Formalin-fixed paraffin-embedded tissue samples of 80 patients with OTSCC who were surgically treated at the Department of Oral and Maxillofacial Surgery at the Seoul National University Dental Hospital between January 2008 and December 2012 were included in the present study. The age of patients ranged from 27–83 years, with a median age of 56 years. Tumors were staged according to the TNM system recommended by the 8th Edition of the American Joint Committee on Cancer Staging Manual, Head and Neck Section. The clinicopathological characteristics of the patients including age, sex, differentiation status, tumor size (pT), depth of invasion (DOI), LN metastasis (including late metastasis), PNI, TNM stage, and recurrence are shown in Table I. All procedures followed in the present study were in accordance with the guidelines (approval no. CRI 20003) of the Institutional Review Board of the Seoul National University Dental Hospital (Seoul, Korea).

Tissue specimens were sectioned at 4-µm thickness, deparaffinized in Neo-clear (Merck KGaA) for 25 min, and rehydrated by passage through a graded alcohol series for 25 min. Heat-induced epitope retrieval was performed in Target Retrieval Solution pH 9 (Dako; Agilent Technologies, Inc.) for 10 min using a microwave. Endogenous peroxidase was inactivated by incubation in 3% H₂O₂ solution for 10 min. Immunohistochemical staining for L1CAM was performed using a mouse monoclonal anti-human L1CAM antibody (1:100; clone 14.10; cat. no. 826701; BioLegend Inc.) for 1 h at room temperature (RT). Slides were rinsed in DAKO wash buffer and then incubated for 30 min with peroxidase-labeled polymer conjugated to anti-mouse immunoglobulins (EnVision Detection System; Dako; Agilent Technologies, Inc.). The chromogenic reaction was carried out with DAB chromogen. All sections were counterstained with Mayer's hematoxylin for 1 min at RT.

All samples stained with anti-L1CAM antibody were independently evaluated by two oral pathologists (K-YO and H-JY) using a light microscope (BX53; Olympus Corporation). Nerve tissues in each slide were used as an internal positive control. L1CAM expression was considered positive if 10% or more of the tumor cells showed moderate to strong membranous staining, as previously described (12).

The Ca9-22, HSC-2, HSC-3, HSC-4 and SAS cell lines were kindly provided by professor Masanobu Shindoh of Hokkaido University (Hokkaido, Japan). The HN22 cell line was generously provided by the School of Dentistry, DanKook University (Cheonan, Korea). The cell lines from Japan were authenticated by Japanese Collection of Research Bioresources (JCRB) Cell Bank using the SRT profiling. Ca9-22 (cat. no. JCRB0625) cell line was derived from gingiva SCC; HN22 and HSC-2 (cat. no. JCRB0622) cell lines from unknown sites of oral cavity; and HSC-3 (cat. no. JCRB0623), HSC-4 (cat. no. JCRB0624) and SAS (cat. no. JCRB0260) cell lines from tongue SCC. All OSCC cell lines were cultured in Dulbecco's Modified Eagle's Medium (DMEM)/F-12 (Welgene, Inc.) containing 10% fetal bovine serum (FBS; Welgene, Inc.) and 100 U/ml penicillin and streptomycin in a humidified atmosphere with 5% CO₂ at 37°C.

Total RNA was extracted using TRIzol^® Reagent (Thermo Fisher Scientific, Inc.). RNA (1 µg) was reverse-transcribed using an AMPIGENE cDNA Synthesis Kit (Enzo Life Sciences, Inc.) according to the manufacturer's instructions. RT-qPCR for analyzing L1CAM expression level was performed using Applied Biosystems StepOne Plus Real-time PCR System (Applied Biosystems; Thermo Fisher Scientific, Inc.). Two independent experiments were performed in triplicate. The primer sequences used for RT-qPCR were as follows: L1CAM forward, 5′-ACGAGGGATGGTGTCCACTTCAAA-3′ and reverse, 5′-TTATTGCTGGCAAAGCAGCGGTAG-3′; and β-actin forward, 5′-CACTCTTCCAGCCTTCCTTC-3′ and reverse, 5′-AGCACTGTGTTGGCGTACAG-3′. L1CAM mRNA expression was measured using a SYBR Premix Ex Taq™ kit (Takara Bio, Inc.). β-actin expression was used as a reference. The thermocycling conditions suitable for L1CAM mRNA were the following: 95°C for 2 min, followed by 40 cycles of denaturation at 95°C for 10 sec and synthesis at 60°C for 30 sec. The relative level of L1CAM mRNA was normalized to that of β-actin and calculated by the 2^−∆∆Cq method (24).

