Introduction

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer in the world, and 650 000 new cases are diagnosed every year.1 Despite considerable advances in multimodality therapy, including surgery, radiotherapy and chemotherapy, the overall survival rate for patients with HNSCC is only 15–45%.2, 3 Patients with HNSCC are usually diagnosed at a late stage, and local tumor recurrence and distant metastasis often occur after conventional therapies.2, 3 Metastatic disease is responsible for about 90% of deaths within 12 months of diagnosis in patients with HNSCC.1 Therefore, understanding the molecular mechanisms of the metastatic pathways underlying HNSCC using currently available genomic approaches may improve therapies for and prevention of the disease.

MicroRNAs (miRNAs) are endogenous small noncoding RNA molecules (19–22 bases in length) that function to regulate protein-coding gene expression by repressing translation or cleaving RNA transcripts in a sequence-specific manner.4 To date, a substantial amount of evidence has suggested that miRNAs are aberrantly expressed in many human cancers and have significant roles in human oncogenesis and metastasis.5, 6, 7 We sequentially identified tumor-suppressive miRNAs that contribute to cancer cell migration and invasion based on the miRNA expression signatures of HNSCC.8, 9, 10, 11 Among these tumor-suppressive miRNAs, microRNA (miR)-26a/b, miR-29a/b/c and miR-218 significantly inhibit cancer cell migration and invasion through targeting of genes encoding extracellular matrix (ECM) components and integrins.11, 12, 13

The ECM is composed of secreted molecules that constitute the cell microenvironment, including a dynamic and complex array of glycoproteins, collagens, glycosaminoglycans and proteoglycans.14 The ECM imparts spatial context for signaling events through various cell surface growth factor receptors and adhesion molecules, such as integrins.14 Recent studies have shown that ECM components and ECM-related receptor proteins are aberrantly expressed in cancer tissues and that such dysregulation of the composition and organization of the ECM can induce activation of specific signaling pathways.15 Integrins are a large family of cell surface receptors composed of two subunits (α and β), which bind to ECM components. Most types of cells require integrin-mediated signal pathways for proliferation, migration, invasion and survival.16 Several studies have shown that overexpression or activation of integrin-mediated cancer signals promotes cancer cell progression and metastasis.17, 18, 19 For example, the interaction between laminin-332 and α6β4 integrin triggers a number of signaling cascades in cancer cells, promoting both cell migration and cancer cell survival in HNSCC.12, 13 We also showed that tumor-suppressive function of miR-218 and miR-29a/b/c in prostate cancer and renal cell carcinoma through targeting ECM component genes.20, 21

In cancer cells, aberrant expression of miRNAs disrupts tightly controlled miRNA/protein-coding RNA networks. Therefore, identification of tumor-suppressive miRNAs that act together to regulate oncogenic genes is the first step in elucidating the mechanisms of human oncogenesis and metastasis. The aim of the current study was to identify miRNAs (miR-26a/b, miR-29a/b/c and miR-218) that function in concert to downregulate metastasis-promoting genes in HNSCC. Genome-wide gene expression data and in silico analysis revealed that several putative candidate genes were regulated by these tumor-suppressive miRNAs. Here we focused on the lysyl oxidase-like 2 (LOXL2) gene. Our present data showed that LOXL2 was overexpressed in HNSCC clinical specimens and that silencing of the LOXL2 gene significantly inhibited the migration and invasion of cancer cells. Moreover, LOXL2 was a direct target of miR-26a/b, miR-29a/b/c and miR-218 regulation, as shown using luciferase reporter assays. These data suggested that the pathway involving regulation of LOXL2 by tumor-suppressive miRNAs may be a potential target in the development of novel therapies to treat HNSCC.

