EPB41L5 Mediates TGFβ-Induced Metastasis of Gastric Cancer

Gastric cancer is one of the leading causes of cancer-related mortality globally (1). Patients with advanced or metastatic gastric cancer have a poor 5-year survival rate. Although many studies are under way to find novel therapeutic targets in advanced or metastatic gastric cancer, its molecular mechanisms have poorly elucidated because of biological heterogeneity (2, 3). To date, trastuzumab, a human growth factor receptor 2 (HER2)–targeted agent, and ramucirumab, a VEGFR2-targeted agent, are available for target therapy of gastric cancer. However, HER2 is expressed in fewer than 20% to 30% of gastric cancer patients. Ramucirumab is a second-line remedial agent and is not applicable to hemorrhagic gastric cancer (4–6). Thus, investigation of novel therapeutic targets for effective therapy of advanced gastric cancer is required.

Epithelial–mesenchymal transition (EMT) is a primary cause of gastric cancer metastasis and is associated with tumor molecular subtypes with resistance to chemotherapy (7–9). Many studies have reported that the TGFβ pathway regulates EMT. TGFβ ligand binds to TGFβ receptor I through TGFβ receptor II to activate its serine/threonine kinase. The activated TGFβ receptor dimer disseminates signal by the phosphorylation of Smad2/3 and, in turn, the phosphorylated Smad2/3 recruits Smad4, and this protein complex enters the nucleus. Then the complex regulates EMT-related genes such as PAI-1, ZEB1, Snail, and others (10, 11). Some researchers have suggested that TGFβ1 expression is enhanced in the mucosa, serum, and tissue of gastric cancer patients and provokes scirrhous gastric carcinoma. High expression of TGFβ1 reduces the survival rate of gastric cancer patients (12–14).

EPB41L5 belongs to the NBL4 subgroup of the band 4.1 superfamily, which has a conserved FERM domain (F for 4.1 protein, E for ezrin, R for radixin and M for moesin) at the N-terminus and a nonhomologous sequence at the C-terminus (15, 16). EPB41L5 is involved in the development of the embryos of fruit flies, zebrafish, and mice and interacts with the Crumbs complex to regulate epithelial cell polarity (17–19). Expression of EPB41L5 mRNA and protein is increased by TGFβ in mouse mammary gland cell NMuMG (20). Although functional studies of EPB41L5 in cancer are marginal, recent studies showed that lysophosphatidic acid (LPA) activates a small GTPase Arf6 and recruits AMAP1 which interacts with EPB41L5. This mechanism increases renal cancer cell metastasis (21). The same research group has proposed that ZEB1 is engaged in breast cancer metastasis by the Arf6-AMAP1-EPB41L5 mechanism (22, 23). However, the function and clinical relevance of EPB41L5 in advanced gastric cancer have not yet been reported.

Here, we demonstrated that EPB41L5 overexpression is associated with poor prognosis in gastric cancer patients. Among the several gastric cancer cell lines, the EPB41L5 level was increased by TGFβ1 treatment in KATOIII, MKN28, SNU1, and SNU719. EPB41L5 expression was induced by the Smad-dependent TGFβ signaling pathway. By analysis of the EPB41L5 gene promoter, phosphorylated Smad3 binds to the −265 to −256 region on the EPB41L5 promoter, suggesting that the Smad-dependent TGFβ signaling pathway regulates EPB41L5 gene transcription. Upregulation of EPB41L5 expression by TGFβ1 treatment promotes migration and invasion of gastric cancer cells, which are considerably prevented by TGFβ inhibitor, EPB41L5 siRNA, and anti-EPB41L5 mAb. Lung metastasis of gastric cancer cells in nude mice was facilitated by overexpression of EPB41L5; however, knockdown of p120-catenin abolished the effect of overexpression of EPB41L5 on gastric cancer metastasis. Importantly, we found the promoting effect of EPB41L5 on gastric cancer metastasis was fully blocked by anti-EPB41L5 mAb treatment. This research shows that the TGFβ/EBP41L5/p120-catenin mechanism is primarily involved in gastric cancer metastasis, and EPB41L5 is a potential therapeutic target for advanced gastric cancer.

