Epithelial Splicing Regulatory Proteins 1 (ESRP1) and 2 (ESRP2) Suppress Cancer Cell Motility via Different Mechanisms

Background: The roles of ESRP1 and ESRP2 during carcinogenesis remain unknown.

Results: ESRPs are up-regulated during carcinogenesis but down-regulated in invasive fronts. ESRP1 suppresses expression of the Rac1b isoform, whereas ESRP2 represses epithelial-mesenchymal transition-inducing transcription factors.

Conclusion: ESRP1 and ESRP2 suppress cell motility through distinct transcriptional and/or post-transcriptional mechanisms.

Significance: Our findings reveal a novel molecular network that regulates cancer cell motility.

Abstract

ESRP1 (epithelial splicing regulatory protein 1) and ESRP2 regulate alternative splicing events associated with epithelial phenotypes of cells, and both are down-regulated during the epithelial-mesenchymal transition. However, little is known about their expression and functions during carcinogenesis. In this study, we found that expression of both ESRP1 and ESRP2 is plastic: during oral squamous cell carcinogenesis, these proteins are up-regulated relative to their levels in normal epithelium but down-regulated in invasive fronts. Importantly, ESRP1 and ESRP2 are re-expressed in the lymph nodes, where carcinoma cells metastasize and colonize. In head and neck carcinoma cell lines, ESRP1 and ESRP2 suppress cancer cell motility through distinct mechanisms: knockdown of ESRP1 affects the dynamics of the actin cytoskeleton through induction of Rac1b, whereas knockdown of ESRP2 attenuates cell-cell adhesion through increased expression of epithelial-mesenchymal transition-associated transcription factors. Down-regulation of ESRP1 and ESRP2 is thus closely associated with a motile phenotype of cancer cells.

Introduction

The epithelial-mesenchymal transition (EMT)² is essential for embryonic organogenesis as well as tissue regeneration and wound healing (1), but it is also involved in fibrotic diseases and cancer progression (2, 3). During EMT, individual cells convert from epithelial to mesenchymal phenotypes. EMT is principally characterized by changes in cell morphology, attenuation of cell-cell interaction, and loss of cell polarity, often leading to increased cell motility. These characteristics also contribute to acquisition of malignant phenotypes by cancer cells, including invasion into stromal tissues and distant metastasis (4). The EMT process is controlled by four primary regulatory systems: transcriptional control, alternative splicing, noncoding RNA, and post-translational control (5). Among these, transcriptional control has been explored most extensively. EMT is directly or indirectly controlled by several transcription factors, including δEF1/ZEB1 (zinc finger E-box-binding homeobox 1), SIP1 (Smad-interacting protein 1), Snail, and Twist, which are regarded as the most important regulatory components involved in initiating EMT and/or maintaining mesenchymal phenotypes. These EMT-associated transcription factors repress E-cadherin and up-regulate expression of mesenchymal markers (N-cadherin, vimentin, and fibronectin, etc.) (4, 6, 7).

Recent work has revealed the crucial roles played by RNA splicing events during EMT. The RNA-binding proteins ESRP1 and ESRP2 and the RNA-binding protein FOX2 homologue RBFOX2 have been identified as key regulators of splicing events related to EMT (8). ESRP1 and ESRP2, also called RBM35A and RBM35B, respectively, are involved in maintenance of epithelial cell-specific isoforms, whereas RBFOX2 is required for maintenance of mesenchymal phenotypes (8). We recently demonstrated that ESRP1 and ESRP2 are down-regulated by δEF1 and/or SIP1 during TGF-β-induced EMT (9). Importantly, this down-regulation is essential for EMT progression. Although both ESRP1 and ESRP2 prefer to bind to the UG-rich motif, ESRP2 appears to be less active as a splicing factor (10).

ESRPs play multiple roles in tumor progression. Down-regulation of ESRPs during TGF-β-induced EMT triggers isoform switching of FGF receptors, which sensitizes cells to FGF-2; cooperation of TGF-β with FGF-2 then promotes enhanced EMT with more aggressive phenotypes (11, 12). Furthermore, ectopic expression of ESRP1/RBM35A suppresses malignant phenotypes of colon and breast cancer cells (9, 13), suggesting that ESRPs are tumor suppressors. In contrast, Yae et al. (14) reported that ESRP1 is associated with a lower survival rate in breast cancer patients because it enhances the cysteine/glutamic acid transporter (xCT)-dependent defense against reactive oxygen species in cancer cells, thereby promoting colonization of the lung. Thus, whether ESRPs play positive or negative roles during tumor progression remains controversial. Moreover, although genome-wide determinations of ESRP-regulated exons have predicted that they regulate a large number of splicing events in various genes (10, 15), most of the isoform-specific functions have not been elucidated, except in the cases of CD44, MENA, and Exo70 (14, 16, 17).

In this study, we examined the expression profiles of ESRP1 and ESRP2 in human normal and tumor tissues. The expression levels of both ESRP1 and ESRP2 were low in normal epithelium but up-regulated in precancerous lesions and carcinoma in situ. Expression was maintained in advanced cancer cells but down-regulated in invasive fronts. We also found that knockdown of ESRP1 and/or ESRP2 promoted cell motility in head and neck cancer cell lines that express these proteins. Knockdown of ESRP1 derepressed Rac1b, a self-activating isoform of Rac1, whereas knockdown of ESRP2 resulted in repression of E-cadherin through induction of δEF1/SIP1. We thus conclude that ESRP1 and ESRP2 suppress cell motility through distinct transcriptional and/or post-transcriptional mechanisms.

