Cooperation between integrin α₅ and tetraspan TM4SF5 regulates VEGF-mediated angiogenic activity

All viable tumors must undergo angiogenesis from preexisting vasculature to acquire oxygen and nutrients for continued growth and metastatic dissemination.¹  Angiogenesis requires the production and release of soluble factors, including vascular endothelial growth factor (VEGF), which interacts with cognate receptors resulting in proliferation and invasion of endothelial cells.^2-5  Induction of VEGF can be facilitated either by hypoxia⁶  or oncoprotein activation.^7,8  The identification and function of pro- or antiangiogenic factors enabling tumor cells to modulate endothelial cell responses has been intensively studied. However, little is known about how signals of cancer cells regulate tumor angiogenesis.⁹ 

Integrins are cell-adhesion receptors that activate diverse intracellular signaling molecules to reorganize actin filaments^10,11  and to regulate cell proliferation and gene regulation.^12,13  When integrins interact with extracellular matrix at focal adhesions, the cytoplasmic tails of integrin subunits recruit focal adhesion molecules, including focal adhesion kinase (FAK), paxillin, p130Cas, and c-Src.¹⁴  Because integrins are linked to actin filaments through protein complexes at their cytoplasmic tails,¹¹  the assembly and activation of focal adhesion molecules result in an efficient response to the extracellular environments for regulation of cell functions, including gene expression and protein secretion. Communications between cancer cells and neighboring endothelial cells occur through angiogenic factor expression/secretion and receptor liganding.¹⁵  Integrins transduce signaling by interaction with growth factor receptors^16,17  or TM4SF.¹⁸ 

TM4SF proteins (also known as tetraspanins or tetraspans) are a group of hydrophobic proteins with 4 transmembrane domains, 2 extracellular loops, and 2 short cytoplasmic tails.¹⁹  TM4SFs form complexes with integrins to create massive tetraspanin-web or tetraspanin-enriched microdomain (TERM) structures to collaboratively perform roles in cell adhesion, proliferation, and motility.^18,20,21  TM4SF5 (also known as L6H) is a homolog of tumor-associated antigen L6 (TM4SF1 or L6-Ag) and forms a 4-transmembrane L6 superfamily with L6, IL-TMP, and L6D.²²  TM4SF5 is highly expressed in diverse tumor types, including hepatocarcinoma.^23,24  We observed that TM4SF5 causes actin reorganization through regulation of FAK and RhoA activity, epithelial-mesenchymal transition, and loss of contact inhibition leading to multilayer growth.²⁴  TM4SF5 ectopically expressed in fibroblasts interacts with integrin α₂ to regulate focal adhesion turnover and actin reorganization.²⁵  Therefore, it is probable that TM4SF5 cooperates with the integrins to regulate cellular functions.

In this study, TM4SF5 overexpression in human hepatocarcinoma patients occurred together with VEGF expression and aggressive vascularization. Ectopic expression of TM4SF5 in hepatocytes resulted in higher surface retention of integrin α₅ and subsequent activation of FAK, c-Src, and signal transducer and activator of transcription 3 (STAT3), leading to VEGF expression and secretion, and enhanced angiogenic activity of human umbilical vein endothelial cells (HUVECs). Therefore, TM4SF5 and integrin α₅ enhanced VEGF expression via the c-Src-FAK/STAT3–signaling linkage, indicating that TM4SF5 is proangiogenic.

Stable clones (control SNU449Cp and TM4SF5-expressing SNU449Tp or SNU449T16 cell clones) were described previously.²⁴  HUVECs were maintained until the sixth passage in M199 media containing 20% fetal bovine serum (FBS; JBI, Daegu, Korea), 5 U/mL heparin (Sigma-Aldrich, St Louis, MO), and 5 ng/mL basic fibroblast growth factor on 0.3% gelatin-precoated culture dishes.

Cells were transfected for 2 days with the indicated expression vectors: mock construct, pcDNA3-TM4SF5, myc-(His)₆–TM4SF5, pcDM–wild-type (WT) integrin α₅, pECE-tailless integrin α₅²⁶  (with deletion of the cytoplasmic 27 amino acids), pRc/CMV-(HA)₃-WT FAK, pBS-FRNK, pKH3-WT c-Src, pKH3-inactive Y416F c-Src mutant,²⁷  pRc/CMV-WT STAT3, or pRc/CMV–dominant-negative Y705F STAT3.²⁸  SNU449Tp or TM4SF5-positive Huh7²⁹  cells were microporated with control short, hairpin RNA (shRNA) against GFP or TM4SF5 (shTM4SF5),²⁵  small, interfering RNA (siRNA) against c-Src (QIAGEN, Valencia, CA, targeting against cgg ctt gtg ggt gat gtt tga) or STAT3 (Dharmacon RNA Technologies, Lafayette, CO, targeting against gag att gac cag cag tat att caa cat gtc att tgc tga att cca aca atc cca aga atg ttt and caa cag att gcc tgc att gtt).