The use of specimens from gastric cancer patients for oligonucleotide microarray and survival rate analysis was approved by the Institutional Review Board, Severance Hospital (IRB No: 4-2016-0013), in compliance with the Helsinki Declaration. All patients provided written-informed consent, and the study was conducted after formal approval by the IRB. Cluster analysis was performed with Cluster and Treeview (ref. 24). For the cluster analysis, the log base 2–transformed data were median centered to the values of each gene expression. To generate list of genes having different expression levels among patients, we conducted serial gene filtration by changing filtering criteria. Unsupervised clustering analysis was performed after serial variance filtration, and then, we selected the clusters by cutting the tree at a suitable level where it provided the best separation of clusters.

All mice experiments were conducted according to approved protocol at the Institutional Animal Care and Use Committee of Yonsei University College of Medicine (IACUC NO: 2016-0341).

Human gastric cancer cell lines AGS, NCI-N87, KATOIII, SK-GT-4, MKN1, MKN28, MKN45, SNU1, SNU5, SNU16, SNU216, SNU484, SNU638, SNU668, and SNU719 were cultured in RPMI-1640 supplemented with 10% FBS and 1% antibiotic/antimycotic solution (Corning) at 37°C under 5% CO₂. 293FT cells were cultured in DMEM. TGFβ1 was purchased from Prospec. LY2157299 was purchased from Selleckchem. Stomach adenocarcinoma TMA slide was purchased from US Biomax.

EPB41L5 relevant DNA constructs encoding full-length and Δ619-624 and p120-catenin DNA were generated by PCR and cloned into the pSG5-KF2M1-Flag and HA (Stratagene California). All plasmid constructs were verified by DNA sequencing.

The mice were immunized with human EPB41L5 antigen 386–637 amino acids. Monoclonal antibody was produced by and purchased from ATGen Co. The antibody was validated by Western blot and immunofluorescence (IF) using EPB41L5 constructs and siRNA. A synthetic blocking peptide against EPB41L5 mAb was synthesized and purchased from GenScript.

The cells were lysed in lysis buffer [20 mmol/L Tris-Cl, 150 mmol/L NaCl, 1% Triton X-100, 1.5% MgCl₂, 1 mmol/L ethylenediaminetetraacetate (EDTA), 1 mmol/L Na₂VO₄, 1 mmol/L phenylmethylsulfonyl fluoride (PMSF), and protease inhibitor cocktail, pH 7.5]. The lysates were briefly vortexed and cleared by centrifugation at 13,000 rpm for 20 minutes at 4°C. The supernatants were collected and transferred to fresh tubes. Protein concentrations were determined by 660 nm protein assay reagent (Thermo Fisher Scientific). The samples were precleared with protein A/G agarose beads (Santa Cruz Biotechnology) for 2 hours and then precipitated with anti-Flag-M2 agarose beads or antibodies with protein A/G agarose beads. Equal amounts of protein extracts and immunoprecipitation products were subjected to electrophoresis on SDS-polyacrylamide gels and then transferred to nitrocellulose transfer membranes (Whatman). The membranes were blocked in Tris-buffer (pH 7.4) containing 0.1% (v/v) Tween 20 (Sigma-Aldrich) and 5% (w/v) nonfat Difco skim milk (BD Biosciences) and probed with primary antibodies. The following antibodies were used: polyclonal EPB41L5 (Thermo Fisher Scientific); generated monoclonal EPB41L5 (ATGen Co.); Flag-tag, β-actin (Sigma-Aldrich); p120-catenin, α-tubulin (Abcam); TβRI, TβRII, Smad2, Smad3, Smad4, p-Smad2, p-Smad3, Slug (Cell Signaling Technology); HA, PAI-1 (Santa Cruz Biotechnology); E-cadherin (Cell Signaling Technology and Abcam). The signals were developed by substrate (Thermo Fisher Scientific) according to the manufacturer's protocol.

The cells were cultured on chamber slides (SPL Life Sciences) and fixed in 4% paraformaldehyde for 30 minutes at room temperature. After washing with PBS, the fixed cells were incubated for 1 hour with 3% BSA to block nonspecific antibodies. For IF analysis using human stomach cancer tissue slides, slides immersed in xylene and ethanol for rehydration. Citrate buffer (Dako) was used for antigen retrieval. Anti-EPB41L5, E-cadherin, p120-catenin, and Flag-tag were incubated at 4°C overnight and then stained with Alexa Fluor 488– or Alexa Fluor 549–conjugated goat anti-rabbit or anti-mouse secondary antibody (Thermo Fisher Scientific). The nuclei were stained with Hoechst 33258. The mounted samples were imaged with an LSM710 confocal microscope (Carl Zeiss).