Materials and methods

Clinical HNSCC specimens

A total of 19 pairs of primary HNSCC and corresponding normal epithelial tissue were obtained from patients with HNSCC at Chiba University Hospital (Chiba, Japan) from 2008 to 2015. The patients’ backgrounds and clinicopathological characteristics are shown in Table 1. The patients were classified according to the 2002 Union for International Cancer Control Tumor-Node-Metastasis (TNM) staging criteria before treatment. Written consent for tissue donation for research purposes was obtained from each patient before tissue collection. The protocol was approved by the Institutional Review Board of Chiba University. The specimens were immersed in RNAlater (Qiagen, Valencia, CA, USA) and stored at −20 °C until RNA extraction.

Table 1 Clinical features of 19 patients with HNSCC

Cell lines

The following human HNSCC cell lines were used: FaDu (derived from the primary tumor in a patient with hypopharyngeal SCC) and SAS (derived from the primary tumor of a patient with tongue SCC). Both cell lines were grown in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum in a humidified atmosphere containing 5% CO2 at 37 °C.

RNA isolation

Total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol. The RNA concentrations were determined spectrophotometrically, and molecular integrity was checked by gel electrophoresis. The quality of RNA was confirmed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA).

Quantitative real-time reverse transcription polymerase chain reaction (qRT–PCR)

The expression levels of miR-26a (Assay ID: 000405), miR-26b (Assay ID: 000407), miR-29a (Assay ID: 002112), miR-29b (Assay ID: 000413), miR-29c (Assay ID: 000413) and miR-218 (Assay ID: 000587) were analyzed by TaqMan qRT–PCR and normalized to RNU48 (Assay ID: 001006). TaqMan probes and primers for LOXL2 (P/N: Hs00158757_m1; Applied Biosystems, Foster City, CA, USA), LOX (P/N: Hs00942480_m1) and GUSB (P/N: Hs99999908_ml) as an internal control were obtained from Applied Biosystems (Assay-On-Demand Gene Expression Products).

Immunohistochemistry

Three hypopharyngeal SCC tissue specimens were immunostained following the manufacture’s protocol with the Ultra-Vision Detection System (Thermo Scientific, Fremont, CA, USA). Primary rabbit polyclonal antibodies against LOXL2 were diluted 1:1000. The slides were treated with biotinylated goat antibodies.

Transfection with mature miRNAs and small-interfering RNA

The following mature miRNAs were used in this study: mirVana miRNA mimics for hsa-miR-26a-5p (product ID: PM10249), hsa-miR-26b-5p (product ID: PM12899), hsa-miR-29a-3p (product ID: MC12499), hsa-miR-29b-3p (product ID: MC10103), hsa-miR-29c-3p (product ID: MC10518) and hsa-miR-218 (product ID: AM17100; Applied Biosystems). The following si-RNAs were used in this study: stealth select RNAi si-RNA, si-LOXL2 (P/N: HSS180848; Invitrogen), and negative control miRNA/small-interfering RNA (P/N: AM17111, Applied Biosystems). Transfection methods were described previously.8, 9, 10, 11, 12, 13, 20, 21

Cell proliferation, migration and invasion assays

To investigate the functional roles of LOXL2, we performed cell proliferation, migration and invasion assays using si-LOXL2-transfected FaDu and SAS cells. The experimental procedures were performed as described in our previous studies.9, 10, 11, 12, 13, 20, 21

Identification of putative target genes regulated by miR-26a/b, miR-29a/b/c and miR-218

To investigate putative target genes regulated by these miRNAs, we use in silico analysis. First, we screened genes using TargetScan Release 6.2 (http://www.targetscan.org/). To identify upregulated genes in HNSCC, we analyzed a publicly available gene expression data set in GEO (accession number 9638).