Stable cells were treated with pharmacologic inhibitors, including 20 μM U0126, 40 μM PD98056 (LC Laboratories, Woburn, MA), 40 μM PP2, 40 μM PP3 (A.G. Scientific, San Diego, CA), 30 μM AG490, or 15 μM Y27632 (Calbiochem, San Diego, CA) for 24 hours before lysate preparation. Whole-cell lysates were prepared as previously described.²⁵  SNU449Cp and SNU449Tp cells were suspended in 10 mM ethylenediaminetetraacetic acid and washed with phosphate-buffered saline (PBS), then incubated with normal human IgG or function blocking anti–integrin α₅ (P1D6) antibody (30 μg/mL; Chemicon International, Temecula, CA). Cells were reseeded into tissue-culture dishes, and incubated at 37°C and 5% CO₂ for 10 hours before collecting conditioned media and lysate preparation. Control or tumor tissues were frozen immediately after surgery using liquid N₂, homogenized in liquid N₂ and extracted using radioimmunoprecipitation assay (RIPA) buffer containing 0.1% sodium dodecyl sulfate, before centrifugation at 13 000g for 30 minutes at 4°C. Standard Western blots were performed using antibodies against phospho-Y³⁹⁷FAK, phospho-Y⁵⁷⁷FAK, phospho-Y⁸⁶¹FAK, and phospho-Y⁹²⁵FAK (Invitrogen, Carlsbad, CA), integrin α_v, α₂, α₃, α₄, α₅, β₁, β₃, β₅ (Chemicon International), phospho-Y⁷⁰⁵STAT3, STAT3, HA (AbGent), VEGF-A (sc-152 or sc-7269), phospho-Y⁴¹⁶Src, phospho-Y⁵²⁷Src, c-Src (Santa Cruz Biotechnology, Santa Cruz, CA), FAK (BD Biosciences, San Jose, CA), α-tubulin (Sigma-Aldrich), phospho-Erk1/2, Erk1/2 (Cell Signaling Technology, Danvers, MA), or TM4SF5.²⁴ 

SNU449 cells were transiently microporated for 48 hours with pGL3-human VEGF promoter, pBabe–β-galactosidase, and the indicated expression vectors (human integrin α₅ WT, tailless integrin α₅ mutant, FAK WT, FRNK, c-Src WT, inactive Y416F c-Src mutant, STAT3 WT, or dominant negative [DN; Y705F] STAT3). Luciferase activity was analyzed and normalized using β-galactosidase activity for transfection efficiency.²⁸ 

Subconfluent SNU449Cp and SNU449Tp cells were serum-deprived for 24 hours. Conditioned media were collected from each cell culture. For functional-blocking experiments, conditioned media were collected from SNU449Tp cells that had been suspended and incubated for 20 minutes with either normal mouse IgG or antihuman integrin α₅ (P1D6) monoclonal antibody (30 μg/mL) before reseeding onto normal culture dishes for 10 hours. Angiogenic factors in the conditioned media were determined using RayBiotech Human Angiogenesis Antibody Array (Norcross, GA), following the manufacturer's protocol.

SNU449 cells were microporated with pcDNA3 or pcDNA3-myc-(His)₆-TM4SF5 plasmid. After 1 day, cells were suspended in 10 mM ethylenediaminetetraacetic acid, centrifuged, and resuspended in PBS. The same number of cells was incubated with various concentration of trypsin (1-1000 units/mL PBS) at room temperature for 15 minutes. Digestion was stopped by addition of 1 mM phenylmethylsulfonyl fluoride and cells were washed twice with cold PBS and lysed with RIPA buffer, before immunoblots using either anti–α₅ecto (BD Bioscience) or -α₅cyto (Chemicon International) antibody, recognizing an extracellular or the cytoplasmic region of integrin α₅, respectively. Alternatively, the transfected cells were collected and an equal number of cells in 2 sets were analyzed for integrin α₅ on cell surface by flow cytometry using monoclonal anti–integrin α₅ antibody (clone P1D6; Chemicon International).²⁶  One set was nonpermeabilized, whereas the other set was permeabilized with 0.5% Triton X-100 in PBS before fixation and antibody incubation. Data were analyzed using WinMDI software (The Scripps Research Institute, San Diego, CA).