In situ proximity ligation assay PLA analysis (Sigma-Aldrich) was carried out following the manufacturer's procedure. KATOIII and SNU719 cells fixed in 4% paraformaldehyde were washed with PBS and blocked with blocking solution. After the application of primary rabbit (p120-catenin) and mouse (EPB41L5 and E-cadherin) antibodies, the cells were incubated with PLUS and MINUS secondary PLA probes based on species and then subjected to ligation and amplification using the provided reagents. The samples were mounted with Duolink mounting medium and analyzed using an LSM 710 Laser Scanning Microscope (Carl Zeiss).

Migration was measured by Transwell with 8.0-μm pore polycarbonate membrane insert (Corning). For the invasion assay, inserts of the Transwell were coated with Matrigel (BD Biosciences). Note that 1 × 10⁵ cells per well were added to the upper chamber, and the lower chamber was filled with 600 μL of serum-free medium with or without TGFβ1, LY2157299 (TGFβ inhibitor), or anti-EPB41L5 mAb as a chemoattractant. After 24 hours of incubation, nonmigrating and noninvading cells were carefully removed from the upper chamber with a cotton swab. Migrating and invading cells were stained with 0.2% crystal violet (Sigma-Aldrich) in 20% methanol and counted at ×200 magnification under a microscope. The migration and invasion assays were performed in triplicate.

To make human EPB41L5 gene promoter constructs, a specific region (−1051 to +1109 nucleotides) of the EPB41L5 gene was prepared by PCR amplification of human genomic DNA obtained from KATOIII cells. The amplified product was purified, digested with XhoI and HindIII restriction enzymes, and cloned into pGL4.21[luc2P/Puro] vector (Promega). This EPB41L5 reporter gene was used as a template for all subsequent deletion constructs. To perform the luciferase reporter assay, the cells were transiently transfected with 1 μg of EPB41L5 reporter gene and 100 ng of β-galactosidase gene per well in 6-well plates. After 24 hours of incubation, the cells were treated with 10 ng/mL of TGFβ1 in serum-free medium for another 24 hours. The cells were harvested and lysed to assay luciferase activity using a luciferase assay system (Promega). Luminescence was measured by a MicroLumatPlus LB96V Microplate Luminometer (Berthold). β-Galactosidase activity was measured with colorimetric substrate, ortho-nitrophenyl-β-galactoside (ONPG), at 420 nm of absorbance. Relative luciferase reporter gene activity was calculated as luciferase activity/β-galactosidase activity (fold).

Chromatin immunoprecipitation (ChIP) assay was carried out on 6 × 10⁶ KATOIII cells using a Pierce Agarose ChIP kit (Thermo Fisher Scientific) according to the manufacturer's instructions. The nuclear lysates were precipitated with p-Smad3 antibody (Cell Signaling Technology) or normal rabbit IgG (Thermo Fisher Scientific) overnight at 4°C together with ChIP grade protein A/G agarose beads. The beads were washed, eluted, and treated with proteinase K. The recovered DNA was amplified by qPCR or standard PCR. The primers (−386 to −218) used for PCR were forward (5′-GGGAGGAAATACGTGACAGG-3′) and reverse (5′-CTGCCGAAACCCAGTTCC-3′).

Five-week-old female athymic BALB/c nu/nu mice were obtained from Orient. Control vector or pCDH-EPB41L5–, pLKO.1-EPB41L5–, pLKO.1-p120-catenin–, and pLECE3-GFP–expressing KATOIII cells and TGFβ1-treated KATOIII cells (2 × 10⁷ cells in 200 μL PBS) were injected into the lateral tail vein. Anti-EPB41L5 mAb was administered at 5 mg/kg every other day for 2 weeks. The fluorescence images were taken and analyzed with an IVIS imaging system (Caliper Life Sciences).

The overall survival data obtained from gastric cancer patients were analyzed by the Kaplan–Meier Plotter (http://kmplot.com/analysis; ref. 25). Used Affymetrix ID is 220977_x_at. The log-rank test was used for Kaplan–Meier survival plots. Statistical analysis was performed using the Student t test to compare two groups of independent experiments. The data are presented as mean ± SD. P values < 0.05 were considered to indicate statistical significance.

