These two authors contributed equally to this work.
Cancer Cell Biology
The tetraspanin CD9 inhibits the proliferation and tumorigenicity of human colon carcinoma cells
Article first published online: 20 JUN 2007
Copyright © 2007 Wiley-Liss, Inc.
International Journal of Cancer
Volume 121, Issue 10, pages 2140–2152, 15 November 2007
How to Cite
Ovalle, S., Gutiérrez-López, M. D., Olmo, N., Turnay, J., Lizarbe, M. A., Majano, P., Molina-Jiménez, F., López-Cabrera, M., Yáñez-Mó, M., Sánchez-Madrid, F. and Cabañas, C. (2007), The tetraspanin CD9 inhibits the proliferation and tumorigenicity of human colon carcinoma cells. Int. J. Cancer, 121: 2140–2152. doi: 10.1002/ijc.22902
- Issue published online: 25 SEP 2007
- Article first published online: 20 JUN 2007
- Manuscript Accepted: 9 MAY 2007
- Manuscript Received: 22 SEP 2006
- Fundación de Investigación Médica Mutua Madrileña
- Dirección General de Investigación, Ministerio de Educación y Ciencia. Grant Numbers: SAF2004-01715, BFU2005-02671
- Consejo Superior de Investigaciones Científicas (CSIC)
The implication of the tetraspanin CD9 in cancer has received much recent attention and an inverse correlation between CD9 expression and the metastatic potential and cancer survival rate has been established for different tumor types. In contrast to the well-established role of CD9 in metastasis, very little is known about the involvement of this tetraspanin in the process of development of primary tumors. In the present study, we present evidence on the implication of CD9 in colon carcinoma tumorigenesis. We report here that ectopic expression of CD9 in colon carcinoma cells results in enhanced integrin-dependent adhesion and inhibition of cell growth. Consistently with these effects, treatment of these cells with anti-CD9-specific antibodies resulted in (i) increased β1 integrin-mediated cell adhesion through a mechanism involving clustering of integrin molecules rather than altered affinity; (ii) induction of morphological changes characterized by the acquisition of an elongated cell phenotype; (iii) inhibition of cell proliferation with no significant effect on cell survival; (iv) increased expression of membrane TNF-α, and finally (v) inhibition of the in vivo tumorigenic capacity in nude mice. In addition, through the use of selective blockers of TNF-α, we have demonstrated that this cytokine partly mediates the antiproliferative effects of CD9. These results clearly establish for the first time a role for CD9 in the tumorigenic process. © 2007 Wiley-Liss, Inc.
The tetraspanin CD9 is a widely distributed surface molecule implicated in diverse functions, including cell signaling, growth, adhesion and motility, metastasis and sperm–egg fusion. Like other members of the tetraspanin protein family, CD9 participates in the organization of cell surface protein microdomains, termed “the tetraspanin web,” through association with other transmembrane proteins including members of the integrin family of adhesion receptors.1, 2, 3
The implication of CD9 in cancer has received much attention. An inverse correlation between its expression in primary tumors and the metastatic potential and patient survival rate has been established for melanomas and colon, lung and breast carcinomas.1, 4, 5, 6, 7, 8 The involvement of this tetraspanin in tumor progression has been inferred from the effects of CD9 antibodies or CD9 overexpression on tumor cell motility and migration. In this regard, it has been shown that overexpression of this tetraspanin in melanoma and breast, lung, pancreas and colon carcinoma cells suppresses the motility and metastatic potential of these cells.5, 7, 9, 10 mAbs directed to the CD9 molecule have also been shown to inhibit the migration of different types of carcinoma tumor cells and the transendothelial migration of melanoma cells.11, 12, 13 Therefore, the inverse correlation observed between CD9 expression and the metastatic potential could be explained, at least in part, by the effects mediated by CD9 on cell adhesion and migration through the interactions among this tetraspanin and different integrins.2, 3
We have recently described a functional conformation of CD9, detected by the anti-CD9 mAb PAINS-13, that is imposed upon association of this tetraspanin with members of the beta-1 subfamily of integrins and particularly with α6β1, a cellular adhesion receptor for laminin. This mAb exerts important effects on endothelial cell adhesion, spreading, migration and morphogenetic processes related to angiogenesis,14 and all these effects are more prominently observed when cells are plated on laminin, in clear difference to other anti-CD9 mAbs that display conformation-independent accessibility to epitopes on this tetraspanin.
In contrast to the well-documented implication of CD9 in the metastatic process, very little is known about the involvement of CD9 in the tumorigenic process itself, i.e. in malignant cell transformation and the development of primary tumors. As changes in cell adhesion, apoptosis, proliferation and signaling are crucial cellular processes in the tumorigenesis of many types of carcinomas, we have characterized in this study the effects of ectopic expression of CD9 as well as the effects of mAbs specific for CD9 on these phenomena in different established human colon carcinoma cell lines: Colo320, HT29 and BCS-TC2.2. In addition, in vivo tumorigenesis studies have been performed with the human BCS-TC2.215, 16 and Colo320 cell lines, which possess intrinsic tumorigenic capacity. Ectopic expression of CD9 in CD9-negative Colo320 carcinoma cells resulted in enhanced integrin-mediated cell adhesion and decreased proliferative and tumorigenic capacities, clearly demonstrating the role of this tetraspanin in these processes. In agreement with these results, several CD9-specific mAbs also enhanced integrin-mediated cell adhesion and induced in vitro characteristic morphological changes and an important inhibition of their proliferative and in vivo tumorigenic capacities. We also demonstrate that TNF-α is a mediator of these effects through the use of selective blockers of this cytokine.
Materials and methods
Cell culture, antibodies and reagents
Colo320-CD9 cells were generated by transfection of CD9-negative Colo320 colocarcinoma cells in 2.5% FCS–RPMI-1640 with 20 μg of cDNA coding for human CD9. Cells were electroporated at 412.5 V/cm in an ElectroSquarePorator ECM 830 (BTX, Holliston, MA) and positive clones were selected using G418 (0.8 mg/ml) in culture medium. Parental and transfected colon carcinoma cells were grown in RPMI-1640 medium supplemented with 10% FCS.