Whole cells were lysed in RIPA buffer and protein concentration was determined using the DC protein assay kit II (Bio-Rad Laboratories, Inc.). A total of 30 µg protein extract was separated by 10% SDS-PAGE and transferred to a polyvinylidene fluoride membrane. After being blocked by 5% non-fat milk at RT for 1 h, the membranes were incubated with the following primary antibodies diluted at 1:1,000: L1CAM (cat. no. SC-53386; Santa Cruz Biotechnology, Inc.), E-cadherin (cat. no. BD610181; BD Biosciences), N-cadherin (cat. no. BD610920; BD Biosciences), vimentin (cat. no. BD550513; BD Biosciences), Snail (cat. no. CST3879; Cell Signaling Technology, Inc.), Slug (cat. no. CST9585; Cell Signaling Technology, Inc.), Twist (cat. no. SC-81417; Santa Cruz Biotechnology, Inc.), AKT (cat. no. CST9272; Cell Signaling Technology, Inc.), p-AKT (cat. no. CST9271; Cell Signaling Technology, Inc.), and β-actin (cat. no. SC-47778; Santa Cruz Biotechnology, Inc.). After incubation with goat anti-rabbit (1:3,000; cat. no. GTX213110-01; GeneTex, Inc.) or anti-mouse secondary antibody (1:3,000; cat. no. GTX213111-01; GeneTex, Inc.) at RT for 2 h, the proteins were identified by SuperSignal West Pico Chemiluminescent Substrate (cat. no. SC-2048; Santa Cruz Biotechnology, Inc.), and immunoreactive bands were visualized using ImageQuant LAS 500 (Cytiva). Densitometric analysis of western blot bands was carried out using ImageJ software (version 1.51 k; National Institutes of Health).

Recombinant human L1CAM (rhL1CAM) was purchased from R&D Systems, Inc. Cell viability was first assessed at concentrations of 0, 50, 100, 200, and 400 ng/ml, and a titer of 100 ng/ml of rhL1CAM was used for subsequent experiments.

ON-TARGETplus SMARTpool small interfering RNA (siRNA) targeting L1CAM (siL1CAM) and Non-targeting Control Pool (siControl) were purchased from GE Healthcare Dharmacon, Inc. This SMARTpool siRNA contains four pooled siRNAs, each targeting a separate region of the L1CAM RNA sequence. ON-TARGETplus Non-targeting siRNA (D-001810-10-20) was also used as a non-targeting control. The SMARTpool siRNA and non-targeting siRNA target sequences are as follows: ON-TARGETplus SMARTpool siRNA J-011069-05, L1CAM Target Sequence: CACUACACCUUUAGGGUUA; ON-TARGETplus SMARTpool siRNA J-011069-06, L1CAM Target Sequence: GCAAGAGACAUAUCCACAA; ON-TARGETplus SMARTpool siRNA J-011069-07, L1CAM Target Sequence: GAUACAAUGUGACGUACUG; ON-TARGETplus SMARTpool siRNA J-011069-08, L1CAM Target Sequence: ACACAAUGGUGACCCAAUG; and ON-TARGETplus Non-targeting pool Target Sequences: UGGUUUACAUGUCGACUAA, UGGUUUACAUGUUGUGUGA, UGGUUUACAUGUUUUCUGA and UGGUUUACAUGUUUUCCUA.

Transfection was performed using Lipofectamine 2000^® (Invitrogen; Thermo Fisher Scientific, Inc.). Briefly, anti-L1CAM siRNA and Lipofectamine 2000 were mixed according to the protocol and allowed to form the siRNA-Lipofectamine 2000 lipoplexes at RT for 20 min. HSC-4 or HN22 cells was seeded on 60-mm plates and transfected transiently with 100 nM transfection complex. After 24 h of transfection, subsequent wound healing or Transwell migration/invasion experiment was performed.

A total of 3×10⁵ cells were seeded in six-well plates and incubated overnight. After treatment with 100 ng/ml of rhL1CAM or 100 nM of siL1CAM for 24 h, cells were incubated with 5% CO₂ at 37°C for 24, 48 and 72 h. Viable cells were stained with 0.4% trypan blue solution (Gibco; Thermo Fisher Scientific, Inc.) at RT for 5 min and counted using a hemocytometer at each time-point.