EXPERIMENTAL PROCEDURES

Cell Lines

Seven human head and neck squamous cell carcinoma (HNSCC) cell lines (SAS, HSC2, HSC3, HSC4, Ca9-22, Kuma-1, and Gun-1), a human uterine cervix squamous cell carcinoma cell line (HeLa), and a human breast adenocarcinoma cell line (T47D) were used in this study. SAS, HSC2, HSC3, and HSC4 cells, established from human oral squamous cell carcinomas, were purchased from the Japanese Collection of Research Bioresources Cell Bank (Osaka, Japan). Ca9-22 cells, established from a human gingival squamous cell carcinoma, were purchased from the Japanese Cancer Research Resources Bank (Tokyo, Japan). Kuma-1 cells, established from a human maxillary sinus squamous cell carcinoma, and Gun-1 cells, established from a human hypopharyngeal squamous cell carcinoma (18), were kindly supplied by Dr. Kazuaki Chikamatsu (Gunma University, Gunma, Japan). These cells were cultured in DMEM (Nacalai Tesque) supplemented with 10% heat-inactivated FBS, 500 units/ml penicillin, and 500 μg/ml streptomycin at 37 °C in a humidified atmosphere containing 5% CO₂.

Immunohistochemistry

Formalin-fixed, paraffin-embedded tissue blocks were sectioned (2 μm thick) and mounted on silane-coated glass slides. These slides were then deparaffinized with xylene, followed by rehydration through graded alcohols to water. For antigen retrieval, deparaffinized sections were placed in plastic Coplin jars and heated in an autoclave at 120 °C for 15–20 min in 10 mm citrate buffer solution (pH 6) or target retrieval solution (pH 9) (Dako S2367). Next, endogenous peroxidase was quenched with 3% (v/v) H₂O₂ for 5 min. After blocking with 1% BSA at room temperature for 2 h, the sections were incubated with a primary antibody diluted with 1% BSA in PBS. Slides were incubated overnight at 4 °C with primary monoclonal antibodies. Primary antibodies and working dilution rates were as follows: anti-ESRP1, 1:100 (Sigma-Aldrich HPA023720); anti-ESRP2, 1:100 (Abcam ab113486); anti-E-cadherin, 1:100 (BD Biosciences 610181); and anti-Rac1b, 1:50 (Millipore 09-271). Slides were then incubated with HRP using a ChemMate EnVision kit (Dako) for 2 h. After the slides were washed twice with PBS, immunoreactivity was visualized with 0.6 nm 3–3′-diaminobenzidine tetrahydrochloride (Dojindo), and the slides were counterstained with hematoxylin. Images were acquired with an Olympus BX53 microscope and DP72 camera and analyzed using Olympus cellSens software. All studies were conducted using protocols approved by the ethics committee of the University of Yamanashi and the Tokyo Medical and Dental University.

Immunofluorescence and Phalloidin Staining

Cells were grown on 24 × 24-mm microscope cover glass (Matsunami), fixed in acetone/methanol for 10 min, and permeabilized for 10 min in 0.3% Triton X-100 in PBS. After blocking with Blocking One Histo (Nacalai Tesque) for 15 min, cells were incubated overnight at 4 °C with anti-E-cadherin primary monoclonal antibody diluted in Blocking One Histo. The cells were incubated with Alexa Fluor 633-labeled goat anti-mouse IgG (Invitrogen A-21052) as a secondary antibody for 2 h and then with DAPI for 15 min at room temperature. Fluorescence was examined using an Olympus FV1000 confocal microscope and analyzed with Olympus FV10-ASW imaging software. Phalloidin staining was performed as described previously (19). Images were acquired with an Olympus BX50 microscope and analyzed using ViewFinder Lite software (Better Light, Inc.). Filopodial length was measured at eight random fields on photographs.

Immunoblot Analysis

Immunoblot analysis was performed as described previously (20). Cells were lysed in lysis buffer containing 20 mm Tris-HCl (pH 7.4), 5 mm EDTA, 150 mm NaCl, 1% Nonidet P-40, 1% aprotinin, and 1 mm PMSF on ice. The protein concentration was measured using the BCA protein assay reagent (Thermo Fisher Scientific). Proteins were separated by SDS-PAGE, followed by electrophoretic transfer onto polyvinylidene difluoride membranes and blocking with 50 mm Tris-HCl (pH 7.4), 150 mm NaCl, and 0.1% Tween 20 containing 5% skim milk. The following primary antibodies were used: anti-ESRP1 (Sigma-Aldrich HPA023719), anti-ESRP2 (Abcam ab113486), anti-E-cadherin (BD Biosciences 610181), anti-Rac1 (Millipore 05-389), anti-Rac1b (Millipore 09-271), and anti-α-tubulin (Sigma-Aldrich DM1A). The blots were incubated overnight at 4 °C. After incubation with HRP-conjugated mouse or rabbit IgG (Jackson ImmunoResearch Laboratories) for 1 h, proteins were visualized using Amersham Biosciences ECL Western blotting detection reagent (GE Healthcare). All images were acquired with a Fujifilm LAS-4000 mini imager and analyzed with Image Reader LAS-4000 software. This software was also used to quantify protein bands on immunoblots.

RNA Interference

Transfection of siRNAs into cells (5 × 10⁵) was performed using Lipofectamine RNAiMAX transfection reagent (Invitrogen). The Stealth RNAi siRNA sequences used in this study were as follow: human ESRP1 (262), 5′-GAGAAGGAGUUGAUCCUGCUGUUCU-3′; human ESRP1 (981), 5′-CGGAGAAGCUCUGGUUAGGUUUGUA-3′; human ESRP2 (862), 5′-CCGAGGUGAUAAAGCAGAAAUACGA-′; human ESRP2 (1126), 5′-GCGUCCGCUAUAUUGAGGUGUAUAA-3; human Rac1b 5′-UGGAGAAACGUACGGUAAGGAUAUA-3; human δEF1, 5′-GACCAGAACAGUGUUCCAUGCUUAA-3′; and human SIP1, 5′-CCACCACAGUGUUACGAAUUGUGAU-3′. The final concentration of the siRNAs used was 5 nm, except for the SIP1 siRNA, which was used at 10 nm. Unless noted otherwise, we used ESRP1 (981) and ESRP2 (862) siRNAs.