Subconfluent TM4SF5-null SNU449 or SNU398 cells were cotransfected for 48 hours with mock, mock plus myc-(His)₆-TM4SF5, or myc-(His)₆-TM4SF5 plus either WT integrin α₅, tailless α₅ (α_5/1), or chimeric α₅ tail where the α₅ cytoplasmic tail replaced the tail of integrin α₄ (X4C5).³⁰  Whole-cell lysates were immunoprecipitated with anti-(His)₆ antibody, before immunoblots using anti–integrin α₅ecto (BD Bioscience) or α₅cyto (Chemicon International) antibodies.

Media for SNU449 cell lines (RPMI 1640 media containing 10% FBS) were changed to M199 media containing 5% FBS for a 48-hour incubation. Media supernatant was cleared with a 0.45-μm syringe filter. Before the tube formation assay, HUVECs were serum-starved by incubating in M199 media containing 1% FBS for 8 hours. Matrigel (BD Biosciences) was polymerized (200 μL/well of a 24-well culture plate) for 30 minutes at 37°C. Serum-starved HUVECs were resuspended in 1.2 mL of the conditioned media and reseeded in Matrigel-coated wells (10⁵ cells/well). After incubation without or with anti-VEGF antibody (40 μg/mL) for the indicated periods, images of capillary-like networks were recorded randomly using a phase-contrast microscope (CKX41; Olympus, Tokyo, Japan).

Aorta segments were prepared from a 6-week-old male BALB/c mouse (Orient, Seungnam, Korea). Aorta ring pieces were embedded between 2 layers of Matrigel in 24-well culture plates, in the presence of the conditioned media. Normal IgG or anti-VEGF antibody (40 μg/mL) was mixed with the conditioned media, 20 minutes before application. Four days later, images of sprouts indicating endothelial cell growth were obtained from 5 randomly isolated fields using a phase-contrast microscope (CKX41; Olympus).

Four- or 5-week-old female nude (BALB/cAnNCrjBgi-nu) mice were maintained following the procedures in the Seoul National University Laboratory Animal Maintenance Manual and institutional review board approval. Mice were injected subcutaneously in the flank with 5 × 10⁶ of viable SNU449Cp or SNU449Tp cells in PBS. Tumor volumes were measured with a caliper and calculated using the formula: volume = (a × b²)/2, where a was the width at the widest point of the tumor and b was the maximal width perpendicular to a. Anti–integrin α₅ (P1D6 clone) or -VEGF (Upstate Biotechnology) antibody or normal IgG at 2.5 mg/kg was administered directly to tumors twice a week after the tumor dimension reached approximately 100 mm³.

Human liver tissues (normal or with tumors) were obtained from patients in the Kyungpook National University Hospital (Daegu, South Korea) with informed consent obtained in accordance with the Declaration of Helsinki and institutional review board approval. Normal tissues were cut from an area grossly free of tumor (> 5 times of tumor size distal from the tumor's visible edge). Biologic and clinical information was available for the samples.

Mouse tumor tissues were removed approximately 6 weeks after cell injection. Mouse tumor or clinical hepatocarcinoma tissues were fixed with formalin and embedded in paraffin. Serial sections (6 μm thick) were stained with anti-CD31 or –integrin α₅ before visualization by fluorescent microscopy (BX51; Olympus). Slides were viewed with an Olympus BX51 research microscope for images at DIC (Nomarski) units (UPLFLN 10-100×) using a digital camera (KY-1030BU; JVC, Yokohama, Japan), which were processed with Ky-Link software (JVC) and Adobe Photoshop version 7.0 software (Adobe Systems, San Jose, CA).

HUVECs were seeded into 24-well culture plates (10⁵ cells/well). One day later, triplicate cultures were treated with conditioned media from SNU449, SNU449Cp, SNU449Tp, and SNU449T16 cells. During conditioned media incubation, viable cells were counted by trypan blue staining.

Immunohistochemistry images for CD31 were photographed (× 40) and scanned in Adobe Photoshop (San Diego, CA). The CD31-positive area from at least 5 independent images was quantified with Image-Pro Plus (Media Cybernetics, Silver Spring, MD). Student t tests were performed, where P values less than .05 were considered significant.