To identify potential driver pathways or target molecules in gastric cancer, oligonucleotide microarray analysis was performed in 78 gastric cancer patients. The patients were divided into two groups according to their gene expression patterns. Patients in the group with highly expressed candidate genes, including APEG1, SMPX, GPR177, and EPB41L5, had a higher rate of recurrence or death than patients with lower rates of gene expression (Supplementary Fig. S1A). Among the highly expressed genes, we selected a cell adhesion protein EPB41L5 in this study because of its availability as a druggable target for cancer therapy. Patients with high levels of EPB41L5 expression had a worse survival rate than patients with low levels (Fig. 1A). According to Kaplan–Meier plots analysis in The Cancer Genome Atlas (TCGA) database, gastric cancer patients with high levels of EPB41L5 expression had a lower survival rate (HR, 1.73; P, 2.8E−10; Fig. 1B). These clinical outcomes show that high levels of EPB41L5 expression are related to poor prognosis in gastric cancer patients, thereby warranting further investigation.

Expression of EPB41L5 is increased by TGFβ treatment in epithelial NMuMG cells (20). We first examined which gastric cancer cells were responsive to TGFβ signaling. The protein levels of TGFβ signaling cascades were analyzed in several gastric cancer cell lines by Western blot. Despite the varying expression of EPB41L5 in these cell line panels, EPB41L5 was expressed in most of gastric cancer cells (Fig. 1C). These protein expression patterns were similar to the mRNA level (Supplementary Fig. S2A and S2B). As reported, Smad4 was not detected in NCI-N87, MKN45, and SNU216 cells (26–28). Next, TGFβ1 was administered to gastric cancer cells to examine whether EPB41L5 gene expression is regulated by TGFβ signaling. EPB41L5 gene expression was significantly increased by TGFβ1 treatment in KATOIII, MKN28, SNU1, and SNU719 cells, but not in Smad4-defective gastric cancer cells. Smad4 was normally expressed, but EPB41L5 level was not affected by TGFβ1 treatment in gastric cancer cell lines including AGS, SK-GT-4, MKN1, SNU5, SNU16, SNU484, SNU638, and SNU668 (Fig. 1D). Because of mutations of TGFβ signaling cascades, such as TβRII, Smad3, and Smad4, in these gastric cancer cells, EPB41L5 expression was not likely affected by TGFβ. In fact, it has been reported that there are mutations of TGFβ pathway machinery in some gastric cancer cells, including SNU5 and SNU638 (29, 30). Both the expression of mesenchymal genes, such as PAI-1 and Slug, and the level of phosphorylated Smad3 were significantly increased by TGFβ1 treatment, according to RT-qPCR and Western blot analysis, respectively (Supplementary Fig. S3A; Fig. 1E). IF staining results showed that EPB41L5 expression was increased by TGFβ1 in the membrane region of KATOIII cells (Supplementary Fig. S3B). Interestingly, the morphologic characteristics of gastric cancer cells KATOIII and SNU719, but not of NCI-N87 cells, were changed to mesenchymal characteristics by TGFβ1 treatment (Fig. 1F). These results indicate that TGFβ signaling positively regulates the expression of EPB41L5 and EMT in gastric cancer cells.

To determine whether TGFβ-induced EPB41L5 expression is mediated by Smad-dependent or non–Smad-dependent signaling, we first assessed the mRNA levels of EPB41L5, PAI-1 and Slug in MKN45 and NCI-N87 gastric cancer cells, which are deficient in Smad4. The levels of these mRNAs, as well as the level of EPB41L5 protein, were not changed by TGFβ1 treatment, according to Western blot analysis and IF staining (Supplementary Fig. S3A and S3B; Fig. 1E). In contrast, knockdown of Smad4 in TGFβ signaling–positive KATOIII gastric cancer cells abolished the effect of TGFβ on the induction of EPB41L5 (Fig. 2A and B). Knockdown of Smad4 alone as a control did not affect the level of EPB41L5, PAI-1 and Slug (Supplementary Fig. S4A and S4B). These results suggest that the Smad4 is required for TGFβ-induced regulation of EPB41L5 expression in gastric cancer cells.