Conventional PCR and Quantitative Real-time PCR

Total RNA was extracted from each cell line using the RNeasy mini kit (Qiagen). Two μg of total RNA were then reverse-transcribed into cDNA using the SuperScript VILO cDNA synthesis kit (Invitrogen). Conventional PCR was performed with LA Taq polymerase (TaKaRa). The primers used in conventional PCR are as follows: human CD44, 5′-GCACTTCAGGAGGTTACATC-3′ (sense) and 5′-ACTGCAATGCAAACTGCAAG-3′ (antisense); human Rac1, 5′-GGATCCTTTGACAATTATTCTGCCAATG-3′ (sense) and 5′-CGGACATTTTCAAATGATGCAGG-3′ (antisense); human MENA, 5′-GCTGGAATGGGAGAGAGAGCGCAGAATATC-3′ (sense) and 5′-GTCAAGTCCTTCCGTCTGGACTCCATTGGC-3′ (antisense); and human β-actin, 5′-GGCATCCTCACCCTGAAGTA-3′ (sense) and 5′-GGGGTGTTGAAGGTCTCAAA-3′ (antisense).

All PCR conditions included an initial denaturation for 2 min at 95 °C. Amplification reactions were performed for 30 cycles under the following conditions: 95 °C for 1 min, 98 °C for 20 s, and 60 °C for 30 s, followed by an extension of 1 min at 72 °C. PCR products were separated on 1.5% agarose gels, stained with ethidium bromide, and visualized using a Printgraph AE-6932 gel detection system (ATTO Corp.). The gene encoding β-actin was used as an internal control in conventional PCR. Quantitative RT-PCR using SYBR Green was performed on an ABI 7300 Fast real-time PCR system (Applied Biosystems) as described previously (21). mRNA levels were normalized to the level of the mRNA encoding GAPDH in the same sample. The relative expression levels of target genes were determined by the 2^−(ΔΔCt) method. The primers used were as follows: human ESRP1, 5′-CAATATTGCCAAGGGAGGTG-3′ (sense) and 5′-GTCCCCATGTGATGTTTGTG-3′ (antisense); human ESRP2, 5′-TGCCACAGAGGATGACTTTG-3′ (sense) and 5′-ATTGACTGCTGGGCTCTTTG-3′ (antisense); human E-cadherin, 5′-TGCACCAACCCTCATGAGTG-3′ (sense) and 5′-GTCAGTATCAGCCGCTTTCAG-3′ (antisense); human δEF1, 5′-CAATGATCAGCCTCAATCTGCA-3′ (sense) and 5′-GTCAGTATCAGCCGCTTTCAG-3′ (antisense); human SIP1, 5′-AAGCCCCATCAACCCATACAAG-3′ (sense) and 5′-AAATTCCTGAGGAAGGCCCA-3′ (antisense); human Snail, 5′-TTCTCACTGCCATGGAATTCC-3′ (sense) and 5′-GCAGAGGACACAGAACCAGAAA-3′ (antisense); human Slug, 5′-GCCTCCAAAAGCCAAACTACA-3′ (sense) and 5′-GAGGATCTCTGGTTGTGGTATGACA-3′ (antisense); human Twist, 5′-GGCCGGAGACCTAGATGTCATT-3′ (sense) and 5′-CCACGCCCTGTTTCTTTGAAT-3′ (antisense); and human GAPDH, 5′-GCACCGTCAAGGCTGAGAAC-3′ (sense) and 5′-ATGGTGGTGAAGACGCCAGT-3′ (antisense).

GST-PAK Fusion Protein Binding Assay

Cells (5 × 10⁵) were inoculated, cultured for 48 h, and harvested. Rac1b activation was assessed by GST pulldown assay using the Rac1/Cdc42 activation assay kit (Millipore 17-441).

Wound Healing Assay

Cells were inoculated in 6-well plates. After cells reached confluence, they were scratched with a 10-μl disposable pipette tip. Migration of wound edges was measured at 10 random points on photographs acquired using a Nikon Eclipse TS100 microscope, and then cell migration distance after 12 h was compared with the distance at 0 h. A previous protocol (19) was used for data analysis.

Proliferation Assay

Cells (1 × 10⁵) were inoculated in 6-well plates and cultured for 24, 72, 120, and 168 h. Cells were then trypsinized and counted using hemocytometer.

Statistical Analyses

All statistical analyses were performed using StatMate IV software (ATMS Co.).

RESULTS

ESRP1 and ESRP2 Are Up-regulated during Human Oral Squamous Cell Carcinoma Carcinogenesis

We first assessed ESRP1 and ESRP2 expression during the progression of carcinogenesis from normal epithelium to dysplasia and invasive carcinoma. To this end, we examined human dysplasia and oral squamous cell carcinoma (OSCC) samples by immunohistochemical staining. In normal human squamous epithelium, ESRP1 and ESRP2 were detected only weakly. Most of the ESRP-positive cells were located in the basal layer, whereas only a few were present in the cornified layer (Fig. 1A). To the extent that they were expressed, they were principally localized in the nucleus (Fig. 1). The expression levels of ESRP1 and ESRP2 were higher in dysplastic lesions than in normal epithelium (Fig. 1B). Epithelial dysplasia can be divided into three categories: mild, moderate, and severe. As the degree of cellular atypia rose, the proportion of ESRP1-positive cells increased (Fig. 1C). In carcinoma in situ and advanced OSCC lesions, which can be classified into well or poorly differentiated types, ESRP1 expression in cancer cells was also significantly elevated. The patterns and intensities of ESRP1 expression in each histological type of advanced OSCC were similar to those in dysplastic lesions (Fig. 1D). A similar expression pattern was observed for ESRP2 in dysplasia (data not shown) and advanced OSCC samples (Fig. 1E). In addition, we evaluated the intensities of ESRP1 and ESRP2 immunoreactivity in 49 human OSCC samples. The reactivities for ESRP1 and ESRP2 were elevated relative to those for normal epithelium in 46 of 49 and 40 of 49 OSCC samples, respectively. These results clearly indicate that human ESRP1 and ESRP2 are expressed in both normal epithelium and OSCC and are up-regulated during OSCC carcinogenesis.

FIGURE 1.

