|Home | About | Journals | Submit | Contact Us | Français|
Polysialic acid (PSA), a carbohydrate polymer attached to the neural cell adhesion molecule (NCAM), promotes neural plasticity and tumor malignancy, but its mode of action is controversial. Here we establish that PSA controls tumor cell growth and differentiation by interfering with NCAM signaling at cell-cell contacts. Interactions between cells with different PSA and NCAM expression profiles were initiated by enzymatic removal of PSA and by ectopic expression of NCAM or PSA-NCAM. Removal of PSA from the cell surface led to reduced proliferation and activated extracellular signal-regulated kinase (ERK), inducing enhanced survival and neuronal differentiation of neuroblastoma cells. Blocking with an NCAM-specific peptide prevented these effects. Combinatorial transinteraction studies with cells and membranes with different PSA and NCAM phenotypes revealed that heterophilic NCAM binding mimics the cellular responses to PSA removal. In conclusion, our data demonstrate that PSA masks heterophilic NCAM signals, having a direct impact on tumor cell growth. This provides a mechanism for how PSA may promote the genesis and progression of highly aggressive PSA-NCAM-positive tumors.
The neural cell adhesion molecule (NCAM) exhibits high structural diversity due to alternative splicing and dynamically regulated posttranslational modifications (5, 6). Initially identified as a cell adhesion molecule expressing homophilic binding properties (20), NCAM was later shown to exert heterophilic cis and trans binding interactions with molecules of different classes, such as cell adhesion molecule L1 (21), members of the fibroblast growth factor receptor family (3, 42), extracellular matrix components (4), and perhaps others. In accordance with the protein's complexity, it has been implicated in a multitude of cellular functions, mainly during neural development and plasticity (reviewed in reference 5) but also in oncogenesis (3, 36).
The most prominent and unique posttranslational modification of NCAM is polysialic acid (PSA), a homopolymer of α-2,8-linked sialic acid residues which is added to specific N-glycan attachment sites in the fifth immunoglobulinlike domain of NCAM (30). PSA is abundantly expressed during embryonic development and downregulated in the course of maturation and differentiation (41). In the developing nervous system, PSA-NCAM has been shown to promote plasticity of cell-cell interactions during cell migration and neurite outgrowth (9, 41). In the adult mammalian brain, PSA appears to be involved in persistent neurogenesis and some forms of synaptic plasticity (9), and its expression under different pathological conditions implies a role in neural regeneration and repair (2, 32). As an oncodevelopmental antigen, PSA is reexpressed during progression of several malignant human tumors, including small cell lung carcinoma, Wilms' tumor, neuroblastoma, and rhabdomyosarcoma (11, 14, 17). In these tumors, polysialylation of NCAM appears to increase the metastatic potential and has been correlated with tumor progression and a poor prognosis (7, 8, 14, 16, 17, 48).
Despite the abundant evidence that polysialic acid is critically involved in neural development and tumor malignancy, its mode of action on the cellular level is still unclear. According to the prevailing model, steric inhibition of membrane-membrane apposition by PSA causes a general attenuation of cell adhesion (41), and this antiadhesive effect of PSA appears to be independent of NCAM-mediated interactions (13). Besides modulating cell adhesion, PSA may act as a receptor of secreted factors (22), and only recently it was shown that PSA is needed for the adequate sensitivity of neurons to the brain-derived neurotrophic factor (BDNF) (49). In contrast to the antiadhesive properties and the possible receptor functions of PSA, its potential impact on NCAM signals is unresolved. As demonstrated previously, enzymatic removal of PSA induces marked inhibition of cell growth in human SH-SY5Y neuroblastoma cells (19).
Using a panel of PSA-NCAM expressing neuroblastoma and rhabdomyosarcoma cell lines, the present study identifies PSA as a negative regulator of heterophilic NCAM interactions at cell-cell contacts. Removal of PSA releases NCAM signals, inducing growth inhibition as well as mitogen-activated protein kinase (MAPK)-dependent survival and differentiation. The present data demonstrate for the first time that expression and downregulation of PSA are decisive for NCAM-mediated regulation of tumor cell growth.
