RNA‐binding protein‐regulated fibronectin is essential for EGFR‐activated metastasis of head and neck squamous cell carcinoma

There is a higher expression level of epidermal growth factor receptor (EGFR) in up to 90% of advanced head and neck squamous cell carcinoma (HNSCC) tissue than in normal surrounding tissues. However, the role of RNA‐binding proteins (RBPs) in EGFR‐associated metastasis of HNSCC remains unclear. In this study, we reveal that RBPs, specifically nucleolin (NCL) and heterogeneous nuclear ribonucleoprotein A2/B1 (hnRNPA2B1), correlated with the mesenchymal phenotype of HNSCC. The depletion of RBPs significantly attenuated EGF‐induced HNSCC metastasis. Intriguingly, the EGF‐induced EMT markers, such as fibronectin, were regulated by RBPs through the ERK and NF‐κB pathway, followed by the enhancement of mRNA stability of fibronectin through the 5′ untranslated region (5′‐UTR) of the gene. The upregulation of fibronectin triggered the integrin signaling activation to enhance tumor cells’ attachment to endothelial cells and increase endothelial permeability. In addition, the concurrence of EGFR and RBPs or EGFR and fibronectin was associated with overall survival and disease‐free survival of HNSCC. The in vivo study showed that depletion of NCL, hnRNPA2B1, and fibronectin significantly inhibited EGF‐promoted extravasation of tumor cells into lung tissues. The depletion of fibronectin or treatment with integrin inhibitors dramatically attenuated EGF‐induced HNSCC metastatic nodules in the lung. Our data suggest that the RBPs/fibronectin axis is essential for EGF‐induced tumor‐endothelial cell interactions to enhance HNSCC cell metastasis.


| INTRODUCTION
Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide and can be classified as human papillomavirus (HPV)-negative and HPV-positive HNSCC. 1 HPV-positive HNSCC can be prevented by vaccination and generally has a more favorable survival than HPV-negative HNSCC. 2 Treatment of HNSCCs, specifically those derived from the oral cavity and larynx, are now referred to as HPV-negative HNSCC, requires multimodality approaches, including surgical resection, followed by adjuvant radiation, chemotherapy, targeted therapy, and immune checkpoint inhibition. 3However, despite aggressive treatment, the 5-year survival rate remains poor (approximately 50%). 4The mutation or loss of tumor suppressors such as p53 and frequent focal amplification of genes, including epidermal growth factor receptor (EGFR), human epidermal growth factor receptor 2 (HER2), and fibroblast growth factor receptor 1 (FGFR1), are required for a complete transformation to invasive and progressive HNSCC. 5,6or example, EGFR is overexpressed in 80%-90% of HNSCC tumors and is associated with poor overall and progressionfree survival. 7Although elucidation of the molecular genetic landscape of HNSCC has revealed new opportunities for therapeutic intervention, the use of molecular targeting of EGFR with monoclonal antibodies such as cetuximab has been reduced in the treatment of HNSCC upon activation of other receptor tyrosine kinases, including HER2 and c-Met. 8,9In addition, the use of afatinib, an irreversible inhibitor of EGFR, HER2, and HER4 kinases in treating metastatic or recurrent HNSCC patients, exhibited comparable outcomes to cetuximab. 10These results suggest that improving HNSCC therapeutic efficacy may rely on combining EGFRtargeting agents, the inhibition of receptor tyrosine kinase, and their downstream signaling molecules.
Metastatic cancer spreads from its site of origin to another region and causes 90% of deaths in cancer patients. 11ore than half of HNSCC patients develop metastasis or recurrence and have a lower survival rate of less than 30%. 12,13ultiple pathways contribute to HNSCC tumor cell metastasis.5][16] Intriguingly, sustained fibronectin expression in the stromal compartment of human HNSCC strongly correlates with tumor metastasis and decreased survival of patients. 17Fibronectin is also significantly upregulated in HNSCC compared with para-carcinoma tissues in clinical samples. 18In addition, overexpression of α5β1 integrin is observed in aggressive HNSCC, and its binding with fibronectin mediates RAP1promoted migration of HNSCC cells. 19,20Our previous studies also showed that the upregulation of fibronectin activated the Rac1/cdc42 signaling pathway and was associated with PDK1-promoted lactate production and HNSCC cell metastasis. 21Therefore, fibronectin might be a critical prognostic biomarker for HNSCC patients.Several pathways have been identified to be involved in the transcriptional activation of fibronectin, such as the binding of the p21-activated kinase 1/NF-κB complex with the fibronectin promoter and the activation of MAPK in pancreatic cancer and hepatocellular carcinoma, respectively. 22,23Moreover, EGF significantly induces the expression of fibronectin in triple-negative breast cancer cells. 24The depletion of PDK1 or cyclooxygenase-2 also inhibited EGF-induced fibronectin expression and HNSCC cell metastasis. 21,25However, the molecular mechanism involved in regulating fibronectin expression in EGFR-activated HNSCC cells remains incompletely explored.Meanwhile, our current understanding of the molecular factors in regulating metastatic dissemination of HNSCC is still very limited.
RNA-binding proteins (RBPs) regulate gene expression through post-transcriptional regulation, such as by modulating microRNA (miRNA) processing, stability, and RNA translation.Dysregulation of RBPs affects the expression levels of target RNAs related to cancer proliferation, apoptosis, and metastasis. 26Understanding the molecular functions of RBPs can lead to improving therapeutic strategies.8][29][30][31] However, the roles of RBPs in the diverse steps of tumor metastasis remain incompletely understood.In addition, despite several small cells into lung tissues.The depletion of fibronectin or treatment with integrin inhibitors dramatically attenuated EGF-induced HNSCC metastatic nodules in the lung.Our data suggest that the RBPs/fibronectin axis is essential for EGF-induced tumor-endothelial cell interactions to enhance HNSCC cell metastasis.