ESRPs are up-regulated during human OSCC carcinogenesis. Expression of human ESRP1 and ESRP2 was investigated by immunohistochemical analysis. A and B, ESRP1 (left) and ESRP2 (right) expression (brown) in normal epithelium (A) and the border between normal epithelium and the dysplastic lesion (B). C, ESRP1 expression (brown) in different stages of dysplasia: mild (left), moderate (middle), and severe (right). D, ESRP1 expression (brown) in well (left) and poorly differentiated (right) OSCC tissues. E, ESRP1 or ESRP2 expression (brown) in advanced OSCC. Serial sections were stained with hematoxylin and eosin (H-E; left), anti-ESRP1 antibody (middle), and anti-ESRP2 antibody (right). Scale bars = 100 μm.

Reduction of ESRP1 and ESRP2 Expression in Invasive OSCC

ESRPs are down-regulated in basal-like breast cancer cell lines, which are invasive (9). To determine whether ESRP expression levels are altered in cancer cells during invasion into surrounding stromal tissues, we next analyzed expression of ESRP1 in carcinoma in situ and OSCC with invasive phenotypes. In cancer cells that penetrated through the basement membrane to invade stromal tissues, ESRP1 expression was significantly reduced (Fig. 2, A and B). We also examined ESRP1 and ESRP2 expression in cancer cells that invaded the stroma to deeper locations, as well as in cells that metastasized to a lymph node. In both locations, ESRP1 was re-expressed in the cancer nest (Fig. 2, C, left; D; and E, left) but repressed in the invasive fronts (Fig. 2, C, right; and E, right). These results suggest that expression of ESRPs is plastic during invasion and metastasis of cancer cells.

FIGURE 2.

ESRPs are down-regulated in invasive fronts. Expression of human ESRP1 and ESRP2 was investigated by immunohistochemical analysis. A, ESRP1 expression (brown) in the border between carcinoma in situ (CIS) and the invasive lesion. B, ESRP1 expression (brown) in the basal layer where the basement membrane (BM) was intact (left) or where cancer cells penetrated through the basement membrane (right). C, ESRP1 (left) or ESRP2 (right) expression (brown) in the invasive fronts of advanced OSCC. Arrows indicate the direction of tumor invasion. Each high-power field is shown in the panels surrounded by dashed lines. The gradient expression of ESRP1 or ESRP2 is schematically represented as a black slope. D and E, ESRP1 (left) or ESRP2 (right) expression (brown) in a lymph node with metastasis: cancer nests (D, arrows) or those with invasive fronts in the lymphatic tissue (E, arrows). Each high-power field is shown in the panels surrounded by dashed lines. The gradient expression of ESRP1 or ESRP2 is schematically represented as a black slope. F, representative image of the invasive fronts in advanced OSCC. Serial sections were stained with hematoxylin and eosin (H-E), anti-ESRP1 antibody, anti-ESRP2 antibody, and anti-E-cadherin antibody (brown). High-power fields of the noninvasive region (panels a, c, and e) and the tip of the invasive front (panels b, d, and f) are shown. Scale bars = 100 μm.

We previously reported a positive correlation between expression of ESRPs and E-cadherin in human breast cancer cell lines (9). Consistent with our earlier finding, in invasive fronts, where ESRP1 and ESRP2 levels were reduced, we observed loss or internalization of junctional E-cadherin (Fig. 2F). Thus, ESRP1 and ESRP2 may suppress invasion or cell motility in HNSCC.

Expression of ESRPs in HNSCC Cell Lines

To explore the possible suppressive effects of ESRPs on cell motility, we performed in vitro experiments using human HNSCC cell lines. We first examined ESRP mRNA expression in seven HNSCC cell lines, using HeLa cells for comparison. ESRP1 gene expression was higher in all seven HNSCC cells than in HeLa cells, whereas there was no striking difference in ESRP2 expression between HNSCC cell lines and HeLa cells (Fig. 3A). We also examined endogenous ESRP1 and ESRP2 proteins by immunoblotting (Fig. 3B); the protein expression profiles were similar.

FIGURE 3.

Expression profiles and activities of ESRPs in HNSCC cell lines. A, mRNA levels of ESRP1 and ESRP2 in HNSCC cell lines as determined by quantitative RT-PCR. Expression levels are presented relative to those in T47D breast carcinoma cells, which express both ESRP1 and ESRP2, and each value was normalized to expression of GAPDH in the same sample. HeLa cells were used as a control. Values are means ± S.D. of triplicate measurements. B, protein levels of ESRP1 and ESRP2 in HNSCC cell lines as determined by immunoblot analysis. α-Tubulin was used as a loading control. Bands for each protein were confirmed by siRNAs against ESRP1 (siESRP1) and ESRP2 (siESRP2) in SAS cells. Arrows indicate ESRP1 (upper) and ESRP2 (lower). Asterisks denote nonspecific bands. C, profiles of CD44 variant (CD44v) and standard (CD44s) isoforms in HNSCC cell lines examined by conventional RT-PCR. β-actin was used as an internal control. D and E, knockdown of ESRP1 (D) and ESRP2 (E) in SAS and HSC4 cells. Cells were treated with control siRNA (siControl), ESRP1 siRNA, or ESRP2 siRNA for 48 h. The expression levels of mRNAs and proteins were examined by quantitative RT-PCR (upper) and immunoblotting (lower), respectively. mRNA levels are presented relative to those in cells treated with control siRNA after normalization to the level of GAPDH mRNA (upper). Values are means ± S.D. of triplicate measurements. α-Tubulin was used as a loading control (lower). In E, the arrow and asterisk denote ESRP2 protein and nonspecific bands, respectively. F, expression of CD44 isoforms was examined by conventional RT-PCR in SAS and HSC4 cells after knockdown of ESRP1. β-Actin was used as an internal control. G, expression of CD44 isoforms was examined by conventional RT-PCR after knockdown of ESRP2 in SAS and HSC4 cells. mRNAs extracted from ESRP1 knockdown in SAS and HSC4 cells were used as positive controls. β-Actin was used as an internal control. H, expression of MENA isoforms was examined by conventional PCR in SAS and HSC4 cells after knockdown of ESRP1 or ESRP2. β-Actin was used as an internal control.