The anti-β1-integrin mAbs used were TS2/16,17 Lia1/218 and HUTS-21.19 The anti-CD9 antibodies used were VJ1/20 and VJ1/10,20 PAINS-10, PAINS-22, PAINS-26 (unpublished results) and PAINS-13,14 all obtained in our laboratories. The TP1/8 mAb21 was used as an isotype control. Monoclonal antibodies were purified from ascitic fluid by affinity chromatography using Protein A/G-sepharose columns. When necessary, purified antibodies were biotinylated with biotin-3-sulfo-N-hydroxysuccinimide ester (Sigma-Aldrich, Madrid, Spain) as previously described.14 Other antibodies used were anti-E-cadherin (Calbiochem, Darmstadt, Germany), antipancytokeratin (Sigma-Aldrich), anti-human TNF-α AM91822 and the humanized mAb anti-TNF-α Infliximab (Remicade®, Schering Plough, Kenilworth, NJ). Etanercept (Enbrel®), a soluble form of a fusion protein containing the extracellular region of human TNF-RII, was purchased from Wyeth (Madrid, Spain). Recombinant human TNF-α was purchased from Peprotech Ec (London, UK). Laminin-1 (Ln), fibronectin (Fn) and poly-L-lysine (PLL) were all purchased from Sigma-Aldrich. FITC-labeled annexin-V was obtained as previously described.23
Flow cytometric analysis
Cells were washed twice, incubated with primary antibodies at 4°C for 30 min, washed three times and incubated for 30 min on ice with fluorescein–isothiocyanate-conjugated anti-mouse IgG (Sigma). After washing, cells were fixed in 2% formaldehyde and fluorescence measured using a FACSCalibur™ flow cytometer (Becton Dickinson, San Jose, CA).
Cell adhesion studies
BCS-TC2.2, HT-29, Colo320 and Colo320-CD9 cells were incubated in medium containing 10 μg/ml of appropriate purified antibodies or 10 ng/ml of human TNF-α in 96-microwell plates coated with 10 μg/ml Ln, 5 μg/ml Fn or 50 μg/ml PLL for 2 hr at 37°C. Adhered cells were fixed with 3.7% formaldehyde, permeabilized with 2% methanol and stained with 70% crystal violet. Adhesion was analyzed by measuring absorbance of eluted crystal violet at 540 nm as previously described.14 Percentage of cell adhesion was referred to untreated control cells, with the exception of the percentages of Colo320-CD9, which were referred to parental Colo320 untreated control cells. Changes in morphology were photographed on a DM-IRE2 inverted microscope (Leica, Barcelona, Spain).
β1-integrin affinity analysis
BCS-TC2.2 cells were treated with MnCl2 (500 μM) or antibodies (10 μg/ml) for 2 hr at 37°C and incubated with the biotinylated HUTS-21 antibody followed by avidin–FITC (Sigma-Aldrich). Cells were analyzed using a FACSCalibur™ flow cytometer.14
β1-integrin clustering analysis
BCS-TC2.2 cells were allowed to adhere on glass coverlips precoated with PLL in the presence of PAINS-13 or VJ1/20 for 2 hr at 37°C. Cell fixation with 3.7% formaldehyde was performed before the incubation with the primary biotinylated antibody to avoid antibody-dependent clustering of β1-integrins. Nonspecific binding sites were blocked with blocking reagent (Boehringer Mannheim, Germany). Cells were incubated with the biotinylated antibody TS2/16 followed by streptavidin-Alexa488 (Molecular Probes, Leiden, Netherlands). Series of optical sections were obtained with a Leica TCS-SP confocal laser scanning unit equipped with Ar and He/Ne laser beams and attached to a Leica DM-IRBE inverted epifluorescence microscope using a 63× oil immersion objective.
To study the role of CD9 in cell proliferation, 3 × 105 Colo320 and Colo320-CD9 cells were grown on Fn (5 μg/ml) and collected at the indicated times. To study the effects of CD9 mAbs, 5 × 105 BCS-TC2.2, HT29, Colo320 or Colo320-CD9 cells were grown on Fn (5 μg/ml) or Ln (10 μg/ml) precoated plates in 0.5% FCS–DMEM in the absence or presence of the respective antibodies (10 μg/ml) for 1, 3, 5 or 8 days (cultures were supplemented with the corresponding fresh antibodies every 2 days) or in presence of increasing concentrations (5, 10 and 20 μg/ml) of the indicated mAb for 24 hr. To study the effects of short-period treatment, Colo320-CD9 cells were treated with 10 μg/ml of the mAb for 1 hr at 4°C and excess mAb was washed prior to seeding the cells on Fn-precoated dishes and proliferation of cell cultures was measured after 24, 48 and 72 hr. To measure the effect of TNF-α, cells were treated with varying TNF-α concentrations (0–20 ng/ml) for 24 hr. In TNF-α blocking assays, cells were pretreated for 30 min with 50 μg/ml of either infliximab or etanercept, and these agents were also present over the duration of the assays (24 hr). In all cases, cells were recovered and counted in a Neubauer chamber or in a FACSCalibur flow cytometer for 30 sec. Percentage of proliferation was obtained relative to untreated control cells. In particular, the percentages of Colo320-CD9 were referred to parental Colo320 untreated control cells. Cell proliferation was also quantified by CFSE labeling experiments, in which 10 × 106 cells/ml were labeled with 1 μM CFSE (Molecular Probes) for 15 min at 37°C, washed and then the CFSE fluorescence measured in a FACSCalibur flow cytometer at different times (1, 3, 5 and 8 days), or by MTT-assays performed as described.24, 25
Cell cycle, apoptosis and necrosis assays
BCS-TC2.2 cells in suspension were treated with the anti-CD9 antibodies for 1 hr at 4°C or were plated on Ln- or Fn-precoated plates in medium containing the indicated mAbs for 2–120 hr. Chenodeoxycholic acid (CDCA) was added to the cells during 2 hr as a positive apoptotic inducer.25 In both cases, cells were collected and prechilled 70% ethanol was added and incubated at 4°C for 2 hr. DNA was stained with 0.1% Triton X-100, 20 mg/ml propidium iodide in TBS in the presence of an equal volume of DNase-free RNase (200 μg/ml; Boehringer Mannheim) and analyzed by flow cytometry. Debris and clumps were excluded by monitoring FL2-A and FL2-W parameters. Fluorescence signals from 10,000 cells were collected.