Western blotting

Cells were collected 72 h after transfection and lysates were prepared. Next, 20 μg of protein lysates were separated on Mini-PROTEAN TGX gels (Bio-Rad, Hercules, CA, USA) and transfered to PVDF membranes. Immunoblotting was performed with rabbit anti-LOXL2 antibodies (1:1000; ab96233, Abcam, Cambridge, UK); anti-GAPDH antibodies (1:4000; ab8245, Abcam) were used to detect GAPDH as an internal loading control. Membranes were washed and incubated with anti-rabbit IgG horseradish peroxidase-linked antibodies (7074; Cell Signaling Technology, Danvers, MA, USA). Complexes were visualized with Clarity Western ECL Substrate (Bio-Rad). The experimental procedures were described previously.8, 9, 10, 11, 12, 13, 20, 21

Plasmid construction and dual-luciferase assays

Partial wild-type sequences of the LOXL2 3′-untranslated region (UTR) or those with deleted miR-26a/b, miR-29a/b/c and miR-218 binding sites were inserted between the XhoI-PmeI restriction sites in the 3′-UTR of the hRuc gene in the psiCHECK-2 vector (C8021; Promega, Madison, WI, USA). The procedure for the dual-luciferase reporter assay was described previously.8, 9, 10, 11, 12, 13, 20, 21

Identification of genes regulated by LOXL2 in HNSCC

To identify the genes regulated by LOXL2, we performed gene expression analysis using si-LOXL2-transfected FaDu cells. An oligo-microarray (human 60k v; Agilent Technologies) was used for gene expression studies.

Statistical analysis

The relationships between two groups and the numerical values obtained by real-time RT–PCR were analyzed using paired t-tests. Spearman’s rank test was used to evaluate the correlation between the expression of miR-26/b, miR-29a/b/c and miR-218 and the expression of LOXL2. The relationships among more than three variables and numerical values were analyzed using Bonferroni-adjusted Mann–Whitney U-tests. All analyses were performed using Expert Stat View (version 4, SAS Institute Inc., Cary, NC, USA).

Results

Selection of candidate genes regulated by tumor-suppressive miRNAs (miR-26a/b, miR-29a/b/c and miR-218) in HNSCC

To identify putative target genes regulated by tumor-suppressive miRNAs, particularly those inhibiting cancer cell migration and invasion in HNSCC cells (that is, miR-26a/b, miR-29a/b/c and miR-218), as described our previous studies,11, 12, 13 we used in silico analysis and genome-wide gene expression analysis. Our strategy for selection of target genes is shown in Figure 1. We screened target genes using the TargetScan database and identified 334 genes. Next, these genes were then analyzed with a publicly available gene expression data set in GEO (accession number: GSE9638) and upregulated genes (log2 ratio >1.5) were chosen. As a result, 26 genes were identified as putative target genes regulated by miR-26a/b, miR-29a/b/c and miR-218 (Table 2). To show the effectiveness of Table 2, we checked the public database. Previous studies showed that HAPLN1 (hyaluronan an proteoglycan link protein 1) and POFUT1 (protein O-fucosyltransferase 1) were upregulated in cancer tissues and act as oncogenes in oral squamous cell carcinoma.22, 23 The new knowledge of HNSCC might be offered by analyzing the gene included in this list. Among these candidate genes, we focused on LOX and LOXL2 because these genes have been reported to promote metastasis in several types of cancers. Furthermore, our recent study of renal cell carcinoma showed that miR-29a/b/c significantly inhibited cancer cell migration and invasion directly targeting LOXL2.21 These data suggest that LOXL2 member deeply contribute to cancer cell metastasis.

Figure 1
figure 1

Selection for putative target genes regulated by tumor-suppressive miRNAs (miR-26a/b, miR-29a/b/c and miR-218) in HNSCC. A total of 334 genes were identified as target genes containing binding sites for miR-26a/b, miR-29a/b/c and miR-218. Among these, 26 genes were upregulated in HNSCC clinical specimens (accession number: GSE9638). HNSCC, head and neck squamous cell carcinoma.