K E Y W O R D S
EGF, fibronectin, metastasis, RNA-binding protein molecules that have been discovered to inhibit the function of RBPs involved in oral cancer progression, RBPs are generally considered difficult to target. 32Therefore, the alternative strategy to interfere with the function of an RBP is to modulate the interaction with its target RNA.
In this study, we reveal for the first time that RBPs, including nucleolin and hnRNPA2B1, are essential for fibronectin induction in EGF-triggered metastasis of HNSCC cells.We also demonstrate that the RBPs-mediated expression of fibronectin is essential for EGF-promoted HNSCC cell and endothelial cell interactions, increased endothelial permeability, and enhanced tumor cell metastasis.These findings suggest that fibronectin and RBPs such as nucleolin and hnRNPA2B1 may be potentially used as prognostic markers for HNSCC.Moreover, blocking RBPs target genes such as fibronectin-activated integrin signaling provides a therapeutic approach for treating EGF-associated HNSCC metastasis.

| Cell culture and reagents
The FaDu head and neck cancer cell line was purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA).The human microvascular endothelial cell line (HMEC-1) was kindly provided by Dr. Trai-Ming Yeh (Department of Medical Laboratory Science and Biotechnology, Medical College, National Cheng Kung University).Dr. Kwang-Yu Chang kindly provided the head and neck cancer cell lines TU183, UM-SCC1, SAS, and HONE-1 (National Health Research Institutes, NHRI, Taiwan). 33Mouse embryo fibroblast cell line NIH/3T3 and human monocytic leukemia cell line THP-1 were kindly provided by Dr. Ju-Ming Wang (Department of Biotechnology and Bioindustry Sciences, College of Bioscience and Biotechnology, National Cheng Kung University).Cell lines were maintained at 37°C in 10-cm plastic dishes containing 8 -mL medium.Each medium was supplemented with 10% fetal bovine serum (Hyclone), 100 μg/mL Streptomycin (Sigma-Aldrich), and 100 units/mL Penicillin (Sigma-Aldrich).FaDu cells were maintained in an MEM culture medium (Gibco).HMEC-1 cells were maintained in an MCDB 131 culture medium (Gibco).NIH/3T3, UM-SCC1, SAS, and TU183 cells were kept in a DMEM-high glucose culture medium (Gibco).THP-1 and HONE-1 cells were maintained in RPMI 1640 culture medium (Gibco).

| Reverse transcription quantitative polymerase chain reaction (RT-qPCR)
The total RNA of cells was extracted from cells by using REzol C&T (PROtech).A reverse transcriptase reaction was performed using a high-capacity cDNA transcription kit (Thermo) according to the manufacturer's instructions.Following cDNA synthesis, gene-specific primers were designed using NCBI primer design software.Reverse transcription quantitative PCR was performed in triplicate using PowerUp™ SYBR™ Green Master Mix (Thermo) and the StepOne Plus RT-qPCR System (Applied Biosystems).Each well contained the following reaction mixture: 5 μL of 2 × PowerUp™ SYBR™ Green Master Mix, 4 μL of RNase-free water containing 30 ng cDNA, 100 nM sense, and antisense primers.Universal cycling conditions were used.Relative gene expression was calculated using the comparative Ct method.All values were normalized to the housekeeping gene GAPDH.Primer sets for RT-qPCR are shown in Table 1.

| Western blotting
An analytical 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed, and 30 μg of protein was analyzed.For immune blotting, an electroblotting apparatus transferred proteins in the SDS gels onto polyvinylidene difluoride membranes (Millipore).Primary and secondary antibodies are shown in Table 2. Mouse or rabbit IgG antibodies coupled to HRP were used as secondary antibodies.An enhanced chemiluminescence kit (Millipore) was used for detection using the iBright Imaging System (Thermo Fisher Scientific).The intensity of protein level was measured by ImageJ followed by the protocol. 34

| RNA stability assay
To determine the half-life of fibronectin mRNA, cells were treated with 4 μM actinomycin D (Sigma-Aldrich) to block transcription.Total RNA was extracted from cells by using REzol C&T (PROtech).cDNA was prepared by reverse transcriptase reaction, and fibronectin gene expression was analyzed by RT-qPCR.

| Plasmid construction
The fibronectin gene's 5′-and 3′-untranslated regions were constructed into pGL3 Luciferase Reporter Vector (Promega) using NcoI and HindIII restriction enzymes XbaI and EcoRV restriction enzymes, respectively.Synthetic primers are shown in Table 3.The vector sequence was confirmed by DNA sequencing.

| Luciferase assays
For determination of luciferase activity by using the pGL3 basic Luciferase Reporter Vectors system (Promega), the transfected cells (3 × 10 5 ) were washed with phosphate-buffered saline (PBS) and lysed in 150 μL of Cell Culture Lysis Reagent (Promega) and stood at room temperature for 15 min.The transfection efficiency was estimated by cotransfected with Renilla plasmid, and the Renilla luciferase activity was performed using Renilla Luciferase Assay System (Promega).The cell lysate was centrifuged at 7200 × g for 30 s, and 30 μL of the supernatant solution was mixed with luciferase assay substrates according to the manufacturer's protocol.The luciferase activity was measured by a MiniLumat LB 9506 luminometer (EG&G Berthold) and normalized to Renilla luciferase activity.