We next examined whether Smad directly regulates the transcription of EPB41L5 in response to TGFβ1 treatment. The Smad-binding motifs CAGAC or GTCTG on the EPB41L5 gene promoter were searched using the PROMO 3.0 tool (31, 32). The schematic diagram illustrated multiple Smad-binding elements on EPB41L5 promoter (Fig. 2C). Next, luciferase reporter gene assays were conducted with approximately 2-, 1.5-, and 1-kb sized promoters of the EPB41L5 gene. Promoter activity was significantly decreased with the approximately 1-kb (−51 to 1109 bp) promoter of the EPB41L5 gene compared with the approximately 1.5 kb (−551 to 1109 bp) and 2 kb (−1051 to 1109 bp) promoters (Fig. 2D). The EPB41L5 gene promoter with deleted #4 (−265 to −256) Smad-binding element caused significant reduction of luciferase activity compared with control. Luciferase activity of this #4 deleted mutant promoter was not affected by TGFβ1 treatment (Fig. 2E). These results show that the #4 (−265 to −256) Smad-binding element is critical for transcriptional regulation of the EPB41L5 gene. To verify whether Smad binds to the #4 Smad-binding element on the EPB41L5 promoter and regulates EPB41L5 transcription, ChIP assay was carried out with phospho-Smad3 antibody by using the #4 site-specific PCR primers (−386/−218) in KATO3III cells. Phosphorylated Smad3 was significantly recruited by TGFβ1 and dissociated by a potent TGFβ receptor I inhibitor, LY2157299, on the #4 position of the EPB41L5 gene promoter (Fig. 2F). Thus, these data demonstrated that phosphorylated Smad3 binds to and activates EPB41L5 gene transcription in response to TGFβ signaling.

Using LY2157299, we next tested the effect of TGFβ inhibition on the expression of EPB41L5, mesenchymal genes such as PAI-1 and Slug, and mesenchymal transition of gastric cancer cells. These gene and protein expression levels of EPB41L5 and mesenchymal genes were decreased by LY2157299 in a dose-dependent manner (Fig. 3A; Supplementary Fig. S5A). IF staining results showed that EPB41L5 expression was increased by TGFβ1 and inhibited by LY2157299 in KATOIII cells (Supplementary Fig. S5B). To analyze cell mobility, migration and invasion assays were conducted utilizing Transwell chambers coated or not coated with Matrigel. TGFβ1 treatment enhanced migration and invasion of KATOIII and SNU719 cells; the enhanced migration was fully inhibited by LY2157299 (Fig. 3B). Mesenchymal morphologic changes of KATOIII and SNU719 cells were also blocked by LY2157299 treatment (Fig. 3C).

To determine whether EPB41L5 affects the migration of gastric cancer cells in response to TGFβ signaling, experiments were performed using EPB41L5 siRNAs in KATOIII or SNU719 cells. In the absence of TGFβ1, knockdown of EPB41L5 decreased the protein and mRNA level of EMT-related genes, PAI-1 and Slug, in KATOIII cells (Supplementary Fig. S6A and S6B). EPB41L5-knocked down KATOIII cells significantly decreased the TGFβ1-induced gene and protein levels of EPB41L5, PAI-1, and Slug (Fig. 3D and E). EPB41L5 deficiency abrogated the TGFβ1-mediated morphologic changes of KATOIII cells (Supplementary Fig. S5C). Moreover, the knockdown of EPB41L5 abolished the TGFβ1-enhanced migration and invasion of KATOIII and SNU719 cells (Fig. 3F). Notably, knockdown of EPB41L5 alone diminished the migration of KATOIII cells, although to a lesser extent (Supplementary Fig. S6C). To further corroborate our findings in KATOIII and SNU719 cells, we performed the same experiments in MKN28 cells which express relatively low level of EPB41L5. The results were similar to those from KATOIII and SNU719 cells (Fig. 3F; Supplementary Fig. S5D). Collectively, these results demonstrate that EPB41L5 alone as well as TGFβ-induced EPB41L5 is required for mesenchymal transition and invasiveness of gastric cancer cells.

We demonstrated that TGFβ1-enhanced EPB41L5 expression is required for migration and invasion of gastric cancer cells in vitro. Next, we assessed the effect of overexpression of EPB41L5 on migration of gastric cancer cells by Transwell migration assay. Overexpression of EPB41L5 dose-dependently enhanced migration and invasion of KATOIII cells (Fig. 4A). It has been reported that the cell adhesion complex modulates epithelial cell motility and is crucial for cancer metastasis (33, 34). EPB41L5 was shown to interact with cell adhesion molecules, including p120-catenin, paxillin, and MPP5 (18, 20). Because the molecular mechanism of EPB41L5-mediated metastasis of gastric cancer cells is not yet known, we next examined which molecule is required for EPB41L5-mediated migration and invasion of gastric cancer cells. Strikingly, knockdown of p120-catenin abolished the effect of overexpression of EPB41L5 on migration of gastric cancer cells (Fig. 4B). Knockdown of p120-catenin alone had no effect on the migration and invasion (Supplementary Fig. S7A). It is noteworthy that the other known EPB41L5-interacting proteins, including paxillin and MPP5, had negligible enhancing effects on the activity of EPB41L5 on migration of gastric cancer cells (Fig. 4C), suggesting the important role of p120-catenin in EPB41L5-mediated invasiveness of gastric cancer cells.