To study the induction of apoptosis, cells were cultured in presence of corresponding mAbs for varying periods of time (1, 3, 5 or 8 days), TNF-α for 1 day or CDCA for 2 hr. Cells were trypsinized and resuspended in 10 mM HEPES (pH 7.4) containing 150 mM NaCl, 5mM KCl, 1 mM MgCl2 and 1.8 mM CaCl2 and incubated for 15 min in the presence of 2 μg/ml annexin-V–FITC. To discern necrotic cells, 0.005% propidium iodide was added. Analyses were performed in a FACSCalibur as previously reported.16 Leakage of lactate dehydrogenase (LDH) was measured as an index of lethal membrane injury or necrosis. Briefly, release of LDH into the culture medium was evaluated in triplicate experiments after 24–72 hr incubation of BCS-TC2.2, Colo320-CD9 and HT29 cells with antibodies (10 μg/ml) or TNF-α (0.2–15 ng/ml). LDH activity was determined following the oxidation of NADH at 340 nm and 37°C using a Beckman DU640 spectrophotometer.
Tumor development in nude mice
In order to assess the effect of the different mAbs on the in vivo growth of BCS-TC2.2 tumors, these were previously formed by subcutaneous (s.c.) inoculation of 106 cells in the lumbar region of male nude mice (8 weeks old; Harlan Ibérica, Barcelona, Spain). When tumors reached 4–6 mm in size, mice received every 2–3 days intratumoral injections of PBS (as control) or the different mAbs. Tumor size was calculated by measuring the dimensions of the tumor mass (average of the two right-angle diameters, the largest and its perpendicular) with a Vernier caliper.16 The relative tumor size variation was calculated as St/S0, being St the tumor size at each time point and S0 the tumor size at time 0 (before the first mAb injection).
Ex vivo effect of PAINS-13 mAb treatment was analyzed by incubation of BCS-TC2.2 cells (1 hr at 4°C; 1 or 2 mg/ml of purified mAb) prior to their inoculation into nude mice. Cells were then s.c. injected into nude mice as described above and animals were surveyed daily to determine when tumors first became apparent upon visual inspection (latency time; size > 1 mm) and tumor sizes were measured every 2–3 days.
Tumorigenicity of Colo320 and Colo320-CD9 cells was assessed as described above but after s.c. injection of 4 × 106 cells into each mice. At day 14, tumors were excised, cleaned of surrounding tissue and homogenized in extraction buffer (10 mM Tris, pH 8.0, containing 140 mM NaCl, 2% Triton X-100, 2 mM PMSF, 1 μg/ml leupeptin and aprotinin, 1 μM pepstatin and 1 mM DTT). CD9 expression in tumor and cell extracts was analyzed by Western blot using VJ1/10 anti-human CD9 mAb as previously described.14, 19
Membrane TNF-α expression analysis
For flow cytometry detection, BCS-TC2.2, HT29 or Colo–CD9 cells were cultured for 24 hr at 37°C in 24-well plates previously coated with 10 μg/ml Ln in the presence of the corresponding purified antibodies (10 μg/ml). Cells were detached and incubated with biotinylated anti-TNF-α AM918 mAb for 30 min at 4°C, washed and incubated with streptavidin-FITC for 30 min on ice. Fluorescence was detected in the flow cytometer. Membrane TNF-α on carcinoma cells was alternatively measured using a cytotoxicity assay with mouse fibroblastic L929 cells. Briefly, colon carcinoma cells were incubated in the absence (control) or presence of the respective mAbs (10 μg/ml) for 24 hr, and then culture supernatants were collected and cells were fixed in 4% paraformaldehyde and washed extensively. The biological activity of membrane TNF-α expressed on fixed cells was determined by measuring in triplicates their cytotoxicity at 2:1 and 4:1 ratios against monolayers of L929 cells (5 × 104 L929 cells/well in 96-well culture plates), using the LDH method as described.26, 27
The normal distribution of data was determined by the Kolmogorov–Smirnov test. Two-way repeated measurements ANOVA analysis was performed to determine if there were significant differences in the in vivo studies.
Anti-CD9 antibodies induce integrin-dependent cell adhesion and morphological changes in colon carcinoma cells
The tetraspanin CD9 is known to associate with several members of the integrin family and alterations in integrin-mediated tumor cell adhesion and morphology are frequently associated with changes in tumor cell differentiation and proliferation. Therefore, we analyzed the functional effects of anti-CD9 mAbs on the adhesion to the integrin ligands laminin (Ln) and fibronectin (Fn) of different human colon carcinoma cells lines that differ in their level of CD9 surface expression. As shown in Figure 1a, BCS-TC2.2 and HT29 cells abundantly express CD9 on their surface as detected with the mAb VJ1/20 whereas no expression of CD9 can be detected by flow cytometry on the surface of Colo320 cells. Upon transfection of these CD9-negative Colo320 cells with the human CD9 cDNA we have generated a cell line (Colo320-CD9) which stably expresses high CD9 levels on their surface, as detected by mAb VJ1/20 (Fig. 1a). The expression of the CD9 epitope recognized by the specific mAb PAINS-13 on these cell lines as well as expression of β1-integrin detected by mAb TS2/16 is also shown in Figure 1a. Ectopic expression of CD9 on Colo320 cells resulted in enhanced adhesion (compared to the adhesion of CD9-negative Colo320 cells) to the integrin ligand Ln (Fig. 1b) and Fn (not shown), clearly implicating this tetraspanin in cell adhesion. Consistently with this finding, the anti-CD9 mAbs VJ1/20 and PAINS-13 also affected integrin-mediated cell adhesion as evidenced by the increased adhesion of BCS-TC2.2, Colo320-CD9 and HT29 colon carcinoma cells to Ln. Similar augmentation of adhesion of these cell lines to another major integrin ligand, fibronectin (Fn), was also induced by mAb VJ1/20 (not shown). The effects of mAb PAINS-13 on carcinoma cell adhesion to Fn were less evident, as previously described for endothelial cells.14 As expected, cell adhesion to Ln (and to Fn, not shown) was mainly mediated by β1 integrins as evidenced by the strong inhibition observed in all cases with Lia1/2 mAb, a blocking anti-β1 integrin antibody, as well as by the enhancing effect of the stimulatory mAb TS2/16.17 Furthermore, none of the antibodies used, including the anti-CD9 mAbs, exerted any effects on the adhesion of these colon carcinoma cells to the nonspecific ligand PLL (Fig. 1b and results not shown), confirming again the integrin-dependence of the CD9-mediated effects on cell adhesion.