Isoform switching of CD44 mRNA is regulated by ESRPs (9, 14, 22). Therefore, to assess ESRP activities in HNSCC cell lines, we analyzed splicing isoforms of CD44 mRNA by conventional PCR (Fig. 3C). In HNSCC cell lines expressing higher levels of ESRP1 (SAS, HSC4, and Ca9-22), CD44 variant isoforms, including CD44v2-10, CD44v6-10, and CD44v8-10, were exclusively expressed. In contrast, only the CD44 standard isoform was expressed in HeLa cells lacking ESRP1. We then designed two siRNAs against ESRP1 and ESRP2 and transfected these siRNAs into SAS and HSC4 cells, in which ESRP1 and ESRP2 are highly expressed (Fig. 3, D and E). Silencing of ESRP1 resulted in switching from the CD44 variant isoforms to the CD44 standard isoform (Fig. 3F), indicating that ESRP1 is active in these cell lines. In contrast, silencing of ESRP2 did not affect CD44 isoform switching (Fig. 3G). We next analyzed splicing isoforms of MENA mRNA, also identified as an ESRP target. There are three isoforms of MENA transcripts, MENA, MENA^11a(+), and MENAΔv6; MENA^11a(+) and MENAΔv6 are the epithelial and mesenchymal types, respectively (16). In control cells, MENA^11a(+) was most abundantly expressed. Knockdown of ESRP1 resulted in down-regulation of MENA^11a(+) and up-regulation of MENA and MENAΔv6. Knockdown of ESRP2 increased the MENA isoform, although it did not induce the MENAΔv6 isoform (Fig. 3H). These results indicate that ESRP2 is also active in these cell lines.

Knockdown of ESRPs Promotes HNSCC Cell Migration

Because we observed down-regulation of ESRPs at invasive fronts, we next examined the phenotypic effects of silencing endogenous ESRP1 and/or ESRP2 in HNSCC cell lines. First, we confirmed knockdown of ESRP1 and/or ESRP2, as shown in Fig. 4A. Next, we examined changes in cell morphology. Knockdown of ESRP1 and/or ESRP2 induced cell scattering, and knockdown of ESRP2 appeared to be more effective in SAS cells (Fig. 4B). Although knockdown of ESRPs did not affect cell proliferation in either of these cell types (Fig. 4C), wound closure was faster in knockdown cells than in control cells (Fig. 4D). These findings indicate that knockdown of either ESRP1 or ESRP2 increases cell motility, although double knockdown does not increase motility further.

FIGURE 4.

Knockdown of ESRP1 and/or ESRP2 results in increased cancer cell motility. Cells were treated with control (siControl), ESRP1 (siESRP1), or ESRP2 (siESRP2) siRNA 48 h before the following assays. A, protein expression of ESRP1, ESRP2, and E-cadherin in SAS and HSC4 cells in which ESRP1 and/or ESRP2 were knocked down as determined by immunoblotting. α-Tubulin was used as a loading control. B, phase-contrast images of morphological changes in SAS cells after knockdown of ESRP1 and/or ESRP2. Scale bars = 10 μm. C, proliferation of ESRP1- or ESRP2-silenced SAS (left) and HSC4 (right) cells was examined. Values are means ± S.D. of triplicate measurements. D, wound healing assay of SAS and HSC4 cells after knockdown of ESRP1 and/or ESRP2. The results were quantitated as shown (lower). Migration of the wound edge was measured at 10 randomly chosen points in the photograph. Cell migration distances after 12 h at 10 randomly chosen points were compared with the distances at 0 h. Values are means ± S.D. of triplicate measurements. Similar results were obtained in three independent experiments. p values were determined by Student's t test. *, p < 0.01; n.s., not significant.

Mechanisms of Enhanced Cell Motility after ESRP Knockdown

To further elucidate the mechanisms underlying enhanced cell motility in ESRP-silenced cells, we examined E-cadherin expression and actin dynamics. E-cadherin protein expression was decreased by knockdown of ESRP2, but not ESRP1 (Fig. 4A). Furthermore, knockdown of ESRP2 down-regulated E-cadherin at the transcriptional level (Fig. 5A); knockdown of ESRP1 had a similar but more modest effect. Localization of E-cadherin was not affected by knockdown of ESRP1 alone (Fig. 5B), indicating that ESRP1 affects the E-cadherin status only minimally.

FIGURE 5.

Mechanisms of enhanced cancer cell motility by ESRP knockdown. A, expression of E-cadherin in ESRP1- and/or ESRP2-silenced SAS and HSC4 cells as determined by quantitative RT-PCR analysis. mRNA levels are presented relative to those in cells treated with control siRNA after normalization to the level of GAPDH mRNA. Values are means ± S.D. of triplicate measurements. Similar results were obtained in three independent experiments. p values were determined by Student's t test. *, p < 0.05; ***, p < 0.001. siESRP1, ESRP1 siRNA; siESRP2, ESRP2 siRNA. B, subcellular localization of E-cadherin (white) in ESRP1 and/or ESRP2 knockdown SAS and HSC4 cells as visualized by immunofluorescence. Nuclei were stained with DAPI (blue). Scale bars = 2.5 μm. siControl, control siRNA. C, phalloidin staining of ESRP1 and/or ESRP2 knockdown SAS and HSC4 cells. High-power fields are shown (lower). Scale bars = 1.0 μm. D, quantification of filopodial length. Each number below the graphs indicates the number of measured filopodia. p values were determined by the median test. *, p < 0.001; n.s., not significant.

We also examined the status of the actin cytoskeleton in knockdown cells. Remarkably, knockdown of ESRP1 dramatically promoted formation of long filopodia relative to control cells (Fig. 5, C and D), whereas knockdown of ESRP2 had no detectable effect on the actin cytoskeleton. These results indicate that ESRPs play suppressive roles in the control of cell motility in HNSCC: ESRP1 is involved in the regulation of actin dynamics, whereas ESRP2 principally maintains E-cadherin expression.