Table 2 Putative target genes regulated by miR-26a/b, miR-29a/b/c and miR-218

Expression levels of LOXL2 in HNSCC clinical specimens and cell lines

Next, we evaluated the expression levels of LOXL2 mRNA in HNSCC clinical specimens. The expression of LOXL2 was significantly upregulated in cancer tissues compared with that in normal tissues (P=0.0057), as demonstrated by qRT–PCR (Figure 2a). We also determined the expression levels of LOXL2 protein in hypopharyngeal SCC specimens by immunohistochemical staining. LOXL2 was strongly expressed in several cancerous lesions, whereas no or low expression was observed in normal tissues (Figure 2b–d). The patients’ backgrounds and clinicopathological characteristics are summarized in Table 1.

Figure 2
figure 2

Expression levels of LOXL2 in HNSCC clinical specimens and cell lines. (a) Expression levels of LOXL2 mRNA in HNSCC clinical specimens and cell lines (FaDu and SAS). GUSB was used as an internal control. (bd) Immunohistochemical staining for detection of LOXL2 and HE staining in three patients with hypopharyngeal squamous cell carcinoma. HNSCC, head and neck squamous cell carcinoma.

Finally, the expression levels of LOX in HNSCC clinical specimens were evaluated by qRT–PCR (Supplementary Figure 1). Overexpression of LOX was observed in clinical specimens. The mRNA expression of LOX was significantly repressed in miR-29a/b/c or miR-218 transfectant compared with mock- or miR-control-transfectant cells. However, no suppressive effect was observed in miR-26a/b transfectant cells (Supplementary Figure 1). In addition, high expression of LOX was not found in cancer cell lines (FaDu and SAS). Therefore, we omitted this gene as a target of further analysis in the study.

Expression levels of tumor-suppressive miRNAs in HNSCC clinical specimens and cell lines

The expression levels of miR-26a, miR-26b, miR-29a, miR-29c and miR-218 were significantly lower in tumor tissues and cell lines (FaDu and SAS) than in corresponding normal epithelial tissues, as shown by qRT–PCR (Figures 3a, 4a and 5a). On the other hand, miR-29b expression was not significantly downregulated in clinical cancer tissues (Figure 4a).

Figure 3
figure 3

Expression levels of miR-26a and miR-26b in HNSCC clinical specimens and cell lines. (a) Expression levels of miR-26a and miR-26b in clinical specimens and cell lines (FaDu and SAS). RNU48 was used as an internal control. (b) Correlations between the expression levels of LOXL2 and miR-26a or miR-26b were determined in HNSCC clinical specimens. Spearman’s rank test was used to evaluate the correlations. HNSCC, head and neck squamous cell carcinoma.

Figure 4
figure 4

Expression levels of miR-29a, miR-29b and miR-29c in HNSCC clinical specimens and cell lines. (a) Expression levels of miR-29a, miR-29b and miR-29c in clinical specimens and cell lines (FaDu and SAS). RNU48 was used as an internal control. (b) Correlations between the expression levels of LOXL2 and miR-29a, miR-29b or miR-29c were determined in HNSCC clinical specimens. Spearman’s rank test was used to evaluate the correlations. HNSCC, head and neck squamous cell carcinoma.

Figure 5
figure 5

Expression levels of miR-218 in HNSCC clinical specimens and cell lines. (a) Expression levels of miR-218 in clinical specimens and cell lines (FaDu and SAS). RNU48 was used as an internal control. (b) The correlation between LOXL2 and miR-218 expression was determined in HNSCC clinical specimens. Spearman’s rank test was used to evaluate the correlation. HNSCC, head and neck squamous cell carcinoma.

Spearman’s rank test showed negative correlations between the expression of miR-26a, miR-26b, miR-29a, miR-29c and miR-218 and the expression of LOXL2 in HNSCC clinical specimens (Figures 3b, 4b and 5b). However, there was no negative correlation between the expression levels of miR-29b and LOXL2 (Figure 4b).