Several other anti-CD9 mAbs generated in our laboratories (VJ1/10, PAINS-10, PAINS-22 and PAINS-26) (data not shown) also enhanced cell adhesion to Ln and Fn similarly to VJ1/20 mAb, indicating that their effects can be attributed to the CD9 molecule and not to a particularity of a specific mAb. The specific involvement of the CD9 molecule in the observed enhancement of integrin-mediated cell adhesion induced by the anti-CD9 mAbs is further demonstrated by the complete lack of effect of these anti-CD9 mAbs on the adhesion of the CD9-negative parental Colo320 cells (Fig. 1b), ruling out the possibility that the observed effects on cell adhesion could be attributed to any potential contaminants present in the preparations of antibodies.
Increased integrin-mediated cell adhesion can result from either enhanced affinity or induction of clustering on the cell surface. We were interested in defining which of these two basic mechanisms were relevant in the observed augmentation of adhesion of colon carcinoma cells induced by anti-CD9 antibodies. Changes in integrin affinity were probed with the anti-β1 integrin activation reporter HUTS-21 mAb.19, 28 As shown in Figure 1c, treatment of BCS-TC2.2 cells with PAINS-13 or VJ1/20 mAbs had no effect on the expression of HUTS-21 epitope, indicating that integrin affinity was not affected. As positive controls for affinity changes, cells were treated with Mn2+ or with the stimulatory TS2/16 mAb, and both treatments clearly induced increased expression of the HUTS-21 epitope.
To assess if the induction of integrin clustering is involved in the increased cell adhesion, BCS-TC2.2 cells were treated with the anti-CD9 antibodies, allowed to adhere to the nonspecific PLL-coated substrate (to avoid any ligand-induced effects on integrins) and clustering of β1 integrin molecules was visualized by immunofluorescence and laser confocal microscopy. As shown in Figure 1d, the anti-CD9 antibodies clearly increased the clustering of integrin molecules on the cell surface. Similar induction of β1 integrin clustering was observed in HT29 and Colo320-CD9 cells treated with CD9 mAbs (not shown).
In many tumor cells, augmented integrin-mediated adhesion is followed by changes in cell polarization and morphology. Thus we assessed whether the observed effects of anti-CD9 antibodies on the integrin-dependent adhesion of BCS-TC2.2, HT29 and Colo320-CD9 cells were followed by phenotypical changes. As shown in Figure 2, colon carcinoma cells dispersely seeded on Ln readily adhered and spread over a period of 2 hr, retaining after this time an overall round or polygonal flat shape. As expected, the stimulatory anti-β1 TS2/16 mAb induced an important enhancement in cellular flattening and spreading in all cells lines but they still conserved their overall round/polygonal morphology. In contrast, the anti-CD9-specific antibodies induced striking morphological changes in these three cell lines, evidenced by the acquisition in most cells of an elongated/fusiform phenotype and the emission of multiple short cytoplasmic projections. Similar changes in morphology were also observed when cells were plated on Fn, although they were less evident than on Ln (data not shown). These morphological changes were completely integrin-dependent, as they did not take place when cells were plated on PLL (Fig. 2). These results obtained with different colon carcinoma cell lines clearly demonstrate that the effects induced by anti-CD9 mAbs on cell morphology are not a peculiarity restricted to a particular carcinoma cell line.
The adhesive and morphological changes caused by anti-CD9 mAbs could be suggestive of some sort of differentiation in the carcinoma cell phenotype. To assess whether this was the case, the effects of anti-CD9 mAbs on the expression of several carcinoma differentiation markers, including E-cadherin, cytokeratins, dipeptidyl peptidase IV (CD26) and alkaline phosphatase activity, were analyzed in BCS-TC2.2 cells, but no significant changes in any of these markers could be detected (data not shown).
CD9 is involved in inhibition of colon carcinoma cell proliferation
We have studied the effects of CD9 in the proliferation of colon carcinoma cells. For this purpose, we first compared the proliferation rates of Colo320 and Colo320-CD9 cells. As shown in Fig. 3a, ectopic expression of CD9 resulted in a marked decrease in cell proliferation which was still evident after 6 days in culture. Consistent with this finding, treatment with the anti-CD9 mAb VJ1/20 (and other anti-CD9 mAbs, not shown) also inhibited the proliferation of Colo320-CD9 cells as well as that of BCS-TC2.2 and HT29 cells (Fig. 3b). These inhibitory effects on cell proliferation were evident after 24 hr of mAb treatment and also at longer culture times (results not shown). Noteworthy, none of the anti-CD9 mAbs exerted any inhibitory effects on the proliferation of CD9-negative parental Colo320 cells (Fig. 3b), again clearly demonstrating the implication of the CD9 molecule in this process. As expected, the conformation-specific anti-CD9 PAINS-13 mAb also inhibited proliferation when cells were plated on Ln, but only marginally on Fn, consistently with the observation that expression of PAINS-13 epitope on CD9 depends upon interaction of this tetraspanin with the integrin α6β1.14 The TS2/16 or TP1/8 mAbs had no significant effect on the proliferation of the different colon carcinoma cells. Similar inhibitory effects by anti-CD9 mAbs on these three colon carcinoma cell lines were observed using alternative proliferation assays, such as CFSE or MTT cell labeling or direct cell counting in Neubauer chamber (data not shown). As shown in Figure 3c, the antiproliferative effect of anti-CD9 mAb VJ1/20 (and of other anti-CD9 mAbs, not shown) on colon carcinoma cells was dose-dependent. Interestingly, the inhibitory effect of the anti-CD9 mAb VJ1/20 (and of other anti-CD9 mAbs, not shown) on proliferation could also be observed when colon carcinoma cells were pretreated with antibody for a short period of time (1 hr) and then plated on Fn in the absence of mAb (Fig. 3d), indicating that the antiproliferative signal of CD9 specific mAbs is delivered at an early stage.