We recently reported that TM4SF5 expression in hepatocytes causes epithelial-mesenchymal transition and loss of contact inhibition.²⁴  To clarify the role of TM4SF5 in angiogenesis, we examined whether the pattern of TM4SF5 expression in normal or tumor liver tissues correlated with VEGF expression and/or vessel formation. Immunoblots of tissue lysates revealed that 7 hepatocarcinoma tissues (cases 1, 3-6, 8, and 9, Figure 1) showed enhanced levels of TM4SF5 and VEGF compared with normal tissues. Two hepatocarcinoma samples (cases 2 and 7) did not show a correlation between TM4SF5 and VEGF expression (Figure 1). Immunohistochemical detection for vessel formation further showed that 7 of 9 hepatocarcinoma tissues were positive for the endothelial cell marker CD31 (cases 1, 3-6, 8, and 9, Figure 1B). Therefore, the TM4SF5-positive tumors showed a tendency for enhanced VEGF expression and vessel formation, compared with normal tissues, indicating that TM4SF5 may function in tumor angiogenesis.

The correlation between TM4SF5 and VEGF expression in hepatocarcinoma tissues suggests that TM4SF5 may mediate VEGF expression. Therefore, we examined the effects of transient TM4SF5 expression on VEGF promoter activity in hepatocytes. When TM4SF5-null SNU449 hepatocytes were transfected with TM4SF5 plasmids, VEGF promoter activity was enhanced by an average of 3- to 4-fold, compared with that of mock control-transfected cells (Figure 2A). Consistently, expression of VEGF (presumably 121, 165, and 189 amino acid human VEGF-A splice variants) was enhanced in transiently TM4SF5-transfected SNU449 or stably transfected SNU449Tp or T16 cells (Figure 2B,C). We then examined whether the TM4SF5-expressing cells secreted VEGF. Conditioned media were prepared from stable SNU449Cp cells lacking TM4SF5 and SNU449Tp cells expressing TM4SF5 and arrayed with human angiogenic antibodies. Interestingly, the conditioned media from TM4SF5-expressing cells showed a significant amount of secreted VEGF in the culture media compared with negative controls (Figure 2D). These observations indicate that TM4SF5 expression results in effective VEGF induction and secretion.

We next investigated which signaling molecules are responsible for TM4SF5-mediated VEGF induction. We treated SNU449Cp and SNU449Tp cells with pharmacologic inhibitors for 24 hours before lysate preparation for immunoblots. Pharmacologic reagents included dimethyl sulfoxide (as a control vehicle), U0126 and PD98056 (as specific MEK/Erk1/2 inhibitors), PP2 (a specific c-Src family kinase inhibitor), PP3 (a negative control compound of PP2), AG490 (a specific JAK inhibitor), and Y27632 (a specific ROCK inhibitor). TM4SF5-enhanced VEGF induction was inhibited by PP2 treatment, but not by treatment with the other reagents (Figure 2E). This observation indicates that TM4SF5-mediated VEGF induction requires c-Src family kinase activity, but not MEK/Erk1/2 and JAK activities.

We then analyzed proteins involved in the signaling networks upstream or downstream of c-Src family kinase by immunoblots. In particular, we examined FAK and c-Src levels, which are major signaling mediators downstream of integrins,^12,13  that may participate in cross-talk with TM4SFs and/or tetraspans.¹⁸  We previously reported that phosphorylation of FAK Tyr577 (pY⁵⁷⁷FAK) correlated with c-Src phosphorylation (pY⁴¹⁶c-Src) in TM4SF5-expressing cells,²⁴  and that STAT3 downstream of c-Src can induce VEGF promoter activity.²⁸  TM4SF5 expression resulted in enhanced pY⁵⁷⁷FAK, pY⁸⁶¹FAK (but not pY³⁹⁷FAK or pY⁹²⁵FAK), pY416c-Src, pY⁷⁰⁵STAT3, and VEGF protein levels. Expression levels of major integrin subunits were not altered by transient transfection of TM4SF5 (Figure 2F). These enhanced signaling activities and VEGF levels were blocked by transfection with shRNA against TM4SF5 (Figure 2G), or with siRNA against c-Src (Figure 2H) or STAT3 (Figure 2I), but not with shRNA against GFP protein, without subsequent alteration of expression of major integrin subunits. Furthermore, shTM4SF5 introduction to the TM4SF5-endogenously expressing Huh7 cell line showed similar results (Figure 2J).