LOXL2 was directly regulated by tumor-suppressive miRNAs (miR-26a/b, miR-29a/b/c and miR-218) in HNSCC cells

Next, we performed qRT–PCR and western blotting in FaDu and SAS cells to investigate whether LOXL2 expression was downregulated by restoration of miR-26a/b, miR-29a/b/c or miR-218. The expression levels of LOXL2 mRNA and protein were significantly repressed in cells transfected with miR-26a/b, miR-29a/b/c or miR-218 compared with that in mock- or miR-control-transfected cells (Figures 6a, 7a and 8a).

Figure 6
figure 6

Direct regulation of LOXL2 by miR-26a and miR-26b. (a) LOXL2 mRNA expression 72 h after transfection with miR-26a and miR-26b. GUSB was used as an internal control. (b) LOXL2 protein expression 72 h after transfection with miR-26a and miR-26b. GAPDH was used as a loading control. (c) miR-26a and miR-26b binding sites in the 3′-UTR of LOXL2 mRNA. Luciferase reporter assays used vectors that included the wild-type sequence of putative miR-26a and miR-26b binding sites or mutants in which the binding sites were deleted (Deletion). Renilla luciferase signals were normalized to firefly luciferase signals. UTR, untranslated region.

Figure 7
figure 7

Direct regulation of LOXL2 by miR-29a, miR-29b and miR-29c. (a) LOXL2 mRNA expression 72 h after transfection with miR-29a, miR-29b and miR-29c. GUSB was used as an internal control. (b) LOXL2 protein expression 72 h after transfection with miR-29a, miR-29b and miR-29c. GAPDH was used as a loading control. (c) miR-29a, miR-29b and miR-29c binding sites in the 3′-UTR of LOXL2 mRNA. Luciferase reporter assays used vectors that included wild-type sequences of putative miR-29a, miR-29b and miR-29c binding sites or mutants in which the binding sites were deleted. Renilla luciferase signals were normalized to firefly luciferase signals. UTR, untranslated region.

Figure 8
figure 8

Direct regulation of LOXL2 by miR-218. (a) LOXL2 mRNA expression 72 h after transfection with miR-218. GUSB was used as an internal control. (b) LOXL2 protein expression 72 h after transfection with miR-218. GAPDH was used as a loading control. (c) miR-218 binding sites in the 3′-UTR of LOXL2 mRNA. Luciferase reporter assays used vectors that included the wild-type sequence of the putative miR-218 binding site or a mutant in which the binding sites were deleted. Renilla luciferase signals were normalized to firefly luciferase signals. UTR, untranslated region.

Furthermore, we performed luciferase reporter assay in SAS cells to determine whether LOXL2 mRNA contained target sites for miR-26a/b, miR-29a/b/c and miR-218. We used vectors encoding either a partial wild-type sequence or a sequence in which the miRNA binding site was deleted from the 3′-UTR of LOXL2 mRNA. We found that the luminescence intensity was significantly reduced by co-transfection with miR-26a/b, miR-29a/b/c, miR-218 and the vector carrying the wild-type 3′-UTR of LOXL2 mRNA (Figures 6c, 7c and 8c).

Effects of silencing LOXL2 on cell proliferation, migration and invasion in HNSCC cell lines

To investigate the functional role of LOXL2, we performed loss-of-function studies using si-LOXL2 transfectants. First, we evaluated the knockdown efficiency of si-LOXL2 transfection in FaDu and SAS cells. Western blotting and qRT–PCR indicated that the si-RNAs effectively downregulated LOXL2 expression in FaDu and SAS cells (Figure 9a).