ESRP1 Regulates Splicing of Rac1 Isoforms

We next examined the mechanism underlying formation of long filopodia after knockdown of ESRP1. Actin dynamics are regulated by the Rho family of small G-proteins, Rho, Cdc42, and Rac1. Rac1 has a self-activating splice variant, Rac1b, which is characterized by inclusion of exon 3b (Fig. 6A) (23). Exon 3b encodes 19 amino acid residues, the insertion of which induces a conformational change in the switch I and II regions in Rac1b, leading to activation by GDP/GTP exchange independent of guanine nucleotide exchange factors, as well as impaired hydrolysis of GTP (24). A previous report showed that Rac1b promotes cell motility (25). In light of these observations, we assessed the alternative splicing of Rac1 mRNA in HNSCC cell lines.

FIGURE 6.

ESRP1 is associated with regulation of alternative splicing of Rac1 mRNA. A, schematic presentation of alternative splicing of Rac1 mRNA. Insertion of exon 3b (57 bp), which yields the Rac1b isoform, was examined. Primers were designed in exons 3 and 4. An ESRP-binding consensus sequence (UGGUGGGUG) is located 274 bp upstream of exon 3b (shown by the asterisk). Fw, forward; Rv, reverse. B and C, changes in Rac1 isoforms in ESRP1-, ESRP2-, and ESRP1/ESRP2-silenced HNSCC cells as determined by conventional PCR (B) or immunoblotting (C). siESRP1, ESRP1 siRNA; siESRP2, ESRP2 siRNA. In C, α-tubulin was used as a loading control. The quantification of Rac1b protein is shown (right). The density of Rac1b protein bands on Western blot images was measured and quantified. Protein levels are presented relative to those in cells treated with control siRNA (siControl) after normalization to the protein level of α-tubulin. Values are means ± S.D. of triplicate measurements. Similar results were obtained in three independent experiments. p values were determined by Student's t test. *, p < 0.05; n.s., not significant. D, detection of the active Rac1b isoform in ESRP1-silenced HNSCC cells by a pulldown assay using GST-PAK fusion protein. Rac1b (brown) was visualized using a specific monoclonal antibody raised against the amino acid sequence encoded by exon 3b. α-Tubulin was used as an input control. E, immunohistochemical detection of Rac1b in human advanced OSCC. The arrow indicates the direction of tumor invasion. The gradient expression of Rac1b is shown schematically as a black slope.

ESRP1 knockdown increased the level of the Rac1b isoform in SAS and HSC4 cells, whereas ESRP2 knockdown did not (Fig. 6B). Rac1b protein was also significantly up-regulated in ESRP1 knockdown cells (Fig. 6C); the protein was active, as assessed by binding to the GST-PAK fusion protein (Fig. 6D). Thus, ESRP1 appears to repress expression of the Rac1b isoform by excluding variant exon 3b from splicing of Rac1 mRNA. Intriguingly, we noted that an ESRP-binding motif (UGGUGG) (10) is present in the intron upstream of exon 3b (Fig. 6A), suggesting that ESRP1 directly regulates the alternative splicing of Rac1 mRNA. Next, we examined expression of Rac1b in human OSCC tissues by immunohistochemical staining. Consistent with our in vitro findings, the expression levels of Rac1b were increased in invasive fronts (Fig. 6E). ESRP1 and Rac1b are thus reciprocally expressed in OSCC tissues.

We next investigated the role of Rac1b in the promotion of cell motility by ESRP1 knockdown. First, we knocked down Rac1b in ESRP1-silenced HNSCC cell lines by transfecting siRNA against the Rac1b-specific exon 3b (Fig. 7, A and B). Knockdown of Rac1b significantly suppressed the elevated cell motility induced by silencing of ESRP1 (Fig. 7C). In addition, formation of long filopodia was significantly attenuated after knockdown of Rac1b in ESRP1-silenced cells (Fig. 7, D and E). These results demonstrate that ESRP1 suppresses cancer cell motility by repressing the Rac1b isoform.

FIGURE 7.

Knockdown of ESRP1 increases cell motility through up-regulating Rac1b. Shown is the knockdown of the Rac1b isoform in SAS or HSC4 cells. Cells were treated with control (siControl), ESRP1 (siESRP1), or Rac1b (siRac1b) siRNA. After 48 h, RNAs or proteins were extracted and subjected to conventional RT-PCR (A) or immunoblotting (B) using anti-Rac1 antibody. In A, β-actin was used as an internal control. In B, the asterisk denotes Rac1, and α-tubulin was used as a loading control. C, wound healing assay of SAS and HSC4 cells in which ESRP1 alone or with Rac1b were knocked down. The results were quantitated as shown (lower). Migration of the wound edge was measured at 10 randomly chosen points in the photograph. Cell migration distances after 12 h at 10 randomly chosen points were compared with the distances at 0 h. Values are means ± S.D. of triplicate measurements. Similar results were obtained in three independent experiments. p values were determined by Student's t test. *, p < 0.01. D, phalloidin staining of SAS and HSC4 cells in which ESRP1 and Rac1b were knocked down. Scale bars = 1.0 μm. E, quantification of filopodial length. Each number below the graphs indicates the number of measured filopodia. p values were determined by the median test. *, p < 0.001; n.s., not significant.

Knockdown of ESRP2 Increases EMT-associated Transcription Factors in HNSCC

Finally, we sought to clarify the mechanism underlying down-regulation of E-cadherin after knockdown of ESRP2. Because E-cadherin expression is directly regulated by EMT-associated transcription factors, including δEF1, SIP1, Snail, Slug, and Twist, we examined the expression of these factors after knockdown of ESRP2. δEF1 and SIP1 levels were significantly higher in ESRP2-silenced SAS cells, whereas knockdown of ESRP2 in HSC4 cells induced only SIP1 (Fig. 8A). Together, these observations suggest that E-cadherin expression is repressed by δEF1 and/or SIP1 upon knockdown of ESRP2. We then knocked down both δEF1 and SIP1 in ESRP2-silenced SAS cells and SIP1 alone in ESRP2-silenced HSC4 cells (Fig. 8B, upper). Knockdown of δEF1 and SIP1 was not complete, but it was sufficient to compensate for the increase in expression due to knockdown of ESRP2. E-cadherin mRNA and protein levels were significantly restored by knockdown of δEF1 and SIP1 (Fig. 8, B, lower; and C). As shown by immunofluorescence staining (Fig. 8D), E-cadherin re-accumulated, and cell-cell adhesion partially recovered after silencing of δEF1 and SIP1 in ESRP2-silenced SAS cells. Similarly, E-cadherin expression was restored by knockdown of SIP1 in ESRP2-silenced HSC4 cells (data not shown). These results indicate that ESRP2 represses δEF1 and SIP1 to maintain E-cadherin expression in HNSCC cells. In addition, we examined the motility of cells in which δEF1, SIP1, and ESRP2 were silenced. As described above, knockdown of ESRP2 promoted cell motility (Fig. 4D). Additional knockdown of both δEF1 and SIP1 or SIP1 alone canceled the effect of ESRP2 in SAS or HSC4 cells, respectively (Fig. 8E). Thus, we conclude that ESRP2 inhibits cancer cell motility by suppressing EMT-related transcription factors.