Inhibition of cell growth can result either from a slower progression through the cell cycle or from an increase in the rate of apoptosis; thus, we also assessed the possible effects of anti-CD9 mAbs on these processes. As shown in Figure 4a, incubation of BCS-TC2.2 cells plated on Ln with different anti-CD9 antibodies for 24 hr did not induce any detectable changes in the percentage of cells found in the different stages of cell cycle, as evidenced by flow cytometric analysis of the DNA content. Longer incubation times up to 120 hr showed also no changes (data not shown). As a control, cell treatment for 2 hr with the bile acid CDCA induced an important increment in the hypoploid cell population, clearly indicating cell apoptosis.
Direct analysis of cell apoptosis was also performed by annexin-V–FITC and propidium iodide double staining of BCS-TC2.2 cells. Figure 4b shows that treatment with anti-CD9 mAbs did not induce any significant increase in the percentage of cells undergoing apoptosis over the control or TS2/16 mAb-treated cells. We also confirmed that the anti-CD9 mAbs were not inducing necrosis of colon carcinoma cells by measuring the LDH leakage from the culture supernatants. No significant LDH activity was detected in the supernatants of either untreated or anti-CD9-treated cells (24–72 hr), ruling out that the observed inhibition of cell proliferation by anti-CD9 mAbs could be attributable to the induction of nonapoptotic cell death (not shown).
Ectopic expression of CD9 and anti-CD9 antibodies inhibit the in vivo tumorigenic capacity of colon carcinoma cells
All the data presented regarding the in vitro effects of CD9 ectopic expression as well as those of anti-CD9 mAbs on the adhesion and proliferation of colon carcinoma cells pointed out to the possible involvement of the CD9 tetraspanin in the tumorigenic process itself. Accordingly, we decided to analyze and compare the in vivo effects of the anti-CD9 mAbs with those of other control antibodies on tumor development in a nude mice tumorigenesis model.
For such purpose, we first analyzed the effect of the direct intratumoral injections of the different mAbs on the in vivo growth of BCS-TC2.2 tumors. Mice bearing tumors were injected PBS or TP1/8 mAb (as controls) or the corresponding antibody every 2–3 days and tumor size was measured at each time point. Figure 5a shows the time-dependent growth of the tumors in relative units up to 18 days after the initial antibody injection. No significant variations in tumor growth were observed when mice received injections of either TS2/16 or TP1/8 mAbs compared with the PBS control group. On the other hand, intratumoral injection of VJ1/20 or PAINS-13 mAbs significantly reduced tumor growth, being the effect more evident for the later. The relative tumor growth per day decreases from around 0.09–0.1 in the controls (PBS and TP1/8 mAb) and TS2/16 to 0.07 and 0.05 in VJ1/20 and PAINS-13 mAbs, respectively. The differences in the overall behavior between the anti-CD9 mAbs and controls are statistically highly significant (p < 0.001). Differences with the controls were significant from day 4 in PAINS-13 mAb and from day 13 in VJ1/20 mAb. Representative tumors after 18 days treatment are shown in Figure 5b, where the reduction in tumor size induced by VJ1/20 or PAINS-13 mAbs can be clearly observed. Figure 5c shows the statistical differences between the experimental groups at day 9 and 18 after injection. Additionally, we have studied a second group of animals (n = 6) in which both flanks were injected with BCS-TC2.2 cells. Mice bearing tumors in both flanks were then injected with PBS in one flank and with PAINS-13 mAb in the other. Under these conditions, similar differences in tumor growth were detected between the groups (data not shown). Thus, the decrease in tumor mass in vivo is not the result of systemic absorption and delivery of PAINS-13 mAb, but instead is a direct intratumor effect.
When BCS-TC2.2 cells were treated ex vivo for 1 hr with PAINS-13 mAb prior to their subcutaneous injection into the nude mice, differences in the tumor incidence, latency time and growth rate could be observed with a dependence on the mAb concentration used (Fig. 5d). Tumor incidence declined dramatically with PAINS-13 mAb treatment from 100% (6/6) in untreated cells to 67% and 33% (only 4/6 or 2/6 mice developed tumors) when BCS-TC2.2 cells were treated with 1 and 2 mg/ml PAINS-13 mAb, respectively. Tumors formed in the control group presented a latency time of 14 ± 2 days whereas those from PAINS-13 mAb group required 22 ± 1 days. Additionally, the use of a higher antibody concentration induces a slower tumor growth in the two mice that finally developed tumors. All these results demonstrate that anti-CD9 antibodies dramatically reduce the tumorigenicity of BCS-TC2.2 cells in a dose-dependent manner.