F I G U R E 1
The expression of HNRNPs and nucleolar genes in HNSCC.(A,B) The box plots showed the transcriptional expression of HNRNPs and nucleolar genes from the cohort of TCGA-HNSC (n = 604). 62Clinical data were analyzed by UCSC Xena functional genomics explorer. 63The higher expression of genes in HNRNPs and nucleolar such as hnRNPA2B1 and nucleolin (NCL) were labeled with a red arrow.(C) Kaplan-Meier curves showed HNSCC patient overall survival that was retrieved from the TCGA database and categorized into four subtype groups (Atypical n = 66, Basal n = 96, Classical n = 48, and Mesenchymal n = 74).The p-value of the Log-rank test and Hazard ratio (HR) compared patients with high and low expression of nucleolin and hnRNPA2B1 that were shown in the figures.The 50 μL of conditioned media harvested from cultured HNSCC cells were transferred to a black 96-well plate and followed by the procedure of Amplite Universal Fluorimetric MMP Activity Assay Kit (Green fluorescence, AAT bioquest).Briefly, 2 mM amino-phenyl mercuric acetate (APMA) was mixed with a conditioned medium and then incubated at 37°C for 3 h.After incubation, 100 μL of green substrate was then added into each well and further incubated for 1 h.The fluorescence intensities of the background and samples were measured using SpectraMax iD3 Multi-Mode Microplate Reader (Ex: 495 nm, Em: 525 nm).
2.9 | Cell adhesion assay: the dynamic flow circulation system HMEC-1 or any other cell types were seeded into the channel slides (Ibidi) coated with fibronectin (F4759, Sigma-Aldrich) for 24 h to form a monolayer.Tumor cells were treated with 50 ng/mL EGF for 3 h and then labeled with 1,1′-Dioctadecyl-3,3,3′,3′-Tetramethylindocarbocyani ne Perchlorate (DiI) (Invitrogen).DiI-labeled tumor cells were added to the dynamic flow circulation system for 30~60 min and then gently washed with PBS to remove nonadherent cells on slides and fixed with 4% PFA.The adherent cells were imaged using an Upright fluorescence microscope Olympus BX-61.Adherent cells were calculated by Image J. The circulation rate at 3 mm/s mimics the flow rate of microvascular in humans. 35The experiments were conducted in a humidified 5% CO 2 incubator at 37°C.

| Transfection of cells with siRNA oligonucleotide or plasmids
Transient transfection of cells with 40 nM siRNA oligonucleotides and plasmids was performed using RNAiMAX and Lipofectamine 2000 (Invitrogen), respectively, according to the manufacturer's instructions with slight modifications.The siRNA IDs were as follows: Scramble (siRNA ID: D-001810-10-50) (Dharmacon); fibronectin (siRNA IDs: HSS103780, HSS103782); nucleolin (siRNA IDs: HSS106985, HSS106986); hnRNPA2B1 (siRNA IDs: S6713) (Thermo).For transfection, the lipid complex was diluted in Opti-MEM medium (Invitrogen) and incubated with cells for 6 h at room temperature.Following the removal of Opti-MEM medium and replacement with fresh culture medium, the cells were further incubated for an additional 18 h unless stated otherwise.

| shRNA lentiviral vector
RNA interference vectors used in this study were obtained from the National RNAi Core Facility in the Institute of Molecular Biology, Academia Sinica (Taipei, TWN).The lentivirus of hairpins targeting fibronectin, integrin β1 (ITGB1), and LacZ were packaged and obtained from RNAi Core of National Cheng Kung University Hospital in the pLKO.1 lentiviral backbone.The accession numbers in the RNAi core were as follows: shFibronectin: TRCN0000064828, shITGB1: TRCN0000029648, shLacZ: TRCN0000072223.Briefly, 1 × 10 5 cells were seeded to each well in 6-well plates overnight.Before lentivirus infection, the culture medium was changed to fresh media containing 8 μg/mL polybrene (Sigma-Aldrich).Lentivirus was added to the cells at MOI = 3.Following additional incubation of infected cells overnight, infected cells were selected with 2 μg/mL puromycin (Sigma-Aldrich) for 1 week and then maintained with 1 μg/mL puromycin before experiment use.mL streptavidin-HRP (Sigma-Aldrich).The medium in the lower chambers was then collected and assayed for HRP activity by mixing with 100 μL of NeA-Blue tetramethylbenzidine (TMB) substrate (Clinical Science Products).After incubation for 30 min, the reaction was terminated using 100 μL 1% HCl.The absorbance (OD, 450 nm) was detected by using SpectraMax iD3 Multi-Mode Microplate Reader.

| Transwell migration and invasion assay
Transwell migration and invasion assay was performed using transwell inserts (polyethylene terephthalate membranes with 8 μm pores, Millipore).Briefly, for the migration assay, cells (5 × 10 4 ) were seeded in upper chamber containing a serum-free medium and a lower chamber containing a 10% FBS culture medium.After 24 h, cells in the upper chamber were removed, and migrative cells at the bottom of the membrane were fixed with 4% paraformaldehyde and stained with 0.5% crystal violet (Sigma-Aldrich).Crystal violet staining cells were then solubilized with 10% acetic acid, and absorbance (OD, 595 nm) was measured using SpectraMax iD3 Multi-Mode Microplate Reader.For invasion assay, cells were seeded in the upper chamber coated with 1:20 dilution of Matrigel in a 2% FBS culture medium.After 72 h, the invaded cells were measured as the description above.

| Tumor extravasation and metastasis assay in an animal model
Tumor metastasis was determined by intravenous (tail vein) injection of cancer cells into 6 -week-old male severe combined immunodeficient (SCID) mice.Briefly, 1,1 ′dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) labeled cells (5 × 10 5 ) were resuspended in 100 μL of PBS and then injected into the tail vein of mice.Animals were sacrificed up to 48 h after injection.All mice were treated followed by the ethical methods.The lungs were fixed with 4% paraformaldehyde and 30% sucrose and finally embedded in FSC 22 (#3801480) (Leica) for cryosectioned (8 μm).Immunohistochemistry (IHC) was then performed to determine the location of the vessel with antibody CD31 (ab28364) (Abcam).For the lung nodule formation assay, 5 × 10 5 tumor cells were intravenously injected into 6-week-old male SCID mice.Mice were sacrificed after 9 weeks.The tissues were paraffinembedded and then H&E staining was done by the NCKU core lab.The whole slide screening was performed using TissueGnostics FACS-like Tissue Cytometry.All mice were obtained from the National Cheng Kung University (NCKU) Laboratory Animal Center (Tainan, TWN).
The animal study was approved (Approved No. NCKU-IACUC-108-093) by the IACUC of the Laboratory Animal Center, Medical College, NCKU.

| Statistical analyses
Statistical analysis was performed with either t-test (for comparison between two groups) or a one-way ANOVA analysis of variance (Tukey's or Newman-Keuls' post-tests, for comparison among multiple experimental groups) using GraphPad Prism 6.0 software (GraphPad Software, San Diego, CA, USA).The in vitro data are presented as the mean ± standard deviation (SD) determined from three independent experiments.A Fisher's exact test and Kendall's tau (τ)-b correlation analysis were used to examine the relationship between the expression levels of fibronectin, nucleolin, MMPs, and various clinicopathologic features.A p-value less than .05shows the significance and is denoted by an asterisk (*).
The p-values less than .01 and .001are indicated by ** and ***, respectively.n.s.: no significant difference.