FIGURE 8.

ESRP2 maintains E-cadherin expression by repressing the δEF1 family of transcription factors. A, expression of EMT-associated transcription factors in SAS and HSC4 cells in which ESRP2 or both ESRP1 and ESRP2 were knocked down. After incubation for 48 h, gene expression levels were measured by quantitative RT-PCR. mRNA levels are presented relative to those in cells treated with control siRNA after normalization to the level of GAPDH mRNA. Values are means ± S.D. of triplicate measurements. Similar results were obtained in three independent experiments. p values were determined by Student's t test. **, p < 0.01; ***, p < 0.001; n.s., not significant. B–E, SAS cells were transfected with control siRNA (siControl), with ESRP2 siRNA (siESRP2), or with ESRP2 plus δEF1 (siδEF1) and SIP1 (siSIP1) siRNAs. HSC4 cells were transfected with control siRNA, with ESRP2 siRNA, or with ESRP2 plus SIP1 siRNAs. After incubation for 48 h, cells were used for the following assays. B, mRNA expression of ESRP2 (upper), δEF1/SIP1 (middle), and E-cadherin (lower) examined by quantitative real-time PCR. mRNA levels are presented relative to those in cells treated with control siRNA after normalization to the level of GAPDH mRNA. Values are means ± S.D. of triplicate measurements. Similar results were obtained in three independent experiments. p values were determined by Student's t test. *, p < 0.05; **, p < 0.01; ***, p < 0.001. C, protein levels of E-cadherin and ESRP2 as determined by immunoblotting. α-Tubulin was used as a loading control. D, subcellular localization of E-cadherin (white) in SAS cells in which ESRP2 or both ESRP2 and δEF1/SIP1 were knocked down as visualized by immunofluorescence. Nuclei were stained with DAPI (blue). Scale bars = 2.5 μm. E, wound healing assay of SAS and HSC4 cells in which ESRP2 or both ESRP2 and δEF1/SIP1 were knocked down. The results were quantitated as shown (lower). Migration of the wound edge was measured at 10 randomly chosen points in the photograph. Cell migration distances after 12 h at 10 randomly chosen points were compared with the distances at 0 h. Values are means ± S.D. of triplicate measurements. Similar results were obtained in three independent experiments. p values were determined by Student's t test. **, p < 0.01.

DISCUSSION

ESRP1 and ESRP2, which belong to the RBM family of RNA-binding proteins, were originally identified as epithelium-specific splicing regulators in a genome-wide cDNA expression screen aimed at finding key factors that enable alternative splicing of FGF receptor-2 mRNA in epithelial cells (26). They have molecular masses of ∼75 kDa and consist of three conserved tandem RNA recognition motifs (26). Although several recent in vitro studies explored the functions of ESRPs as regulators of EMT-associated alternative splicing, the roles of these proteins in cancer progression remain to be elucidated. Ueda et al. (27) recently reported that ESRP1 is weakly expressed in normal pancreatic ductal cells. Using in situ RNA hybridization, Revil and Jerome-Majewska (28) demonstrated that ESRP1 is expressed in the head region of developing mouse embryos. ESRP1 is also expressed in human breast and pancreatic cancers (9, 27). In contrast, in vivo expression of ESRP2 has not been reported. Moreover, it remains unclear how ESRP1 and ESRP2 expression is regulated in the process of carcinogenesis.

In this study, we examined the expression profiles of ESRP1 and ESRP2 during carcinogenesis using 49 samples of human HNSCC tissue. The process of squamous cell carcinogenesis consists of several developmental steps, including dysplasia, carcinoma in situ, and invasive carcinoma. HNSCC samples are favorable for evaluating expression of ESRP1 and ESRP2 during carcinogenesis because multiple steps associated with HNSCC carcinogenesis can be observed in the same specimens.

We found that ESRP1 and ESRP2 were only weakly expressed in normal oral epithelium, where both could be detected in the basal layers. The expression levels of both proteins in dysplasia and carcinoma in situ were higher than in normal epithelium. Moreover, they were also highly expressed in advanced OSCC and cancer nests in metastatic lymph nodes. These findings indicate that expression of ESRPs is elevated during HNSCC carcinogenesis. In contrast, ESRP1 and ESRP2 were repressed in cancer cells that penetrated through the basement membrane into the stroma and those invading from cancer nests into stromal tissues.

In light of these findings, we asked how altered expression of ESRPs affects cellular phenotypes. ESRPs regulate alternative splicing events in a number of genes associated with reorganization of the actin cytoskeleton and maintenance of cell-cell adhesion, tight junctions, and cell polarity during EMT (29). However, these splicing products are often uncharacterized. In our study, we demonstrated that ESRP1 and ESRP2 suppress cell motility through distinct mechanisms, although involvement of ESRPs in cell motility regulation has been reported previously (10, 27).

First, we found that knockdown of ESRP1 resulted in inclusion of variant exon 3b in alternative splicing of Rac1 mRNA, thereby increasing expression of the Rac1b isoform. In ESRP1 knockdown cells, Rac1b modulated actin dynamics to induce formation of long filopodia and augment cell motility. Variant products of other ESRP1-targeted genes (MENA and Exo70, etc.) are also associated with actin dynamics (16, 17). Taken together, these data indicate that ESRP1 principally inhibits cell motility via regulation of actin dynamics.

