly reduced levels of VEGFA mRNA were observed in HPSE1-silenced SCC9 cells (p < 0.001), whereas markedly higher expression of VEGFA was detected in HPSE1-upregulated cells (p < 0.01) than the controls.