EGF-enhanced metastasis of HNSCC cells
RBPs such as HuR and DDX3 levels were correlated with oral cancer progression. 29However, the molecular family, respectively (Figure 1A,B).Therefore, we focus on investigating whether hnRNPA2B1 and nucleolin impact the progression of HNSCC.As shown in Figure 1C, hnRNPA2B1 and nucleolin were significantly associated with overall survival in mesenchymal but not atypical, basal, and classical tissue samples.The previous report also showed that hnRNPA2B1 could promote epithelial to mesenchymal transition in head and neck cancer. 36o verify the correlation between hnRNPA2B1 and mesenchymal properties, we examined the expression of MMPs in hnRNPA2B1 overexpressing HNSCC cells.The hnRNPA2B1 could induce expressions of MMP1, MMP2, MMP3, and MMP9 (Figure 2A).In addition, the MMP activity was increased in hnRNPA2B1 overexpressing cells (Figure 2B).Since the activation of growth factor receptors, including EGFR, is essential for HNSCC progression, 5,6 and we also found that EGF could activate EGFR in various HPV-positive and -negative HNSCC cell lines (Figure S2), the correlation between RBPs and molecules involved in EGFR pathway-based gene signatures, such as EGFR, KRAS, BRAF, and PIK3CA in HNSCC, was analyzed.The results showed that RBPs (hnRNPA2B1/nucleolin) and EGFR signaling pathways (EGFR/KRAS/BRAF/PIK3CA) were coincidently presented in HNSCC tissue samples (Figure 2C).These results indicate the possible effect of RBPs on EGFR signaling-mediated HNSCC progression.Indeed, we found that the depletion of nucleolin and hnRNPA2B1 significantly inhibited EGF-induced genes in regulating mesenchymal properties (Figure 2D).The inhibition of EMT proteins such as vimentin and N-cadherin (CDH2) in the knockdown of nucleolin and hnRNPA2B1 was also verified using various HNSCC cell lines (Figure S3).These results suggest that RBPs may regulate metastasis of HNSCC.Intriguingly, among EMT markers, the inhibition of EGF-induced fibronectin was the most significant (Figure 2D).In addition, EGF-triggered migration and invasion of HNSCC cells were also inhibited in the depletion of nucleolin and hnRNPA2B1 (Figure 2E).These results reveal that RBPs, specifically hnRNPA2B1 and nucleolin, regulate the EGF-induced metastasis of HNSCC cells.

| EGF enhances the stability of fibronectin mRNA through the 5′-UTR in HNSCC cells
To further pursue the molecular mechanism in regulating the interplay between EGF-enhanced HNSCC metastasis and RBPs-targeted mesenchymal markers, we try to analyze whether one of the most responsive RBPstargeted genes, such as fibronectin, regulates RBPsassociated tumor metastasis.Because the molecular mechanisms in regulating EGF-induced fibronectin remain unknown, thus, we first analyzed the correlation of EGFR and fibronectin in HNSCC tissues.The results showed that EGFR and fibronectin presented higher expression in tumors than that in normal tissues (Figures 3A and S4A).The concurrent high expression of fibronectin and EGFR was observed in HNSCC originating from squamous cells compared with the surrounding normal tissues (Figure 3B).The high expression of EGFR and fibronectin was also correlated with a lower survival rate in HNSCC patients (Figure 3C).Intriguingly, the concurrent expression of EGFR and fibronectin was significantly correlated with late-stage HNSCC (Figure 3D), as like RBPs correlated with mesenchymal tissues (Figure 1C).These results suggest the possibility that the expression of fibronectin might be dependent on the activation of EGFR signaling and confer tumor metastasis.To investigate the dynamic expression of RBPs-targeted fibronectin in EGF-treated HNSCC cells, cell lines developed from original tumor tissues oral, nasopharyngeal, and hypopharyngeal were used.The fibronectin mRNA expression was quantified using RT-qPCR.As shown in Figures 4A and S4B, EGF significantly induced fibronectin gene expression in a timedependent manner in both HPV-negative and -positive HNSCC cell lines.In addition, protein expression and secretion of fibronectin were also increased in EGFtreated cells (Figures 4B and S4C).The increase in gene expression might occur through modulation of transcriptional activation or mRNA stability.Therefore, the Lucia luciferase reporter gene system and cells-treated actinomycin D were used to study transcriptional F I G U R E 4 EGF induces fibronectin expression through the 5′ untranslated region.(A,B) FaDu cells were treated with or without 50 ng/ mL EGF for periods of time as indicated.RT-qPCR analysis was performed to detect fibronectin and GAPDH mRNA levels.Ctrl: control.(A) The expression of fibronectin and αtubulin harvested from cell lysate and conditioned medium (CM) was examined by western blotting.(B).(C) FaDu cells were transfected with the expression vector of pDRIVE5Lucia-hFibronectin (InvivoGen) and then treated with 50 ng/ mL EGF (i) or 400 μM oleic acid (OA) (ii) for 24 h.Relative Lucia activities from total lysate and CM were measured and normalized with protein concentration.(D) FaDu cells were treated with 50 ng/mL EGF for 6 h and then treated with 4 μM actinomycin D for periods of time as indicated.Fibronectin mRNA was examined by RT-qPCR.The first time point of mRNA expression levels was quantified as 100%.(E) The luciferase reporter constructs containing 3′-UTR and 5′-UTR of fibronectin were illustrated (upper panel).FaDu cells were transfected with the 5′-UTR or 3′-UTR constructs and treated with 50 ng/mL EGF for 24 h.The luciferase activity was examined (lower panel).n.s., no significance; FN1, fibronectin.
activity and mRNA stability.The results showed that EGF did not affect Lucia luciferase activity analyzed from cell lysates and conditioned medium compared with the positive control of the oleic acid-treated condition (Figure 4C).On the other hand, EGF dramatically enhanced fibronectin mRNA stability (Figure 4D).To analyze the possible effect of 5′-and 3′-untranslated regions of the fibronectin gene on regulating mRNA stability, the constructed plasmids containing the 5′-UTR or 3′-UTR of the fibronectin gene were used (Fig- ure 4E).The results showed that the 5′-UTR but not the 3′-UTR of the fibronectin gene was activated by EGF (Figure 4E).These results suggest that 5′-UTR was essential for the EGF-enhanced mRNA stability of the fibronectin gene.