Because the dissociation of p120-catenin from E-cadherin is required for invasiveness of cancer cells (35, 36), we next examined the changes in interaction among EPB41L5, p120-catenin, and E-cadherin in response to TGFβ1. Overexpression of EPB41L5 efficiently reduced the levels of E-cadherin, as is commonly observed in metastasis of gastric cancer cells (Fig. 4D; ref. 7). In both KATOIII and SNU719 cells, IF and in situ PLA analysis showed that colocalization between EPB41L5 and p120-catenin was robustly induced by TGFβ1, but colocalization between E-cadherin and p120-catenin was not (Fig. 4E and F). Knockdown of EPB41L5 prevented dissociation of the E-cadherin and p120-catenin complex induced by TGFβ1 (Fig. 4G). This result was again verified by an endogenous coimmunoprecipitation experiment (Supplementary Fig. S7B). Taken together, these results show that the EPB41L5–p120-catenin complex is required for TGFβ-induced invasive activity of gastric cancer cells.

Understanding the crucial interactions of cell adhesion molecules in metastasis leads to the development of novel therapeutics using mAbs and peptides (37, 38). Therefore, to examine whether the EPB41L5 protein may be used as a therapeutic target for gastric cancer, we next generated an mAb against EPB41L5. Western blot and IF analyses were performed in gastric cancer cells transfected with Flag-EPB41L5 or EPB41L5 siRNA to substantiate antibody specificity. Overexpressed EPB41L5 was detected (Supplementary Fig. S8A), and silencing of endogenous EPB41L5 was confirmed with anti-EPB41L5 mAb (Supplementary Fig. S8B). IF staining showed that endogenous EPB41L5 apparently was detected by anti-EPB41L5 mAb in KATOIII cells (Supplementary Fig. S8C). By validation of antibody with constructs of full-length and Δ619-624 of EPB41L5, we confirmed that engineered anti-EPB41L5 mAb specifically recognizes amino acids 619–624 of the C-terminus of EPB41L5 (Supplementary Fig. S8D). To verify whether generated EPB41L5 mAb recognizes the EPB41L5 protein in cell surface, we next performed nonpermeabilized IF analysis after the transfection of EPB41L5 siRNA and full-length or Δ619-624 of EPB41L5 (siEPB41L5-resistant). EPB41L5 mAb clearly recognizes full-length EPB41L5 protein in the cell surface but not Δ619-624 of EPB41L5 protein (Supplementary Fig. S8E). Consistently, EPB41L5 was colocalized with p120-catenin (Supplementary Fig. S8C). Moreover, we observed a high expression of membranous EPB41L5 colocalized with a well-known cell surface protein, E-cadherin in stomach cancers compared with normal stomach tissues (Supplementary Fig. S9). These results demonstrate that generated EPB41L5 mAb recognizes EPB41L5 protein in cell surface.

Given the validation of the specificity of anti-EPB41L5 mAb, we next assessed the effect of anti-EPB41L5 mAb on the migration and invasion of gastric cancer cells with or without TGFβ1 treatment. Transwell migration and invasion analysis showed that TGFβ1 efficiently increased the migration and invasion of KATOIII cells; this effect was dramatically inhibited by anti-EPB41L5 mAb treatment (Fig. 5A). The effect of anti-EPB41L5 mAb treatment was reversed by a blocking peptide against EPB41L5 (Fig. 5B). In addition, the treatment of EPB41L5 mAb inhibited the migration and invasion of KATOIII cells even in the absence of TGFβ1 (Supplementary Fig. S8F). The reduction of E-cadherin by TGFβ1 was abolished by anti-EPB41L5 mAb (Fig. 5C). Exogenous coimmunoprecipitation analysis showed that anti-EPB41L5 mAb treatment leads to dissociation of the EPB41L5–p120-catenin complex, and this effect is prevented by a synthetic blocking peptide (Fig. 5D). These results demonstrated that anti-EPB41L5 mAb inhibits the migration and invasion of KATOIII cells by dissociation of the EPB41L5–p120-catenin adhesion complex.