We next examined whether conditioned media from TM4SF5-expressing cells could enhance viability and angiogenic activity of primary HUVECs. Conditioned media prepared from cultures of diverse cells were added to primary HUVEC cultures. HUVEC viability increased substantially after treatment with conditioned media from TM4SF5-postive cells, whereas conditioned media from TM4SF5-null SNU449 or SNU449Cp cells did not have an effect (Figure 3A). We examined whether vessel-like tube formation by primary HUVECs on Matrigel is influenced by conditioned media from TM4SF5-expressing cells. Primary HUVECs efficiently formed vessel-like tubes on Matrigel. When HUVECs were treated with conditioned media from either TM4SF5-null or TM4SF5-positive cells for 10 hours, similar vessel-like tubes formed (Figure 3B left). However, maintenance of the tube-like structure was longer (up to 24 hours) with treatment with conditioned media from TM4SF5-positive cells relative to treatment with media from TM4SF5-null cells (Figure 3B right). This sustaining effect was blocked by preincubation of cells with anti-VEGF antibody, but not with normal IgG (Figure 3C).

Because we observed that TM4SF5 expression resulted in VEGF induction and secretion (Figure 2) and that conditioned media from TM4SF5-expressing cells enhanced angiogenic activity of the HUVECs (Figure 3A,B), we next examined whether conditioned media from TM4SF5-expressing cells contains sufficient amounts of VEGF to promote angiogenesis using a mouse aorta ring assay. Aorta ring segments embedded in Matrigel were incubated with conditioned media from TM4SF5-expressing and control cells (Figure 3D top and middle). Enhanced outgrowth of endothelial cells was observed after treatment with the conditioned media from TM4SF5-expressing cells but was blocked by preincubation of the conditioned media with anti-VEGF antibody (Figure 3D bottom). These results indicate that secreted VEGF indeed facilitates endothelial cell outgrowth.

The expression levels of major integrin subunits were not changed by transient expression of TM4SF5 during immunoblotting (Figure 2). Because integrin-mediated signaling appeared to be involved in TM4SF5-mediated VEGF induction, we next explored cell-surface expression of integrins by flow cytometric analysis, using cells mock- or TM4SF5-transfected cells without or with permeabilization before antibody incubation. Integrin α₅ levels on TM4SF5-transfected cells were higher than on mock-transfected cells, under conditions where the membrane is not permeabilized, thus allowing detection solely of surface expressed integrins. Meanwhile, total integrin α₅ levels visualized after permeabilization were similar between mock- and TM4SF5-transfected cells (Figure 4A). Increased integrin α₅ expression on the surface of TM4SF5-transfected cells would result in more extracellular domain-cleaved integrin α₅ on protease treatment. To test this possibility, equal numbers of cells were treated with various concentrations of trypsin for 15 minutes at room temperature. As more trypsin was treated, immunoblot analysis using an antibody that recognizes an extracellular integrin α₅ region (ie, α₅ecto) showed lower levels of integrin α₅ in TM4SF5-transfected cells, compared with mock-transfected cells. This indicates that intact and intracellular integrin α₅ were recognized by α₅ecto antibody more in mock-transfected cells. Immunoblots using an anti–α₅cyto antibody to recognize the cytoplasmic tail of integrin α₅ revealed similar levels of integrin α₅ in both mock- and TM4SF5-transfected cells (Figure 4B). These observations indicate that TM4SF5 expression increased integrin α₅ on the cell surface. We next examined whether functional-blocking antibody of integrin α₅ (clone P1D6) would inhibit TM4SF5-mediated VEGF induction and secretion. Cells were incubated for 20 minutes with normal mouse IgG or P1D6 integrin α₅ antibody, and reseeded on normal culture dishes for an additional 10 hours. Immunoblots showed decreased TM4SF5-mediated signaling activity and VEGF induction with functional-blocking antibody (Figure 4C lanes 3 and 4). Anti-body array analysis of conditioned media revealed that VEGF secretion was inhibited when TM4SF5-positive cells were incubated with integrin α₅-blocking antibody, although other angiogenic molecules detected using the array were unchanged (Figure 4D). These observations suggest that TM4SF5 expression in hepatocytes requires integrin α₅ on the cell surface for VEGF induction and secretion.

We next examined whether TM4SF5 associates with integrin α₅, using a coimmunoprecipitation approach. Whole-cell lysates obtained from SNU449 cells that were transfected with mock-, mock plus myc-(His)₆-TM4SF5, or myc-(His)₆-TM4SF5 plus either WT integrin α₅, tailless α₅,²⁶  or chimeric α₅ tail (X4C5³⁰ ). Anti-(His)₆ antibody immunoprecipitates were immunoblotted for integrin α₅. Interestingly, WT integrin α₅ and X4C5 chimera were coimmunoprecipitated by TM4SF5, although X4C5 was coimmunoprecipitated less presumably because of less expression, whereas tailless integrin α₅ was not (Figure 4E). However, anti-(His)₆ immunoprecipitates did not precipitate integrin α₅, when myc-(His)₆–TM4SF5 was not transfected (Figure S1, available on the Blood website; see the Supplemental Materials link at the top of the online article). These TM4SF5-mediated effects were shown also in another cell line (SNU398; Figure 4F). Therefore, TM4SF5 appeared to associate at least with the cytoplasmic tail of integrin α₅.