RBPs are essential for EGF-induced fibronectin expression in HNSCC cells
Our previous report showed that EGF induces PTX3 and ANGPTL4 expression by activating the ERK and NF-κB signaling pathways in HNSCC cells. 37,38To confirm whether EGFR-activated downstream signaling pathways, such as ERK and NF-κB, are also involved in EGF-induced fibronectin expression, the effects of ERK and NF-κB inhibition on fibronectin expression were examined by treating cells with U0126 and parthenolide, respectively.As shown in Figure 5A, inhibition of EGFR using EGFR-TKI significantly blocked EGF-induced fibronectin mRNA expression, which was consistent with the inhibitory effects of U0126 and parthenolide on fibronectin expression as determined by mRNA, protein, and 5′-UTR activation analysis (Figure 5B).These results indicate that the activation of EGFR, ERK, and NF-κB is essential for EGFinduced fibronectin expression.The UTRs of mRNA are regulated by various RBPs, including nucleolin and hnRNPs. 39Our pilot survey from the TCGA-HNSC dataset also revealed that nucleolin expression strongly correlated with hnRNPA2B1 (Figure S5A).Although the depletion of RBPs inhibited EGF-induced fibronectin gene expression (Figure 2D), to investigate whether nucleolin and hnRNPA2B1 are involved in regulating EGF-enhanced fibronectin mRNA stability, siRNA knockdown of nucleolin and hnRNPA2B1 was carried out.As shown in Figure 5C,D, the depletion of nucleolin and hnRNPA2B1 significantly inhibited EGF-enhanced fibronectin mRNA expression and stability.In addition, EGF-induced fibronectin protein expression was also blocked in the knockdown of nucleolin and hnRNPA2B1 (Figure 5E).The role of nucleolin in regulating the EGFactivated 5′-UTR of fibronectin was further confirmed, as shown by the inhibitory effect of nucleolin siRNA on luciferase activity (Figure S5B).The overexpression of hnRNPA2B1 also promoted fibronectin protein, mRNA expression, and mRNA stability (Figures 6A-C and S6).These results reveal that nucleolin and hnRNPA2B1 are essential for EGF-induced fibronectin expression through stabilizing mRNA.

| The fibronectin/integrin signaling contributes to EGF-induced interaction between HNSCC and endothelial cells
The tumor microenvironment in HNSCC is composed of a mix of tumor cells and stromal cells, such as endothelial cells, which stimulate neovascularization and supply oxygen and nutrients to the tumor. 3In addition, the interaction between tumor and endothelial cells supports tumor growth and is a committed step for tumor cell intravasation and extravasation. 40To determine whether EGFR/RBPs axis-induced fibronectin mediates tumor-endothelial cell interactions, we developed an in vitro circulating model that mimics the circulating system.The system contains one-way forced-flowing tumor cells and cultured endothelial cells on the slide (Figure 7A).Intriguingly, the results F I G U R E 6 Overexpression of hnRNPA2B1 promotes fibronectin mRNA stability and expression.(A,B) FaDu cells were transfected with pCMV6-A2B1 expression vector or treated with 50 ng/mL EGF for 18 h.The protein expressions of fibronectin (FN1), hnRNP A2/B1 (A2B1), and αtubulin were examined using western blotting.(A).The expression of fibronectin and GAPDH mRNA was examined using RT-qPCR (B).(C) Cells were transfected with the pCMV6-A2B1 expression vector and then treated with 4 μM actinomycin D for periods as indicated.Fibronectin mRNA was examined by RT-qPCR.The first time point of mRNA expression levels was quantified as 100%.
showed that EGF specifically triggered tumor-endothelial cell interactions but not tumor-tumor, tumor-fibroblast, or tumor-monocyte cell interactions (Figures 7B and S7).Moreover, EGF-induced tumor-endothelial cell interactions were also inhibited by fibronectin depletion (Figure 7C).To further confirm the effect of fibronectin on cell-cell interactions, fibronectin was overexpressed in HNSCC cells.The results showed that fibronectin significantly triggered tumor-endothelial cell interactions (Figure 7D).These results suggest that fibronectin expression is essential for EGF-induced tumor-endothelial cell interactions.
Previous studies have shown that fibronectin-mediated focal contact occurred through the binding of integrin α5β1. 41The inhibition of α5β1 decreased the adhesion, proliferation, and invasion of ovarian cancer cells. 42We also found that downstream molecules of fibronectin/ integrin signaling, such as FAK, were activated by EGF (Figure S8A).Thus, we further studied whether the activation of integrin signaling was required for EGF-and fibronectin-triggered tumor-endothelial cell interactions.As shown in Figure S8B,C, the inhibition of integrin signaling by the RGD-mimetic integrin inhibitor significantly repressed tumor-endothelial cell interactions in EGFtreated or fibronectin-overexpressed cells.In addition, the depletion of integrin β1 inhibited the interaction, confirming that EGF-induced tumor-endothelial cell interactions occurred through the integrin signaling pathway (Figure S8D).These results suggest that the fibronectin/integrin axis is essential for EGF-induced unique interactions between tumors and endothelial cells.