Taken together, these in vivo results show that anti-CD9 antibodies dramatically reduce the tumorigenicity of BCS-TC2.2 cells. To further establish the suppressor role of CD9 in colon carcinoma tumorigenicity, we compared the tumorigenic capacity of parental Colo320 (CD9-negative) and stably CD9-transfected Colo320 (CD9-positive) cells in nude mice. As shown in Figure 6a, expression of CD9 clearly reduced the tumorigenic capacity of these carcinoma cells, as determined by the increased latency time (5 ± 1 vs. 8 ± 1 days), reduced incidence (8/8 vs. 4/8 tumors) and a modest but statistically significant reduction in mean tumor size (9.95 ± 0.67 vs. 8.53 ± 1.1 mm at 14 days after injection), although no significant variation in the tumor growth rate was observed (1.1 ± 0.1 vs. 1.2 ± 0.1 mm/day). We ruled out the possibility that loss of CD9 expression during the development of the tumors generated by CD9-transfected Colo320 cells could account for the similar rate of tumor growth observed, as CD9 was readily detected by immunoblotting in extracts of these tumors but not in tumors derived from parental CD9-negative Colo320 cells (Fig. 6b).
TNF-α is involved in morphological changes and in inhibition of proliferation of colon carcinoma cells treated with CD9-specific antibodies
Several soluble mediators, including TNF-α, have been reported to exert important differentiating and antiproliferative actions on different types of epithelial and carcinoma tumor cells.29, 30, 31 We were therefore interested in determining whether this cytokine could be a mediator of the observed effects of anti-CD9 mAbs on colon carcinoma cell morphology and proliferation. For this purpose, we first assessed whether exogenous TNF-α was able to induce changes in the morphology and inhibition of proliferation similar to those caused by anti-CD9 mAbs. Treatment of different colon carcinoma cells with TNF-α resulted in the acquisition of an elongated/fusiform cell phenotype (Fig. 7a) and inhibition of cell proliferation (Fig. 7b) and these effects were very similar to those induced by anti-CD9 mAbs (respectively shown in Figs. 2 and 3b). We have also established that this TNF-α-induced inhibition of cell proliferation was not due to induction of apoptosis—as determined by propidium iodide and Annexin-V staining—or necrosis—as determined by LDH activity from culture supernatants (data not shown). These results suggested that TNF-α could indeed be a mediator of the effects exerted by anti-CD9 mAbs in these carcinoma cells. If this was the case, then we could expect these antibodies to induce changes in the levels of either secreted or membrane-bound TNF-α on BCS-TC2.2 cells. Using a highly sensitive biological assay based on the TNF-α-dependent cytotoxicity of the fibroblastic murine L929 cell line26, 27 we have determined that treatment of BCS-TC2.2 cells with anti-CD9 mAbs did not affect the amount of secreted TNF-α, which was in all cases below the 6 pg/ml detection limit of this assay (not shown), but resulted in a clear increase in membrane TNF-α, as evidenced by the increased cytotoxicity of L929 cells (Fig. 7c). This increase in membrane TNF-α on BCS-TC2.2 cells after treatment with anti-CD9 mAbs was also confirmed by flow cytometry detection (Fig. 7d, upper panel). Similar increases in membrane TNF-α in Colo320-CD9 and HT29 cells were also detected after treatment with anti-CD9 mAbs (data not shown). Interestingly, expression of membrane TNF-α on Colo320 cells also increased after ectopic expression of CD9 in Colo320 cells (Fig. 7d, lower panel).
The role of TNF-α as a mediator of the antiproliferative effects exerted by CD9 was demonstrated by using two specific inhibitors, infliximab and etanercept, which have been demonstrated to bind to and interfere with the cellular effects mediated by both the secreted and the transmembrane forms of TNF-α.32, 33 Infliximab is a humanized anti-TNF-α specific blocking antibody and etanercept is a soluble form of a fusion protein containing the extracellular region of human p75 TNF-RII. As shown in Figure 7e (left panel), treatment of BCS-TC2.2 cells with these anti-TNF-α agents prevented the inhibition of cell proliferation caused by the CD9 mAbs VJ1/20 and PAINS-13. A similar preventive effect was also observed with Colo320-CD9 and HT29 cells (data not shown). These blockers of TNF-α action also reversed the reduced proliferation observed in Colo320 cells after ectopic expression of CD9 (Fig. 7e, right panel). Taken together our results place the transmembrane form of TNF-α as a mediator of the observed effects exerted by CD9 on the proliferation of colon carcinoma cells.
The level of expression of the tetraspanin CD9 in primary tumor cells has been inversely correlated with their metastatic potential and with patient survival rates in different types of cancer.5, 7, 10, 34 The involvement of the CD9 molecule in metastasis has been further inferred from the inhibition of tumor cell motility and migration caused by CD9 specific mAbs or by overexpression of this tetraspanin.5, 7, 9, 10 While this implication of CD9 in metastasis has been documented in numerous studies and is fairly well established, much less attention has been however devoted to the possible involvement of this tetraspanin in the early stages of tumor formation, i.e. the development of a primary tumor itself—the tumorigenic process. In the present study we have gathered experimental evidence on the implication of CD9 in carcinoma tumorigenesis. For this purpose, we have employed different human colon carcinoma cell lines which differ in their level of CD9 surface expression and have analyzed the effects of ectopic CD9 expression and anti-CD9-specific antibodies on their behavior.
Like other tetraspanins, CD9 is known to be engaged in the formation of plasma membrane protein complexes containing several Ln- and Fn-binding integrins.3, 35 The ectopic expression of CD9 in Colo320 carcinoma cells allowed us to analyze the effect of the presence of the CD9 molecule in terms of cellular adhesion to different proteins of the extracellular matrix. In this regard, stably transfected Colo320-CD9 cells displayed an increased adhesive capacity to Ln and Fn compared to the untransfected Colo320-CD9-negative cells. Consistently with this, we have also observed that several CD9-specific mAbs induced increased colon carcinoma cell adhesion, pointing out to an agonistic effect of these antibodies on CD9-influenced adhesion. As these enhancing effects on adhesion were only observed on integrin–ligand proteins, but not on PLL, they must be attributed to the specific associations between CD9 and integrins, as have also been reported by other groups.36, 37, 38 In the context of colon carcinoma cells, we have further established using the activation-reporter mAb HUTS-2119 that the mechanism by which CD9 mAbs cause the enhancement in cell adhesion is mainly mediated by integrin aggregation rather than altered integrin affinity. In this regard, it has been recently reported that the CD9 molecule controls the redistribution and clustering of integrins, including α6β1, that are relevant in other biological processes such as gamete fusion.39 Our findings are in agreement with those reported for other tetraspanins which support the general notion that these molecules do not exert their functional cellular effects through direct alteration of integrin conformation or affinity for ligand but rather through integrin-dependent postadhesion events such as changes in cell morphology, spreading and signaling.3 However, a recent report proposes that the tetraspanin CD151 influences the laminin-binding capacity of integrin α3β1 through a mechanism involving stabilization of its active—high affinity—conformation.40 Therefore, it is possible that different mechanisms are responsible for the modulation of the adhesive capacities of specific integrins by their tetraspanin partners, which may also depend on the cell context.