To examine the in vivo roles of EPB41L5 during metastasis of gastric cancer cells, EPB41L5-overexpressing KATOIII cells were intravenously injected into nude mice. After 4 weeks of injection, we found that lung metastases of gastric cancer cells were significantly increased in the EPB41L5-overexpressing KATOIII cells compared with controls (Fig. 6A; Supplementary Fig. S10A). Lung nodules were significantly increased in EPB41L5-overexpressing KATOIII cells compared with the control group after 16 weeks of injection (Supplementary Fig. S10D). TGFβ1 treatment increased the metastasis of KATOIII cells; however, knockdown of EPB41L5 diminished the enhanced metastasis of KATOIII cells, again verifying the crucial role of EPB41L5 in TGFβ-mediated metastasis of gastric cancer cells (Fig. 6B; Supplementary Fig. S10B). As confirmed by the above in vitro result, knockdown of p120-catenin significantly diminished the enhancing effect of EPB41L5 on lung metastasis of gastric cancer cells (Fig. 6C; Supplementary Fig. S10C). These results suggested that the EPB41L5–p120-catenin complex is required for metastasis of gastric cancer cells.

Finally, to verify the in vivo efficacy of anti-EPB41L5 mAb, the antibody (5 mg/kg) was administered every other day for 2 weeks after injection of vehicle or EPB41L5-overexpressing KATOIII cells into nude mice (Fig. 6D). Increased lung metastasis of gastric cancer cells by overexpression of EPB41L5 was efficiently blocked by anti-EPB41L5 mAb treatment (Fig. 6E). Furthermore, anti-EPB41L5 mAb treatment abolished TGFβ1-induced metastasis of KATOIII cells (Fig. 6F). Collectively, these results show that anti-EPB41L5 mAb may be a promising therapeutic for metastatic gastric cancer.

Cell adhesion proteins play a critical role in cancer metastasis, and treatments using antibodies to target cell adhesion proteins are being tested in clinical trials (37, 39, 40). Unraveling novel adhesion molecules or signaling pathways related to metastasis could provide promising targets for novel cancer therapies. In EMT, epithelial cells lose cell–cell junctions and cell polarity, and the actin cytoskeleton is reorganized to enable the mesenchymal phenotype (11, 41). The mesenchymal-specific protein EPB41L5 is reported to interact with p120-catenin, which destabilizes E-cadherin, and paxillin, a focal adhesion kinase (20). Moreover, EPB41L5 binds to MPP5, a Crumbs complex component that negatively regulates cell polarity (18). Another junction protein, β-catenin, colocalizes with EPB41L5 in the basolateral membrane of kidney epithelial cells, but its binding has not been confirmed (18, 19). A focal adhesion protein, focal adhesion kinase, has an FERM domain in its N-terminus and binds with ASAP1/AMAP1 and paxillin, another binding partner of EPB41L5 in the C-terminus (21, 42, 43). We have provided evidence that EPB41L5 promotes gastric cancer metastasis by dissociation of p120-catenin from E-cadherin. Furthermore, knockdown of p120-catenin abrogated the promoting activity of EPB41L5 on in vitro invasiveness and in vivo metastasis of gastric cancer cells. Knockdown of either MPP5 or paxillin had no apparent effect on the EPB41L5-enhaced migration of gastric cancer cells when compared with p120-catenin. Our findings, for the first time, report that the EPB41L5–p120-catenin adhesion complex mediates gastric cancer metastasis.

We have also demonstrated the clinical relevance of EPB41L5 in gastric cancer tumorigenesis by showing that EPB41L5 is responsible for the poor prognosis of gastric cancer patients. In this regard, a recent study demonstrated that the mesenchymal-characteristic EPB41L5 is highly expressed and is associated with a poor survival outcome in patients with breast cancer, renal cancer, and tongue squamous cell carcinoma (21, 23, 44). It has also been reported that TGFβ promotes gastric cancer metastasis and that its expression is positively correlated with lymph node metastasis and poor prognosis in gastric cancer (13, 14). A great number of Smad-binding motifs have been identified on the EPB41L5 gene promoter. Using the luciferase reporter gene assay and the ChIP assay, we showed that phosphorylated Smad3 directly binds to the promoter region of EPB41L5, leading to activation of transcription of the EPB41L5 gene. In addition, we found multiple transcription factor binding elements on EPB41L5 gene promoter, including p300, C/EBP, c-Jun, c-Fos, and MZF1, as well as Smad3/4. Therefore, we could not exclude the possibility that other signaling cascades are involved in EPB41L5 transcriptional activity. There has been a report that ZEB1 is involved in the expression of EPB41L5 (22, 23). However, we observed an increase of ZEB1 expression as a mesenchymal marker by TGFβ1, but we could not confirm the correlation between ZEB1 and EPB41L5 in gastric cancer cells. Knockdown of EPB41L5 abrogates the TGFβ1-induced invasiveness of gastric cancer cells as well as expression of mesenchymal genes, such as PAI-1 and Slug. At this stage, we do not know whether EPB41L5 is directly involved in the TGFβ1-mediated transcriptional regulation of EMT target genes. However, we cannot exclude the possibility that EPB41L5 participates in a positive feedback loop of TGFβ1 signal. Further study is needed to prove this possibility.