| RBPs-regulated fibronectin is essential for EGF-triggered vessel permeability and tumor cell extravasation
The effect of RBPs-regulated metastasis may be correlated with the permeability of blood vessels in EGF-treated conditions.Therefore, we further investigated whether EGFR/RBPs axis-induced fibronectin also regulates the permeability of endothelial cells and promotes tumor metastasis.As shown in Figure S9A, the conditioned medium harvested from EGF-treated HNSCC cells significantly triggered the permeability of endothelial cells in a timedependent manner.The permeability of endothelial cells was decreased in a conditioned medium harvested from EGF-treated but fibronectin-depleted cells (Figure S9B).Although our previous studies have already shown that the knockdown of PDK1 inhibits EGF-induced tumor cell metastasis and is associated with the downregulation of fibronectin, 21,25 whether the fibronectin directly regulates EGF-promoted HNSCC metastasis remains to be clarified.
To further investigate the molecular mechanisms of fibronectin in regulating HNSCC metastasis, the expression of MMPs and the migratory and invasive abilities of cells were examined in fibronectin-depleted cells.As shown in Figures 8A and S10, EGF-induced expression, and activity of MMPs were significantly inhibited in fibronectindepleted and RGD-treated cells.The inhibition of the fibronectin/integrin axis using RGD peptide and knockdown of fibronectin also significantly inhibited HNSCC cell migration and invasion (Figures 8B and S11).These results suggest that the fibronectin/integrin axis contributes to EGF-induced metastatic properties of HNSCC cells.We further examined whether nucleolin, hnRNPA2B1, and fibronectin were essential for EGF-enhanced tumor cell metastasis in vivo.The extravasation of HNSCC cells was analyzed using a mouse model.Notably, the extravasation of EGF-treated cells increased dramatically in surrounding lung tissues (Figure 9A).Intriguingly, the depletion of fibronectin, nucleolin, and hnRNPA2B1 in cells significantly inhibited EGF-promoted HNSCC cell extravasation (Figure 9A).In addition, EGF-promoted metastatic nodules in the lungs were inhibited in cells with the depletion of fibronectin or treated with RGD peptide (Figure 9B).These results revealed that RBPs-mediated fibronectin expression and activation of the fibronectin/integrin axis are essential for EGF-induced HNSCC cell metastasis.