TM4SF5-mediated VEGF expression correlated with phosphorylation of FAK, c-Src, and STAT3 (Figure 2), and TM4SF5 expression resulted in retained integrin α₅ on the cell surface (Figure 3). We thus investigated whether integrin α₅ was required for the TM4SF5-mediated VEGF promoter activity. First, we performed the VEGF promoter assay using SNU449 cells cotransfected with TM4SF5 and either WT integrin α₅ or tailless integrin α₅. TM4SF5 transfection resulted in approximately 3-fold increase in VEGF promoter activity. Cotransfection with WT integrin α₅ greatly enhanced the promoter activity (to ∼ 7 fold), but cotransfection with tailless integrin α₅ abolished TM4SF5-mediated VEGF promoter activity (Figure 5A). When TM4SF5 alone was transfected, pY416c-Src, pY⁷⁰⁵STAT3, and VEGF protein levels were increased (Figure 5B lanes 1 and 2). Integrin α₅ overexpression caused further signaling enhancement and increased VEGF protein levels, whereas tailless integrin α₅ abolished the TM4SF5-mediated effects (Figure 5B). When SNU449 cells were transfected with TM4SF5 and either WT FAK, FRNK (a naturally occurring dominant-negative inhibitor of FAK), or FAK mutants (Y397F or Y577F), WT FAK positively enhanced TM4SF5-mediated VEGF promoter activity (Figure 5C) and VEGF induction-related signaling activities (Figure 5D). However, cotransfection of FRNK or FAK mutants did not further enhance VEGF promoter activity or VEGF-induction signaling and did not abolish the TM4SF5-mediated effects (Figure 5C,D).

We next examined the requirement of c-Src for the TM4SF5-mediated effects. c-Src cotransfection with TM4SF5 greatly enhanced TM4SF5-mediated VEGF promoter activity and pY⁷⁰⁵STAT3, whereas cotransfection with an inactive Y416F c-Src mutant abolished the TM4SF5 effects (Figure 5E,F). In addition, WT STAT cotransfection with TM4SF5 resulted in increased TM4SF5-mediated VEGF promoter activity and VEGF expression, whereas the dominant negative mutant of STAT3 (DN Y705F) almost completely abolished the TM4SF5 effects (Figure 5G,H). Together, these data indicate that TM4SF5 cooperates with integrin α₅ to induce VEGF through signal transduction, including FAK–c-Src and STAT3 activation.

We next investigated whether the TM4SF5-mediated effects observed in vitro also occur in vivo. We subcutaneously injected nude mice with SNU449Cp or SNU449Tp cells and measured tumor growth. As shown in Figure 6A, mice injected with TM4SF5-expressing SNU449Tp cells formed more aggressive tumors, compared with mice injected with TM4SF5-null control cells. Injection of another cell clone expressing TM4SF5 at half the level of SNU449Tp cells resulted in formation of smaller tumors, approximately 20% plus or minus 3.5% less in average volume (data not shown), indicating that the tumor formation was influenced by the level of TM4SF5 expression. TM4SF5-mediated tumor or normal epidermal tissues (from locations injected with control cells) were immunostained with anti-CD31. The SNU449 cell-derived tumor showed more vessel formation (Figure 6B). Furthermore, immunoblots of tissue extracts from the TM4SF5-mediated tumor showed enhanced levels of integrin α₅, pY⁵⁷⁷FAK, pY416c-Src, pY⁷⁰⁵STAT3, and VEGF, compared with those for normal epidermal tissue (Figure 6C). In human liver tissues, integrin α₅ overexpression was pronounced at the membrane surface (Figure 6D), and a close linkage exists between increased levels of TM4SF5, integrin α₅, pY⁴¹⁶c-Src, pY⁷⁰⁵STAT3, and VEGF (Figures 1A, 6E).