Wound healing assay was carried out as previously described (25). Ca9-22, HSC-2, HSC-4, and HN22 (3×10⁵ cells) were seeded in six-well plates. After being cultured for 24 h, cells were treated with rhL1CAM (100 ng/ml) or siL1CAM (100 nM) using Opti-MEM medium (Gibco; Thermo Fisher Scientific, Inc.). After 24 h, the center of the cell monolayers was scratched vertically with a sterilized 100-µl pipette tip. After rinsing with PBS three times for removing cell debris, the cells were incubated in 5% FBS-containing medium for 12 h to enable wound healing. Images were captured at 0 and 12 h using an inverted microscope (CKX53; Olympus Corporation) and the wound dimensions were measured using ImageJ software (version 1.51 k; National Institutes of Health).

For Transwell migration assay, Ca9-22 (0.8×10⁵), HSC-2 (0.8×10⁵), HSC-4 (1.8×10⁵), and HN22 (1.8×10⁵) cells in serum-free medium were seeded in the upper chambers of 24 well plate with Collagen type I-coated PET membrane of 8.0-µm pore size (BD Biosciences). For invasion assay, culture inserts of 24-well plate were coated with Matrigel (BD Biosciences) in a 37°C incubator for 2 h. Ca9-22 (1.0×10⁵), HSC-2 (1.0×10⁵), HSC-4 (1.2×10⁵), and HN22 (1.2×10⁵) cells in serum-free medium were seeded in the upper chambers. Lower chambers were filled with the media containing 10% FBS as a chemoattractant. After incubation for 24 h, the non-migratory or non-invasive cells in upper chamber were removed with a cotton swab. Cells on the lower surface of the filter were fixed with 100% methanol for 2 min and stained with H&E solution. Images of migratory or invasive cells were captured under an inverted light microscope (CKX53; Olympus Corporation) and the number of migratory or invasive cells were counted in randomly selected areas in three different microscopic fields (magnification, ×100).

Statistical analyses were performed using SPSS software (version 26.0; IBM Corp.). Correlation of L1CAM expression with clinicopathological parameters was assessed with Pearson's Chi-squared test or Fisher's exact test. Survival curves were generated using the Kaplan-Meier method and between-group differences assessed using the log-rank test. For in vitro experiments, the mean ± standard deviation values from three independent experiments are reported. Statistical significance was analyzed using a two-tailed Student's t-test for paired samples. One-way ANOVA followed by Tukey's post hoc test analyses were used for multiple-group comparisons. For all analyses, P<0.05 was considered to indicate a statistically significant difference.

L1CAM expression was not detected in the epithelium of the normal oral mucosa (Fig. 1A), but was strongly expressed in the peripheral nerve tissue, vessels and mast cells (Fig. 1B). In OTSCC tissue samples, tumor cells showed a membranous pattern of expression of L1CAM (Fig. 1C). Of the 80 clinical samples, 26 (32.5%) samples were positive for L1CAM. Tumor cells exhibiting moderate to strong expression were frequently found at the invasive front of tumors or in the less-differentiated and usually spindle-shaped tumor cells (Fig. 1D and E), while negative expression was observed in relatively well-differentiated tumors (Fig. 1F). In the area of PNI, tumor cells expressing L1CAM were found in the perineural tissues or even inside the nerves (Fig. 1G and H).

mRNA and protein levels of L1CAM were examined in six OSCC cell lines (Fig. 1I). HSC-3, HSC-4 and HN-22 presented relatively high levels of L1CAM compared with other cell lines (Ca9.22, HSC-2 and SAS). Thus, Ca9.22 and HSC-2 were used for confirming the effect of L1CAM upregulation induced by treatment with rhL1CAM. By contrast, HSC-4 and HN-22 were used for determining the effect of L1CAM downregulation using siL1CAM.

To assess the correlation of L1CAM expression with clinicopathological parameters, including PNI, PNI was determined more objectively according to the commonly accepted definition by Liebig et al (3): i) Tumor in close proximity to nerve and involving at least 33% of its perimeter or ii) tumor cells within any of the three layers of the nerve sheath. PNI was detected in 40.0% of 80 OTSCC samples. PNI, which is a known adverse prognostic factor in OSCC, revealed a significant correlation with pT (P=0.001), DOI (P<0.001), LN metastasis (P<0.001), advanced clinical stage (P=0.001) and survival status (P=0.002) (Table II). L1CAM expression showed significant correlation with high histologic grade (P=0.013), greater DOI (P=0.009), LN metastasis (P<0.001), presence of PNI (P<0.001), advanced stage (P=0.004), and survival status (P=0.018) (Table III). These findings suggested high clinical relevance of L1CAM in relation to cancer invasion, metastasis, and progression.