Humanized antibodies have had a great success as anticancer drugs for more than 15 years. Therapeutic antibodies inhibit or activate the function of antigens to inhibit the growth of cancer cells, or act through cytotoxicity by the immune system through the mechanism of antibody-dependent cell-mediated cytotoxicity (45, 46). Studies are under way to develop antibodies that inhibit cell adhesion molecules, such as N-cadherin, P-cadherin, and ICAM-1 (47–49). Given our findings that EPB41L5 localizes in cell adhesion molecules and promotes in vitro and in vivo metastasis of gastric cancer cells by forming an adhesion complex with p120-catenin, we successfully generated a monoclonal EPB41L5 antibody and tested its availability as mAb-based therapy. The EPB41L5 mAb efficiently blocked TGFβ1-induced migration and invasion of gastric cancer cells. Moreover, EPB41L5-promoted lung metastasis was significantly suppressed by anti-EPB41L5 mAb treatment. The antimetastatic action of EPB41L5 mAb was validated by confirming that EPB41L5 mAb inhibits the formation of the EPB41L5–p120-catenin adhesion complex. In fact, the in vitro binding analysis revealed that the interaction site between EPB41L5 and p120-catenin was different from the EPB41L5 antibody recognition site in EPB41L5 protein. Nevertheless, we observed a significant decrease in the interaction between EPB41L5 and p120-catenin when treated with EPB41L5 antibody, and confirmed that the in vivo gastric cancer metastasis was completely blocked. Presumably, the binding of EPB41L5 antibody to EPB41L5 induces the destabilization of EPB41L5/p120-catenin complex. Because the formation of protein complexes in vivo occurs through three-dimensional way, in order to prove this, additional studies using protein structure are needed.

Collectively, this study demonstrated that the TGFβ/EPB41L5/p120-catenin signaling cascade promotes metastasis of gastric cancer cells. In addition, high expression of EPB41L5 was highly correlated with poor overall survival of gastric cancer patients, and targeted inhibition of EPB41L5 efficiently inhibited gastric cancer metastasis. Our findings provide the first rationale that EPB41L5 is a promising therapeutic target for advanced gastric cancer.

No potential conflicts of interest were disclosed.

Conception and design: M.-H. Jeong, S.-Y. Park, S.-H. Lee, S. Lee, K.-C. Choi, J.-H. Cheong, H.-G. Yoon

Development of methodology: J.-Y. Yoo, M.J. Kim, S. Jang, K.-C. Choi, H.-G. Yoon

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.-H. Jeong, J.E. Lee, S.-J. Shin, K.-C. Choi, J.-H. Cheong, H.-G. Yoon

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.-C. Choi, J.-H. Cheong, H.-G. Yoon

Writing, review, and/or revision of the manuscript: M.-H. Jeong, K.-C. Choi, J.-H. Cheong, H.-G. Yoon

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Seo, J.-Y. Yoo, S.-H. Park, M.J. Kim, S. Jang, H.-K. Choi, K.-C. Choi, J.-H. Cheong, H.-G. Yoon

Study supervision: K.-C. Choi, J.-H. Cheong, H.-G. Yoon

This work was supported by the National Research Foundation of Korea (NRF) MRC grant funded by the Korean government (MSIT; No. NRF-2018R1A5A2025079 and NRF-2017R1E1A1A01072732 to H.-G. Yoon, and NRF-2018R1A5A2025079 and NRF-2016R1A2B2016286 to J.-H. Cheong) and the National Research Foundation of Korea (NRF) grant funded by the Ministry of Education, Science and Technology (No. NRF-2017R1A2B4007971, K.-C. Choi).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.