We next examined whether anti–integrin α₅ or -VEGF antibody administration would inhibit the TM4SF5-mediated tumor and vessel formation in mice and the underlying signaling. When TM4SF5-mediated tumors reached approximately 100 mm³ in size, antibodies were injected directly to the tumors twice a week. Interestingly, the administration of the antibodies, but not normal IgG, efficiently (∼ 80%) blocked TM4SF5-mediated tumor formation (Figure 7A) and vessel formation (Figure 7B). In addition, the signaling for TM4SF5-mediated VEGF expression was also blocked by anti–integrin α₅ antibody administration, whereas TM4SF5 levels were unchanged (Figure 7C). Possibly, anti-VEGF antibody-mediated sequestering of extracellular VEGF might inhibit the tumor growth, although the VEGF level in tissue extracts was unchanged (Figure 7D).

Observations from this study suggest that tetraspan TM4SF5 cooperates with integrin α₅ to transduce signaling through FAK-c-Src complex activation and STAT3 phosphorylation to induce VEGF in cancerous epithelial cells. Secreted VEGF promotes angiogenic activities in neighboring endothelial cells, presumably allowing efficient angiogenesis (Figure 7E).

Administration of functional-blocking integrin α₅ antibody into TM4SF5-mediated tumors blocked the vessel formation and tumor growth. Functional neutralization of cell-surface integrin α₅ on TM4SF5-expressing SNU449Tp cells using function blocking antibody as well as cotransfection of TM4SF5 with a cytoplasmic tailless integrin α₅ mutant suggests that the integrin α₅ cytoplasmic tail is required for TM4SF5-mediated VEGF expression and secretion. Consistent with these observations, tailless integrin α₅ did not bind TM4SF5, whereas the cytoplasmic tail of integrin α₅ in the X4C5 chimera did bind TM4SF5. TM4SF5 expression appeared to retain integrin α₅ on the cell surface, presumably via TM4SF5-mediated influence on internalization and/or trafficking. This increased integrin α₅ retention on TM4SF5-expressing cell surface resulted in more efficient adhesion and spreading on fibronectin (S.-A.L. and J.W.L., unpublished data, May 2008). It has been previously reported that the TM4SF5-related tetraspan L6-Ag positively modulated cell motility by regulating surface retention and endocytosis of tetraspanins CD63 and CD82 without any changes in the total expression levels of the tetraspanins.³¹  SNU449Tp cells and tumor tissues in nude mice injected with TM4SF5-expressing cells showed enhanced levels of integrin α₅ compared with SNU449Cp cells and normal epidermal tissue around the injection site, presumably because of TM4SF5-mediated integrin α₅ stabilization on the cell surface. However, the binding affinities of TM4SF5 to different integrin subunits may differ. We have observed coimmunoprecipitation between the extracellular domains of TM4SF5 and integrins α₂ or β₁ in SNU449Tp cells, whereas coimmunoprecipitation of TM4SF5 with endogenous integrin α₅ was insignificant (S.-A.L. and J.W.L., data not shown). Moreover, the association between TM4SF5 and integrin α₂ (not α₅) was observed in fibroblasts ectopically expressing TM4SF5 and was abolished by growth factor signaling.²⁵  The cytoplasmic tails of integrin α₅ and TM4SF5 are short and thus recruit other signaling molecules to mediate cooperation between both receptors. Although it is unclear whether the association between TM4SF5 and integrin α₅ tail is direct or indirect, it will be interesting to identify cytoplasmic mediator(s) for the cooperation that bind either receptor.

Integrin α₅β₁ engagement with fibronectin causes recruitment of FAK to focal adhesions through interaction between the FAK C-terminal domain and integrin β₁-binding proteins, such as talin and paxillin.³²  It was recently reported that integrin α₅β₁-mediated neuroblastoma migration involves FAK-mediated c-Src activation.³³  Evidence for a reciprocal influence between c-Src and FAK has been documented; when cells adhere, FAK is autophosphorylated at Tyr397, and c-Src is recruited to form a FAK/c-Src complex, leading to c-Src activation.⁸  Activated c-Src bound to pY³⁹⁷FAK can phosphorylate other tyrosine residues, including pY⁵⁷⁶FAK and Y⁵⁷⁷FAK for full FAK activation.⁸  Interestingly, FAK appears to play a redundant role in TM4SF5-mediated VEGF induction. Although integrin α₅ was required for TM4SF5-mediated VEGF induction and WT FAK overexpression could synergistically enhance the TM4SF5 effects, inhibition of FAK through expression of FRNK or FAK mutants did not abolish the TM4SF5-mediated transcriptional activation of VEGF promoter and induction of VEGF. Therefore, it appears that TM4SF5-mediated VEGF induction may not depend completely on FAK activity, although FAK can increase the TM4SF5 effects. It may be probable that the role of integrin α₅ in c-Src activation is required and that FAK activation is redundant, for TM4SF5-mediated VEGF expression. These observations indicate that cooperative complexes formed between c-Src and FAK may be different depending on the signaling contexts including TM4SF5. Moreover, c-Src was required for the TM4SF5-mediated effects, which were abolished by introduction of siRNA against c-Src or of an inactive Y416F c-Src mutant. Therefore, it cannot be ruled out that c-Src activity emanating directly from TM4SF5 is sufficient for VEGF induction to a certain degree when endogenous integrin α₅ expression on the cell surface may be insignificant. To our knowledge, ligands for most tetraspan(in)s are unknown, and their regulatory roles in cellular signal transduction and function may involve cross-talk with other receptors. It will be interesting to clarify whether TM4SF5-mediated c-Src activation occurs through the TM4SF5 signaling capacity or through cross-talk with other receptors.