We have found that, in addition to their effects on cell adhesion, anti-CD9 mAbs also induced important changes in the morphology of different colon carcinoma cells characterized by the acquisition of an elongated cell phenotype and the emission of multiple short cytoplasmic projections, whereas cell spreading was not significantly augmented. These changes could not be merely attributed to the increased activation of integrins and augmented cell adhesion, since the stimulatory anti-β1 integrin mAb TS2/16, which also increased adhesion of these cells to Ln and Fn, caused an important enhancement in cell spreading but did not induce such phenotypic alterations. The morphological changes caused by anti-CD9 mAbs may indicate induction of a transition into a more differentiated carcinoma cell phenotype. Therefore, we analyzed changes in the expression of several widely used markers for colon carcinoma differentiation, including E-cadherin, cytokeratins, dipeptidyl peptidase IV (CD26) and alkaline phosphatase activity, but could not detect significant changes in any of them, indicating that the observed morphological changes induced by anti-CD9 mAbs expression cannot be correlated with induction of differentiation or epithelial–mesenchymal transition.
Changes in the growth and differentiation of colonic epithelial cells are indeed fundamental to the pathogenesis of colorectal carcinogenesis.41, 42, 43 Our results clearly show that the ectopic expression of CD9 in colon carcinoma cells lacking this tetraspanin results in a marked decrease in their proliferative capacity. In agreement with this, anti-CD9 mAbs also exert an important inhibitory effect on the proliferation of colon carcinoma cells, again pointing out to an agonistic role of these antibodies on CD9-mediated functional effects. A plausible explanation for this agonistic effect is that upon antibody binding to CD9 molecules, the associations of this tetraspanin with other membrane or intracellular proteins which are mediating the effects on proliferation and adhesion are reinforced or stabilized. These results clearly establish the involvement of this tetraspanin in the growth of these tumor cells. Inhibition of cell growth can result either from a slower progression through the cell cycle29 or from an increase in the rate of apoptosis and/or necrosis. Therefore, we assessed the possible effects of anti-CD9 mAbs on the apoptosis and necrosis of colon carcinoma cells, but no significant alteration in these two processes was detected. Thus, the inhibition of cell proliferation in colon carcinoma cells caused by anti-CD9 mAbs is not mediated by an increase in apoptosis or necrosis, and point to a slower progression through the cell cycle without an arrest at a specific cell cycle phase. It has been recently reported that an anti-CD9 mAb (ALB6 mAb) induced morphological changes and inhibited the proliferation of several types of gastrointestinal cancer cells through induction of apoptosis and activation of caspase-3, JNK and p38 MAPKs.44 Although these inhibitory effects of the anti-CD9 ALB6 mAb on cancer cell proliferation are similar to those reported here, we do not observe however any induction of apoptosis in colon carcinoma cells by our anti-CD9 mAbs, which either suggests that the mechanisms involved in the inhibition of proliferation may differ depending on the carcinoma cell type under study or that distinct effects are induced by different anti-CD9 mAbs. Additionally, ectopic expression of CD9 in small-cell lung carcinoma cells also results in inhibition of their proliferation, indicating that the role of CD9 in the regulation of tumor cell growth is not restricted to gastrointestinal carcinomas but seems to be relevant in other types of carcinomas.45 On the other hand, CD9 is not the only tetraspanin that has been implicated in proliferation of tumor cells. In this regard, it is worth to indicate that the tetraspanin CD81 was originally identified as the target of an antiproliferative mAb on human B leukemic cells.46, 47
Taken collectively, all the in vitro data on the induction of adhesion and morphological changes, and the inhibition of cell proliferation, indicate that the tetraspanin CD9 negatively affects the growth capacity of colon carcinoma cells. We assessed directly this possibility in the nude mice model of in vivo tumorigenesis. As could be anticipated from our in vitro data showing that a short (1 hr) pretreatment of BCS-TC2.2 cells before their seeding on Fn resulted in decreased proliferative capacity, the treatment of these cells prior to their subcutaneous injection into mice with anti-CD9 mAbs resulted in an important reduction of their tumorigenicity, evidenced by the decreased incidence, prolonged latency and reduced tumor size. Similar inhibitory effects on the tumorigenicity of these carcinoma cells were observed when tumors were first allowed to develop and grow subcutaneously and then the anti-CD9 mAbs were repeatedly injected intratumorally. Although tumor regression did not occur, tumor growth was significantly delayed probably as a consequence of the inhibition of cell proliferation. Interestingly, the mAb PAINS-13 was more effective at inhibiting tumor growth than the other anti-CD9 mAb used (VJ1/20) in these in vivo experiments despite the fact that expression of the PAINS-13 epitope on BCS-TC2.2 cells is much lower. As we have reported previously,14 the CD9 epitope recognized by mAb PAINS-13 is preferentially imposed upon association of this tetraspanin with integrin α6β1, a cellular receptor for Ln. Accordingly, the functional effects exerted by this mAb on cell adhesion, spreading, migration and morphogenetic processes in endothelial cells are more prominently observed when they are plated on laminin, in clear difference to the effect of other anti-CD9 mAbs (like mAb VJ1/20) with conformation-independent accessibility to epitopes on CD9. In agreement with our previous data with endothelial cells, we show here that the mAb PAINS-13 was again more efficient at enhancing the adhesion to Ln of colon carcinoma cells compared to mAb VJ1/20. In this context, it is well established that laminin is a major component of the extracellular matrix present in the stroma of colorectal carcinomas48, 49, 50 which would account for the potent observed antitumoral effect of mAb PAINS-13 in these in vivo assays. On the other hand, there seems to be no direct correlation between the effects of CD9 mAbs on tumor growth in vivo and their effects on proliferation in vitro since mAb PAINS-13 has a more pronounced effect than mAb VJ1/20 in vivo but a lower effect in vitro. A plausible explanation for this lack of correlation could be that subtle changes induced by each particular anti-CD9 mAb in the adhesive properties of carcinoma cells at the initial steps of in vivo tumorigenesis assays, for instance at the engraftment stage, could be decisive in determining later the outcome of tumor growth. An alternative possibility to explain this lack of complete correlation could be the involvement of distinct isotype-dependent immune antitumoral responses that might be elicited in vivo after injection of the different anti-CD9 mAbs. The inhibitory effects of anti-CD9 mAbs in vivo strongly suggested a supressor role for this tetraspanin in carcinoma tumorigenesis, which was confirmed by the reduced tumorigenic capacity displayed by stably CD9-transfected Colo320 cells compared to that of parental CD9-negative Colo320 cells. Despite the important reduction in tumor incidence and the increased latency time observed with CD9-positive cells compared to CD9-negative cells, only a modest though statistically significant reduction in the final tumor size was detected. These results suggest that the inhibitory effects of CD9 expression mainly concern the initial steps of tumor development. Since in these in vivo experiments with parental and CD9-transfected Colo320 cells no mAb was used, an involvement of an isotype-dependent immune response in the observed negative role of CD9 expression in tumor development could be ruled out.