| DISCUSSION
HNSCC patients may benefit from EGFR-targeted therapies due to EGFR overexpression in over 90% of HNSCC F I G U R E 7 EGF-induced fibronectin enhances the interaction between tumor and endothelial cells.(A) The schematic diagram of the instrument for detecting cell-cell interaction by dynamic flow system.The DiI-labeled tumor cells circulated pass through the monolayer of endothelial cells cultured in chambers.(B) FaDu cells were treated with 50 ng/mL EGF for 3 h, followed by circulating in the dynamic flow system for 30 min.The attachment of DiI-labeled tumor cells was imaged using a microscope (lower panel).Original magnification, ×20.The number of tumor cells was calculated by analyzing at least four sections and five fields (upper panel).(C) FaDu cells were transfected with 40 nM fibronectin (FN1) siRNA or scrambled oligonucleotides (SC), followed by treatment 50 ng/mL EGF for 3 h.The attachment of circulating DiI-labeled tumor cells was calculated by analyzing at least four sections and five fields (i).The protein expression of fibronectin and αtubulin was examined using western blotting (ii).(D) FaDu cells were transfected with various amounts of fibronectin-expressing vector (OE-FN1) for 36 h.The attachment of circulating DiI-labeled tumor cells was calculated by analyzing at least four sections and five fields (i).The protein expression of fibronectin and αtubulin was examined using western blotting (ii).tumors. 43However, the response rate for cetuximab was less than 20% in treating HNSCC, despite HPV status, and there was no improvement of the outcome in combination with platinum-based chemotherapy. 44,45Understanding EGFR signaling in regulating HNSCC metastasis may confer prognosis and combination therapies targeting multiple pathways for treating HNSCC.In this study, for the first time, we identified that RBPs, including nucleolin and hnRNPA2B1, regulated fibronectin expression in EGF-treated HNSCC.The RBPs-regulated fibronectin/integrin axis impacts EGFR-promoted HNSCC-endothelial cell interactions, followed by extravasation and the formation of metastatic lung nodules.These results suggest that fibronectin and RBPs are significantly associated with HNSCC metastasis, which may confer poor prognosis and response to chemotherapy for patients.
Mounting evidence has shown that RBP-mediated RNA modifications, such as stabilizing mRNA strands, are critical for regulating the expression and function of oncogenes. 46For example, hnRNPA2B1 is essential in tumor development and progression. 47,48The depletion of hnRNPA2B1 is correlated with increased E-cadherin expression and downregulation of Twist1 and Snail1 in lung cancer cells. 49The knockdown of hnRNPA2B1 reduces the migratory and invasive abilities of gastric cancer cells. 50Moreover, the depletion of hnRNPA2B1 impaired tumor angiogenesis, as illustrated by the downregulated expression of VEGF-A. 51However, the details of hnRN-PA2B1 in regulating EMT and metastasis are unclear.In the in vitro cell model and animal studies, we found that hnRNPA2B1 was essential for EGF-induced fibronectin expression through the 5′-UTR and the activation of integrin signaling to promote HNSCC cell metastasis.In addition, hnRNPA2B1-regulated fibronectin was also involved in EGF-triggered increases in the permeability of endothelial cells and interactions between tumor cells and endothelial cells.Although the recent report shows that hnRNPA2B1 promotes epithelial to mesenchymal transition through activation of Akt/PKB signaling in head and neck cancer, 36 the impact of hnRNPA2B1 on HNSCC metastasis remains incompletely investigated.Our studies showed for the first time, to our knowledge, that hnRNPA2B1 was not only activated by EGF but also triggered the fibronectin/integrin axis and its downstream signaling to promote HNSCC metastasis.These results reinforce the role of hnRNPA2B1 in HNSCC cell metastasis by regulating tumor-endothelial cell interactions, vascular permeability, and EMT activation.
In addition to hnRNPA2B1, nucleolin is one of the most multifunctional RBPs.3][54][55][56] Although the knowledge accumulated from previous reports to interpret nucleolin plays a role in tumor progression, the mechanisms regulating tumor metastasis by nucleolin are poorly described.On the other hand, nucleolin is overexpressed in several types of cancer and is correlated with increased malignancy and poorer outcomes. 57However, the interplay between nucleolin and tumor metastasis and the effect of nucleolin on EMT gene expression remain unclear.In this study, we reveal that nucleolin contributed to EGF-enhanced HNSCC cell metastasis through the upregulation of EMT markers, including fibronectin and MMPs.The previous report also shows that MMP9 mRNA and translation are regulated by nucleolin, suggesting that nucleolin may contribute to the invasion and metastasis of tumor cells. 54Although nucleolin has been found to interact with diverse mRNAs, including p53, EGFR, IL-2, BCL-2, APP, and COX-2, 57 we further identified the interaction between nucleolin and fibronectin in EGF-treated HNSCC cells.Our studies provide evidence to suggest that nucleolin plays diverse cellular functions by interacting with unique mRNAs, such as fibronectin, to promote cancer EMT.These results strongly support the role of nucleolin in promoting tumor metastasis through the regulation of fibronectin/integrin signaling and EMT markers.
Intra-and extra-tumoral fibronectin expression impacts tumorigenesis in various cancer types to regulate tumor invasion and metastasis.For example, the COX-2/fibronectin axis is essential for nicotine-enhanced colon cancer cell migration. 58Sustained fibronectin expression in the stromal compartment of HNSCC impacts tumor metastasis. 17Specifically, the activation of integrin signaling via fibronectin binding is essential for cancer metastasis.For example, integrin engagement with fibronectin, followed by activation of the FAK/MMP9 axis, resulted in an enhancement of migrating and invasive lung cancer cells. 59Our previous studies showed that fibronectin regulates MMP expression via integrin signaling. 60In this study, we also found upregulation F I G U R E 8 EGF induces tumor cell migration and invasion through the fibronectin/integrin axis.(A) FaDu cells containing the depletion of fibronectin (shFN1) and lacZ control (shLacZ) were treated with 50 ng/mL EGF for 18 h (i).In addition, cells were treated with 100 μM RGD and 50 ng/mL EGF for 18 h (ii).MMP-1, -3, -9, and GAPDH mRNA expression were examined using RT-qPCR.(B) FaDu cells were transfected with 40 nM fibronectin siRNA (si-FN1) or scrambled oligonucleotides (SC) (i) or treated with 100 μM RGD (ii), followed by treating 50 ng/mL EGF for 24 h or 48 h.Migrating and invading cells were stained with crystal violet and imaged with a microscope (right panel) and then solubilized in 10% acetic acid for quantification of migrating and invasive levels, as shown in columns (left panel).The absorbance was measured using SpectraMax iD3 Multi-Mode Microplate Reader (wavelength: 595 nm). of intra-tumor fibronectin expression and secretion induced by the EGFR/RBPs axis, promoting tumor-endothelial cell interactions.The cell interactions were also inhibited when tumor cells were treated with the integrin inhibitor RGD.These results were consistent with a previous report showing that RGD peptides and anti-α5β1 and αvβ3 inhibited the adhesion and proliferation of ovarian cancer cells. 42n addition, we found that EGF-induced cell metastasis was attenuated by inhibiting the fibronectin/integrin axis.These results strongly support that EGF-regulated HNSCC F I G U R E 9 EGF-induced tumor cell metastasis relies on RNA-binding proteins and the fibronectin/integrin axis.(A) FaDu Cells were transfected with 40 nM fibronectin (FN1) and hnRNPA2B1 (A2B1) siRNA or scrambled oligonucleotides (SC), followed by treating 50 ng/ mL EGF for 3 h and then injected intravenously into the tail vein of mice.At 48 h after injection of tumor cells, the mice were sacrificed to examine metastatic tumor cells surrounding the lung tissue as described in "Materials and Methods."Tumor cell penetration was imaged using a microscope (left panel).Original magnification ×20; DiI-labeled tumor cells (red); CD31-labeled blood vessels (green); DAPI-labeled nucleus (blue).The number of tumor cell extravasation was calculated by analyzing at least five sections and six fields (right panel).Dots indicate the number of mice.(B) FaDu cells were transfected with 40 nM fibronectin siRNA (si-FN1) or treated with 100 μM RGD, followed by treating 50 ng/mL EGF for 3 h and then injected intravenously into the tail vein of mice.To evaluate lung metastasis, mice were sacrificed up to 9 weeks after injection.H&E staining of lung tissue was imaged with the whole slide screening using TissueGnostics FACS-like Tissue Cytometry (left panel).The number of micronodules was counted under a microscope (right panel).Dots indicate the number of mice.adhesion, migration, and extravasation rely on the activation of the fibronectin/integrin axis.
In summary, our study highlights several important points.We show that RBPs, specifically nucleolin and hnRNPA2B1, play a critical role in regulating fibronectin expression and tumor metastasis in EGFR-activated HNSCC.We demonstrate that the increase in fibronectin was mediated by EGF-activated ERK and NF-κB signaling pathways, followed by RBPs-regulated 5′-UTR activation of the fibronectin gene.Last, we show that the EGF-induced fibronectin/integrin axis triggers tumor-endothelial cell interactions and permeability of endothelial cells, which confer tumor metastasis.Although the current strategy to interfere with the function of an RBP by small-molecule drugs still has some limitations, 61 the approaches to modulate the interaction with its target RNA remain underdeveloped.We provide an alternative strategy to inhibit RBP cellular function by blocking fibronectin/integrin using RGD.Our findings reveal that fibronectin may also be a diagnostic marker of tumor metastasis in EGFRoverexpressing HNSCC.The blocking RBPs through inhibiting fibronectin/integrin signaling combined with EGFR targeting therapy may provide a better outcome for HNSCC patients.