TM4SF5 mRNA is known to be highly expressed in diverse tumor types, including pancreatic,²³  nonendocrine lung, and corticotropin-negative bronchial carcinoid tumors.³⁴  We recently observed that TM4SF5 protein is highly overexpressed in human hepatocarcinoma,²⁴  and colon carcinoma tissues (S.-A.L. and J.W.L., unpublished data, July 2008). TM4SF5 belongs to a 4-transmembrane L6 superfamily consisting of L6 (or L6-Ag), IL-TMP, and L6D.²²  Tetraspanins or TM4SFs have 4 transmembrane domains, 2 extracellular loops, and 2 short cytoplasmic tails.¹⁹  The extracellular loop 2 of genuine tetraspanins has a variable region that mediates specific protein-protein interactions and a conserved region that includes 4 conserved cysteines for disulfide bonding, correct folding, and homodimerization.¹⁹  However, the extracellular loop 2 loop of TM4SF5 is relatively divergent, containing only 2 cysteine residues.²²  The 4 transmembrane domains are suggested to be responsible for intramolecular and intermolecular interactions during organizing the tetraspanin-web or TERM on the cell surface¹⁹  and for forming complexes with integrins to collaboratively regulate cell adhesion, proliferation, and motility.^18,20  We recently reported that TM4SF5 causes morphologic changes through RhoA inactivation, epithelial-mesenchymal transition, abnormal cell-cycle progression, and loss of contact inhibition in vitro and in vivo in tumor formation in nude mice.²⁴  Our previous study implicated the existence of cross-talk between TM4SF5 and integrin signaling pathways. In addition, ectopic expression of TM4SF5 in fibroblasts regulated actin reorganization and focal adhesion turnover via association with integrin α₂, which appears to be competitively regulated by growth factor receptor signaling.²⁵  In this study, TM4SF5 enhanced surface retention of integrin α₅ without changing total protein levels. Function-blocking of surface integrin α₅ using its antibody abolished TM4SF5-mediated VEGF induction and secretion. Furthermore, anti–integrin α₅ antibody administrated into mice might bind integrin α₅ on cell surface, so that its resultant conformation change might disrupt the intracellular association between TM4SF5 and integrin α₅, leading to inhibition of c-Src/STAT3 signaling, vessel formation, and tumor growth in nude mice. Similarly, tumor-associated antigen L6 (L6-Ag), which has a sequence identity of approximately 50% with TM4SF5,²²  was recruited to the TERM, and the recruitment was correlated with the promigratory activity of L6-Ag through regulation of surface retention and endocytosis of CD63 and CD82 but not of integrin β₁.³¹  Therefore, such as genuine tetraspanins, TM4SF5, and L6-Ag, also appear to play roles in TERM by regulating networks between membrane receptors, including tetraspanins and integrins. TM4SF5 may facilitate efficient communication between cancer cells and extracellular environments (including nearby endothelial cells) through regulation of architectural integrity with diverse membrane receptors. Therefore, developing therapeutics to target TM4SF5 may be a strategy to inhibit tumorigenesis and angiogenesis.

This work was supported by the Research Program for New Drug Target Discovery (2007-03536) and a Korea Science and Engineering Foundation grant (Cell Dynamics Research Center R11-2007-007-01 004-0) to J.W.L.

Contribution: S.C. performed experiments and wrote the manuscript; S.-A.L., T.K.K., H.J.K., and M.J.L. helped perform animal experiments; S.-K.Y., S.-H.K., and S.K. supplied materials and discussed data; J.W.L. designed the research and edited the paper; and all authors interpreted data.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Jung Weon Lee, Cancer Research Institute, College of Medicine, Seoul National University, Seoul 110-799, Korea; e-mail: jwl@snu.ac.kr.