Figure 9
figure 9

Effects of si-LOXL2 transfection on HNSCC cell lines. (a) Expression of LOXL2 mRNA 72 h after transfection with si-LOXL2 in FaDu and SAS cells. GUSB was used as an internal control. (b) Protein expression of LOXL2 72 h after transfection with si-LOXL2. GAPDH was used as a loading control. (c) Cell proliferation 72 h after transfection with 10 nM si-LOXL2. (d) Cell migration 72 h after transfection with 10 nM si-LOXL2. (e) Cell invasion 72 h after transfection with 10 nM si-LOXL2. HNSCC, head and neck squamous cell carcinoma.

XTT assays demonstrated that cell proliferation was significantly inhibited in si-LOXL2-1 transfectants compared with that in mock- or si-control-transfected SAS cells. On the other hand, proliferation was not inhibited in FaDu cells (Figure 9c). Migration and invasion assays demonstrated that cell migration and invasion activities were significantly inhibited in si-LOXL2 transfectants compared with that in mock or si-control transfectants for both cell lines (Figure 9d).

Identification of downstream pathways regulated by LOXL2

Few reports have described the functional significance of LOXL2 in HNSCC. Therefore, we investigated the molecular pathways regulated by LOXL2 in FaDu cells using genome-wide gene expression analysis in si-LOXL2 transfectants. Twenty-five genes were downregulated (log2 ratio <−1) in response to si-LOXL2 transfection and upregulated (log2 ratio >1) in HNSCC clinical specimens (accession number: GSE9638; Table 3).

Table 3 Downregulated genes after transfection with si-LOXL2 in FaDu cells

Discussion

The highly invasive properties of cancer cells frequently cause locoregional recurrence and distant metastasis in patients with HNSCC, leading to an expected 5-year survival rate of ~50% for patients with advanced disease.24, 25 Therefore, understanding the molecular mechanisms of metastatic pathways underlying HNSCC using current genomic approaches, including noncoding RNA networks, would facilitate the development of novel therapies for and prevention of the disease. For elucidation of novel metastatic RNA networks in HNSCC, we sequentially identified tumor-suppressive miRNAs that regulated oncogenic genes and HNSCC-related pathways.11, 12, 13 Our previous studies of the functional significance of downregulated miRNAs in HNSCC expression signatures showed that six miRNAs (that is, miR-26a/b, miR-29a/b/c and miR-218) significantly inhibit cancer cell migration and invasion,11, 12, 13 suggesting that these miRNAs act to suppress metastasis in human cancers. Therefore, we hypothesized that these tumor-suppressive miRNAs (that is, miR-26a/b, miR-29a/b/c and miR-218) would regulate genes with key functions in HNSCC metastasis.

Previous studies have shown that miR-26a and miR-26b act as tumor suppressors by targeting oncogenic genes in several types of cancers, such as breast cancer, nasopharyngeal carcinoma and hepatocellular carcinoma. More recently, the overexpression of miR-26a has been shown to inhibit tongue SCC cell proliferation and promote cell apoptosis.11, 26 Decreased expression of miR-29a/b/c has been observed in cholangiocarcinoma, nasopharyngeal cancer, non-small cell lung cancer, hepatocellular carcinoma, malignant peripheral nerve sheath tumors and mantle cell lymphoma.27 Our recent studies of miRNA expression signatures in urothelial cancers and prostate cancer showed that miR-218 is frequently downregulated in cancer tissues and functions as a tumor suppressor.20, 28 The tumor-suppressive functions of miR-218 in several types of cancers have been described by other research groups.29