T A B L E 3
Synthetic primers for the 5′-and 3′-untranslated regions (UTR) of fibronectin.Sense Antisense 5′UTR CCCAA GCT TCT GCA CAG GGG GAGGAGA CATGC CAT GGT TGA GAC GGT GGGGGAGA 3′UTR AAAGG CCG GAC TAT TCA AAT ACT TTT AAT ATC TTAGTCATTTT CGCGC GAT ATC ACT ATT CAA ATA CTT TTA ATA TCT TAGTCATTTT T A B L E 2 Primary and secondary antibody.

HMEC- 1
cells (2 × 10 5 ) were seeded on a transwell insert (3 μm; Millipore) to form a monolayer.After incubation with a conditioned medium harvested from tumor cell culture, the medium in the upper chamber was refreshed with 200 μL of serum-free medium containing 4 μL/ F I G U R E 2 Nucleolin and hnRNPA2B1 are essential for EGFR-associated EMT signatures and tumor cell migration and invasion.(A) FaDu cells were transfected with the pCMV6-A2B1 expression vector for 24 h.The expressions of MMPs and GAPDH mRNA were analyzed by RT-qPCR.(B) UM-SCC1 cells were transfected with the pCMV6-A2B1 expression vector for 24 h.The conditioned media were harvested for analyzing MMP activity as described in the materials and methods.BG: background, EV: empty vector.(C) Concurrent expression of RBPs (hnRNPA2B1/ nucleolin) and EGFR signaling pathways (EGFR/KRAS/BRAF/PIK3CA) was determined in the HNSCC tumor tissues (n = 518) from the TCGA database and analyzed by GEPIA2 (Pearson's correlation coefficient and R square value were shown in the figures).64TPM: transcripts per million.(D,E) FaDu cells were transfected with 40 nM nucleolin (NCL) and hnRNPA2B1 (A2B1) siRNA or scrambled oligonucleotides (SC) for 24 h, followed by the treatment of 50 ng/mL EGF for 24 h.EMT markers and GAPDH mRNA expressions were analyzed by RT-qPCR (D).Migrating and invading cells were stained with crystal violet and imaged with a microscope (right panel) and then solubilized in 10% acetic acid for quantification of migrating and invasive levels, as shown in columns (left panel) (E).The absorbance was measured at a wavelength of 595 nm.

F I G U R E 3
The expression of EGFR and fibronectin is associated with HNSCC progression.(A) Fibronectin mRNA expression was quantitated from the Oncomine database of Ginos head and neck cancer (normal tissues n = 13; tumor tissues n = 41) (i), Sengupta nasopharyngeal cancer (normal tissues n = 10; tumor tissues n = 31) (ii), and Pyeon oral cavity cancer (normal tissues n = 14; tumor tissues n = 36) (iii) dataset in Oncomine.[65][66][67](B) Concurrent expression of EGFR and fibronectin was determined in subtypes of atypical (n = 67), basal (n = 87), classical (n = 49), and mesenchymal (n = 75) HNSCC in the TCGA database.(C) Kaplan-Meier curves showing HNSCC patient survival were retrieved from the TCGA database and analyzed by GEPIA2 (n = 518).The median value was used to classify patients into high-expression or low-expression EGFR and fibronectin.P-values compared patients with high and low EGFR/fibronectin expression.(D) Concurrent expression of EGFR and fibronectin in tumor tissues with stages I~IV of HNSCC patients (n = 25, n = 96, n = 78, n = 253) in TCGA database was quantitated (Pearson's correlation coefficient and R square value were shown in the figures).TPM: transcripts per million.FN1: fibronectin.mechanisms of RBPs in regulating HNSCC metastasis remain unclear.To verify the role of RBPs in HNSCC progression, we analyzed the gene expression signatures in four phenotypic classes of HNSCC (atypical, basal, classical, and mesenchymal) sourced by the cohort of TCGA-HNSC dataset.Although RBPs such as HNRNPs were not correlated with overall survival and disease-free survival of HNSCC (Figure S1), we found that hnRNPA2B1 and nucleolin were most significantly expressed in the hnRNP family and nucleolar protein | 11 of 21 SHIEH et al.

F I G U R E 5
Nucleolin and hnRNPA2B1 regulate EGF-induced fibronectin expression through ERK/NF-κB pathway.(A) FaDu cells were pretreated with EGFR inhibitors, including 1 μM erlotinib and 1 μM gefitinib for 1 h, followed by treatment of 50 ng/mL EGF for 18 h.The expression of fibronectin and GAPDH mRNA was examined by RT-qPCR.(B) FaDu cells were pretreated with 10 μM U0126 and 10 μM parthenolide for 1 h, followed by treating 50 ng/mL EGF for 24 h.The fibronectin mRNA (i), 5′-UTR activation (ii), and protein expression (iii) were analyzed by real-time quantitative PCR, luciferase activity assay, and western blotting, respectively.(C,D) FaDu cells were transfected with 40 nM nucleolin (NCL) and hnRNPA2B1 (A2B1) siRNA or scrambled (SC) oligonucleotides for 24 h, followed by the treatment of 50 ng/mL EGF for 18 h.The expressions of fibronectin and GAPDH mRNA were analyzed by RT-qPCR (C).After EGF treatment, cells were then treated with 4 μM actinomycin D for periods of time as indicated.Fibronectin and GAPDH mRNA expression were measured by RT-qPCR.The first time point of mRNA expression levels was quantified as 100% (D).(E) FaDu cells were transfected with 40 nM NCL and A2B1 siRNA or scrambled (SC) oligonucleotides for 24 h, followed by the treatment of 50 ng/mL EGF for 24 h.The protein expression of fibronectin (FN1), A2B1, NCL, and αtubulin was examined using western blotting (i).The expression levels of protein intensity were quantified from three independent experiments (ii-iv).

T A B L E 1
Primer sets for RT-qPCR.