Taking into account this negative role of CD9 in tumorigenesis, human colon carcinomas would be expected to be more frequently negative for expression of CD9 than CD9-positive. In one study of 82 surgical specimens of human primary colorectal carcinomas, Mori et al. reported that CD9 mRNA expression was positive in 56 cases and negative in 26. Importantly, the CD9-positive cases showed a significantly higher survival rate and a lower frequency of metastasis as well as a much better prognosis.5 However, in another study of 146 colon cancer patients by Hashida et al., the percentage of CD9-positive tumors was lower (43.8%) than that of CD9-negative tumors (56.2%), though again the overall survival rate of patients with CD9-positive tumors was significantly higher than that of patients with CD9-negative tumors (89.8 vs. 50.8 %, p < 0.001).51 Collectively, these studies and our data support a negative role of CD9 in colorectal cancer both at the initial stages of primary tumor development and in the subsequent metastatic dissemination.
We have shown that exogenous TNF-α was able to induce the acquisition of an elongated cell phenotype and inhibition of cell proliferation. These effects caused by TNF-α were very similar to those induced by the anti-CD9 antibodies VJ1/20 and PAINS-13, indicating that TNF-α could indeed be a mediator of the effects exerted by CD9 mAbs in these carcinoma cells. In this regard, we observed a clear increase in the expression of the membrane form of TNF-α in anti-CD9 mAbs-treated colon carcinoma cells. Further support for the involvement of TNF-α as a mediator of the observed effects of anti-CD9 mAbs on colon carcinoma cells came from the use of two specific blockers of this cytokine (infliximab and etanercept) that are currently used in the therapy of several human inflammatory diseases including rheumatoid arthritis, Crohn's disease and psoriasis.52 Blockade of TNF-α with these compounds partially prevented the antiproliferative effects on colon carcinoma cells derived from ectopic expression of CD9 or treatment with anti-CD9 mAbs. Taken together our results place the cytokine TNF-α as a mediator of the observed effects exerted by CD9 mAbs on morphology and inhibition of proliferation of BCS-TC2.2 cells. In agreement with our data, a recent report has shown that, in CD9-knockout mice, TNF-α production is delayed and reduced compared to wild-type animals.53 Furthermore, a very recent paper demonstrates the inhibitory effect of transmembrane TNF-α in hepatic carcinoma tumorigenesis,27 strongly supporting the findings presented in this work with colon carcinomas. It has been demonstrated that transmembrane TNF-α is biologically active and affects proliferation through interaction with the p75 TNFR-II in a juxtacrine cell-to-cell fashion, but in addition transmembrane TNF-α is itself capable to elicit intracellular signaling impinging on proliferation, a process known as reverse signaling.33, 54 On the other hand, the precise mechanism by which anti-CD9 mAbs influence the levels of membrane TNF-α needs to be elucidated. By quantitative RT-PCR analysis we have not detected any significant differences in TNF-α mRNA between untreated and anti-CD9 mAb-treated colon carcinoma cells (data not shown), indicating that the effects of anti-CD9 mAbs are not exerted at a transcriptional level. Anti-CD9 mAbs could mediate intracellular signaling affecting posttranscriptional events, such as transport or recycling of this cytokine, leading to elevated amounts on the surface. In addition, CD9 could be affecting the stability of the transmembrane form of TNF-α perhaps by inhibiting the activity of associated ADAMs, which are responsible for the processing and secretion of TNF-α, as has been recently proposed in different models.55, 56 Further studies will be required to distinguish between these possibilities.
The results herein presented clearly establish for the first time that in addition to its role as a metastasis-supressor molecule the tetraspanin CD9 acts as a negative regulator of colon carcinoma tumorigenesis.
We are grateful to Mariano Vitón and Marta Ramírez (Hospital de la Princesa, Madrid) for their continuous technical assistance, to Dr. José Alberto García-Sanz (CIB-CSIC, Madrid) for his assistance with the proliferation studies and to Dr José Luis Rodríguez-Fernández (CIB-CSIC, Madrid) for his critical reading of this manuscript. SO is the recipient of a I3P predoctoral Fellowship from Consejo Superior de Investigaciones Científicas (CSIC).
- 1Tetraspanins and malignancy. Expert Rev Mol Med 2001; 31: 1–17., , , .