Kaplan-Meier survival analysis was performed to evaluate the prognostic significance of LN metastasis, PNI, and L1CAM expression for overall survival (OS) of patients with OSCC. OTSCC patients with PNI or LN metastasis demonstrated significantly worse OS rate (P=0.001 and P=0.007, respectively; Fig. 2A and B). As for a novel marker, L1CAM, patients with positive expression of L1CAM showed significantly lower OS than those with negative expression (P=0.014; Fig. 2C). L1CAM expression was found to have a more significant effect on the decrease of OS rate in patients with T1/2 tumor, N0 tumor, or stage I/II tumor (P=0.012, P=0.001 and P=0.003, respectively; Fig. 2D-F). These findings suggested that expression of L1CAM in early OSCC may be an indicator of poor prognosis.

To evaluate the function of L1CAM in cell proliferation, migration and invasion in OSCC, Ca9-22 and HSC-2 cell lines were treated with rhL1CAM. The viability of Ca9-22 cells increased in a concentration-dependent manner up to 400 ng/ml of concentration of rhL1CAM, while viability of HSC-2 cells increased gradually up to 100 ng/ml of concentration and there was no change >100 ng/ml (**P<0.01 and ***P<0.001; Fig. 3A). Overexpression of L1CAM significantly increased proliferation capacity of both cell lines at each time-point (***P<0.001; Fig. 3B). Next, wound healing assay revealed that upregulation of L1CAM significantly increased the migration of both cell lines (***P<0.001; Fig. 3C). Wound closure in the rhL1CAM-treated groups increased more than two-fold compared with that of control groups in both cell lines. Lastly, Transwell migration and Matrigel invasion assays also revealed significantly increased migration and invasion ability in rhL1CAM-treated groups of both cell lines compared with the corresponding control groups (***P<0.001; Fig. 3D). Collectively, these findings indicated that overexpression of L1CAM can promote cell proliferation, migration, and invasion, suggesting its important role in the progression of OSCC.

To assess the effect of L1CAM knockdown, both HSC-4 and HN22 cells were treated with 100 nM siL1CAM for 24 h. HSC-4 and HN22 cell lines revealed decreased cell viability in a concentration-dependent manner (*P<0.05, **P<0.01 and ***P<0.001; Fig. 4A). Suppression of L1CAM expression significantly decreased the proliferation capacity of both cell lines at each time-point (***P<0.001; Fig. 4B). Next, in wound healing assay, L1CAM knockdown significantly affected the migration ability of both cell lines. The ability of wound closure in siL1CAM-treated groups of both cell lines decreased more than three-fold compared with that in the corresponding control groups (***P<0.001; Fig. 4C). Lastly, Transwell migration and Matrigel invasion assays demonstrated significantly decreased cell migration and invasion ability in siL1CAM-treated groups of both cell lines compared with the control groups (***P<0.001; Fig. 4D). These results indicated that knockdown of L1CAM can inhibit proliferation, migration and invasion, suggesting its possibility to serve as a therapeutic target.

Based on the effect of L1CAM in enhancing the migration and invasion of OSCC cells, the association of L1CAM with EMT was assessed. After L1CAM knockdown, changes in the expression of EMT-associated markers were assessed by western blotting. As revealed in Fig. 5A, expression of epithelial cell marker E-cadherin was increased, whereas mesenchymal markers including N-cadherin, vimentin, snail, slug and twist were significantly decreased after L1CAM knockdown. In other words, suppression of L1CAM was found to reverse EMT. These results suggested the involvement of L1CAM in the EMT process of OSCC cells.

The PI3K/AKT and ERK pathways are the main signaling pathways regulating cell proliferation, differentiation, survival, motility and apoptosis. Dysregulation of these pathways is known to contribute to carcinogenesis and cancer progression (18). In order to determine whether the functions of L1CAM in OSCC are mediated through the PI3K/AKT and ERK signaling pathways, the effect of L1CAM knockdown on the activation of these pathways was investigated. As revealed in Fig. 5B, no change was observed in the expression of PI3K/AKT and ERK1/2 in the HSC-4 and HN22 cell lines; however, there was significant decrease of p-PI3K, p-AKT and p-ERK1/2 levels. According to the densitometric analysis, relative protein expression levels and the ratio of p-/total protein were evaluated with bar-chart graphs (*P<0.05, **P<0.01 and ***P<0.001; Fig. 5C). These data indicated that L1CAM likely plays a role in the progression of OSCC via both the PI3K/AKT and ERK pathways.