Second, we found that knockdown of ESRP2 effectively repressed E-cadherin expression in HNSCC cell lines at both the mRNA and protein levels. Several reports have shown that ectopic expression or knockdown of ESRP1 does not affect E-cadherin expression (8, 10, 22, 30). We previously found that overexpression of ESRP1 and ESRP2 in basal-like breast cancer cells results in up-regulation of E-cadherin expression (9). These observations are consistent with our present finding that ESRP2 contributes to maintenance of E-cadherin expression. Furthermore, we found that among EMT-associated transcription factors that repress E-cadherin expression, the levels of δEF1 and SIP1 were significantly elevated in ESRP2-silenced cells. δEF1 and SIP1 bind directly to the E-cadherin promoter (7, 31, 32). Down-regulation of E-cadherin by δEF1 and SIP1 is associated with acquisition of a migratory/invasive phenotype by epithelial tumor cells (32–35). In addition, δEF1 and SIP1 are also direct suppressors of ESRP expression in cancer cells, indicating that there is a double-negative feedback loop between ESRP2 and δEF1/SIP1. It remains to be elucidated how ESRP2 represses δEF1 and SIP1.

Among the many systems involved in the regulation of EMT, microRNAs (miRNAs) have been specifically implicated in the regulation of EMT-associated transcription factors. In particular, a double-negative feedback loop is formed between miR-34, the miR-200 family, and Snail or δEF1/SIP1 (6, 36). It is thus possible that ESRP2 suppresses δEF1 and SIP1 by regulating miRNA expression. Melamed et al. (37) recently reported that the alternative splicing machinery competes with the miRNA processing machinery when a pre-miRNA sequence overlaps with active splice sites. However, EMT-associated pre-miRNA sequences (miR-34, miR-200 family, and miR-9) do not overlap exon-intron junctions, ruling out such a mechanism. Wu et al. (38) reported that ASF/SF2 (SRSF1) has a splicing-independent function in miRNA processing. Previous reports (8, 10) and our own data (this study) have shown that ESRP2 is a less potent regulator of splicing than ESRP1. Therefore, ESRP2 may influence δEF1 and SIP1 expression by regulating miRNA processing via splicing-independent mechanisms. Elucidation of such a mechanism might lead to identification of a novel mechanism of regulation of cancer progression.

One of the important findings of our study is the plasticity of ESRP expression during cancer cell invasion and metastasis (Fig. 9). ESRP1 and ESRP2 were re-expressed in cancer cells that reached the deep stroma and the lymph nodes, whereas cells invading from cancer nests again lost expression of ESRPs. These in vivo data suggest that down-regulation of ESRP1 and ESRP2 is restricted to cells that acquire a motile phenotype during cancer invasion. As noted above, whether ESRP expression has a positive or negative impact on survival rates of cancer patients remains controversial. Ueda et al. (27) reported that in a cohort of pancreatic cancer patients, the overall survival rate of an ESRP1-high group was significantly higher than that of an ESRP1-low group. In contrast, Yae et al. (14) showed that a high level of ESRP1 expression is significantly associated with a lower rate of overall survival in breast cancer patients, suggesting that increased ESRP1 expression is related to the malignant phenotypes of human breast cancer. In this study, we observed plasticity of ESRP1 and ESRP2 expression during malignant progression. Therefore, it would be difficult to draw a definitive conclusion regarding the relationship between ESRP expression and cancer prognosis. Consistent with this, analysis of information on HNSCC patients in public databases (PrognoScan) revealed no difference in prognosis between ESRP1/ESRP2-positive and ESRP1/ESRP2-negative groups.

FIGURE 9.

Schematic illustration of regulation of cancer cell motility by ESRPs, the expression of which is plastic. Upper (in vivo findings), during carcinogenesis, ESRP1 and ESRP2 expression is increased in dysplasia, advanced carcinoma, and metastatic lesions compared with normal epithelium but is down-regulated in cells that acquire a motile phenotype during cancer invasion. CIS, carcinoma in situ. Lower (in vitro findings), ESRP1 and ESRP2 suppress cell motility through distinct transcriptional and/or post-transcriptional mechanisms.

It is important to elucidate the mechanisms underlying the plasticity of ESRP expression. Previous reports showed that EMT-associated transcription factors (δEF1, SIP1, Snail, Slug, and Twist) down-regulate ESRP1 or ESRP2 (9, 26, 39, 40). In contrast, the factors involved in up-regulation of ESRP1 and ESRP2 are relatively unknown. Recently, the Grainyhead transcription factor Grhl2 was shown to induce ESRP1 expression in breast cancer cells (41). Ras signaling is also implicated in up-regulation of ESRP1 and ESRP2 during carcinogenesis because Ha-Ras-transformed MCF10A cells express higher levels of ESRP1 and ESRP2 compared with parental MCF10A cells.³

It remains to be determined why normal epithelial cells with low expression of ESRP1 have low motility. In this regard, it is noteworthy that Rac1b was first identified as a tumor-associated splicing isoform in breast and colon cancers (42, 43). A previous report demonstrated that ASF/SF2, a member of the serine/arginine-rich protein family, acts as an inducer of the endogenous Rac1b isoform by promoting inclusion of exon 3b (23). In contrast, we found that ESRP1 represses Rac1b expression by causing the variant exon to be skipped. Similar to the pattern for ESRPs, ASF/SF2 expression is also weak in normal epithelium but increases significantly during lung carcinogenesis (44). Therefore, even though ESRP1 is expressed weakly in normal epithelium, Rac1b is not expressed due to low expression of ASF/SF2. In cancer tissues, ESRP1 appears to be functionally competitive with ASF/SF2 in alternative splicing of Rac1 mRNA. Upon down-regulation of ESRP1 in invasive fronts, ASF/SF2 is likely to freely induce Rac1b by promoting inclusion of exon 3b. Thus, ESRP1 has context-dependent functions in the regulation of cell motility. Similarly, it remains to be elucidated why δEF1 and SIP1 are not expressed in normal epithelium even though ESRP2 is not expressed at high levels. Understanding the regulatory networks between ESRPs and EMT regulators may lead to new treatment strategies against cancer progression and metastasis.