To better understand cancer cell metastasis, we sequentially identified tumor-suppressive miRNAs mediating novel metastatic pathways using genome-wide gene expression analysis and in silico analysis.11, 12, 13 Among the putative targets of miR-26a/b, miR-29a/b/c and miR-218 regulation in HNSCC, we focused on the LOX and LOXL2 genes as metastatic regulators in HNSCC cells. Importantly, our previous data in renal cell carcinoma showed that the LOXL2 gene is a direct target of tumor-suppressive miR-29a/b/c regulation.21 Past studies have shown that LOX and LOXL2 are overexpressed in several types of cancers.30, 31, 32, 33 In the present study, we confirmed the overexpression of LOX and LOXL2 in HNSCC clinical specimens. LOXL2 was directly regulated by miR-26a/b, miR-29a/b/c and miR-218. LOXL2 was directly regulated by miR-26a/b, miR-29a/b/c and miR-218. However, LOX was regulated by miR-29a/b/c and miR-218, not by miR-26a/b in HNSCC cells (Supplementary Figure 1). Thus, we focused on LOXL2 and investigated the functional significance of LOXL2 regulation in HNSCC. Our data showed that silencing of LOXL2 significantly inhibited cancer cell migration and invasion, consistent with the function of LOX-family proteins in covalent crosslinking of collagen and/or elastin in the ECM.34

Collagens are the most abundant ECM component, and excessive ECM deposition has been observed in many cancers.35, 36, 37 In cancer cells, collagens regulate the physical and biochemical properties of the cancer cell microenvironment, which modulates cancer cell polarity, migration and signaling.38 Collagen crosslinking is extracellularly initiated by the LOX family of secreted enzymes, which are regulated by hypoxia-inducible factor and are induced under hypoxic conditions.39 Hypoxic conditions also induce the expression of collagen-degrading proteins and matrix metalloproteinases (MMPs), thereby contributing to ECM remodeling; these mechanisms are mediated by several families of proteinases that have been suggested to promote cancer cell invasion.40 In this study, we investigated the pathways downstream of LOXL2 in HNSCC cells following knockdown of LOXL2 using si-LOXL2. Among the putative LOXL2 downstream genes, PLOD3 encodes enzymes that mediate collagen lysine hydroxylation,41 and SERPINH1 encodes a member of the serpin superfamily of serine proteinase inhibitors, which has a role in collagen biosynthesis as a collagen-specific molecular chaperone.42 Moreover, MMP1, which is also downstream of LOXL2, encodes a secreted enzyme that breaks down the interstitial collagens, that is, types I, II and III.43 Thus, these proteins increase the crosslinking of collagens and other ECM components, subsequently promoting matrix stiffness. Moreover, LOXL2-mediated ECM deposition seems to function in synergy with MMP activity, which may lead to remodeling of the ECM in such a way as to increase metastasis.40

The intracellular function of LOXL2 is still not clear. Several transcription factors are upregulated in metastatic cells that are undergoing the epithelial-mesenchymal transition, including Snail, TWIST, ZEB and others.44 Transforming growth factor-β has a critical role in activating Snail, which in turn downregulates E-cadherin; this process promotes the epithelial-mesenchymal transition.45 Interestingly, past studies have shown that nuclear LOXL2 interacts with the transcription factor Snail1, represses E-cadherin and induces the epithelial-mesenchymal transition.46 These findings indicated that LOXL2 deeply contribute to activation of metastatic pathways in HNSCC. Recent study showed that LOXL2 binds NOTCH1 promoter region and repressed NOTCH1 transcription.47 Several studies indicated that NOTCH mutation and NOTCH-mediated signal pathways can have either an oncogenic or a tumor-suppressive effects.48 A further study is necessary about LOXL2-NOTCH signaling pathways and HNSCC oncogenesis.

In conclusion, downregulation of miR-26a/b, miR-29a/b/c and miR-218 was frequently observed in HNSCC clinical specimens. These miRNAs functioned as tumor suppressors, inhibiting cancer cell migration and invasion. Our present study showed that LOXL2 was a direct target of all of these tumor-suppressive miRNAs in HNSCC cells. Moreover, LOXL2 was upregulated in HNSCC clinical specimens and contributed to cancer cell migration and invasion, indicating that LOXL2 promoted metastasis. The identification of novel metastatic pathways and targets regulated by the tumor-suppressive miRNA-LOXL2 axis may lead to a better understanding of HNSCC and the development of new therapeutic strategies to treat this disease.