The following monoclonal (MAb) and polyclonal (PAb) antibodies were used: extracellular signal-regulated kinase (ERK) 1- and 2-specific rabbit PAb; dually phosphorylated ERK1/2-specific mouse MAb E10 (9106) or rabbit PAb (9101; New England Biolabs); neurofilament-specific mouse MAb 2F11, recognizing the phosphorylated form of the 200-kDa and 68-kDa subunits (Neomarkers); mouse MAb 14G2a, specific for ganglioside GD2 (kindly provided by R. Handgretinger); bromodeoxyuridine (BrdU)-specific rabbit immunoglobulin G (IgG) (Chemicon); normal mouse or rabbit IgG; alkaline phosphatase- and peroxidase-conjugated goat anti-mouse IgG; peroxidase-conjugated goat anti-rabbit IgG; 10 nm gold-conjugated goat anti-mouse IgG (Sigma); Texas Red-conjugated rabbit anti-mouse IgG1; and fluorescein isothiocyanate-conjugated rabbit anti-mouse IgG2 (all from Biotrend), and CY2-conjugated goat anti-rabbit IgG and CY3-conjugated goat anti-mouse IgG (Dianova).
All commercially available antibodies were used according to the recommendations given by the supplier. PSA-specific mouse MAb 735 (12) and MAb 123C3 (29) (clone kindly provided by R. Michalides), reactive with all isoforms of human NCAM, were used after affinity purification on protein G-Sepharose (Pharmacia) and applied to live cells at 5 μg/ml or for immunoblotting, enzyme-linked immunosorbent assay (ELISA), immunocytochemistry, and immunofluorescence as described (19, 45). N-Acetylneuraminic acid and colominic acid were purchased from Sigma and used at 50 μg/ml. C3d, a synthetic dendrimeric undecapeptide which binds to the Ig1 module of NCAM and its inactive variant C3d2Ala (39) were kindly provided by E. Bock and N. Pedersen. The MEK inhibitor PD98059 was from Alexis, fibroblast growth factor 2 (basic fibroblast growth factor) was from Merck, BDNF was from Calbiochem, and nerve growth factor was from Biomol.
To generate NCAM- and PSA-NCAM-positive LS cell lines, the vectors pAM1, encoding full-length human NCAM-140 in pCDM8 (10), and pcDNAI-PST, kindly provided by M. Fukuda, encoding human full-length ST8Sia IV in pcDNAI (33), were used. pEGFP-C1 (BD Clontech) was used as the cotransfected selection marker (neomycin/G418) and to obtain enhanced green fluorescent protein (EGFP) expression in the cytosol. Recombinant endoneuraminidase NE (endo-NE), specifically degrading PSA, was isolated as described (15) and used in the cell culture medium at 6 or 60 ng/ml to remove PSA from the cell surface (19). For inactivation, endo-NE was heated for 10 min at 60°C.
The human neuroblastoma cell lines SH-SY5Y (subclone of SK-N-SH, ATCC CRL-2266), Kelly (ECACC 92110411), LS (40), LAN-5 (ICLC HTL-96022), and rhabdomyosarcoma cell line TE-671 (ATCC CRL-7774) were used. Cells were cultured at 37°C and 9% CO2 in Dulbecco's modified Eagle's medium-Ham's F12 medium containing 10% (vol/vol) heat-inactivated fetal bovine serum (Biochrom), 2 mM glutamate, 100 units of penicillin per ml, and 100 μg of streptomycin per ml. Media were changed every 2-day, and cells were replated before confluency. Experiments were conducted with cells at densities between 2.5 and 5 × 104 cells/cm2. To avoid effects of serum deprivation, great care was taken in all experiments that the different experimental groups received the same treatment with respect to medium changes, i.e., supplied with fresh serum. For short-term incubations with endo-NE in serum-free medium, cells were kept without serum for 1 h before changing to serum-free medium containing the reagent.
For stable transfections, 3.5 × 106 SH-SY5Y or 1.5 × 106 LS cells were plated in a 60-mm dish in culture medium containing 10% fetal calf serum. After 24 h cells were transfected with 20 μl of Effectene corresponding to the manufacturer's instructions (Qiagen) in culture medium containing 5% fetal calf serum. SH-SY5Y cells were transfected with 1 μg of pEGFP-C1. LS cells were cotransfected with 1 μg of pAM1 and 0.1 μg of pEGFP-C1 for expression of NCAM-140 or with 0.5 μg of pAM1, 0.5 μg of pcDNAI-PST, and 0.05 μg of pEGFP-C1 for expression of PSA, or with 1 μg of pCDM8 and 0.1 μg of pEGFPC1 as a control (mock transfection); 24 h later the medium was replaced by culture medium, and 48 h later the cells were passaged in 12-well plates with culture medium complemented by 400 μg of potent G418-sulfate per ml (Calbiochem). Transfected cells were subcloned by limited dilution to obtain single-cell clonal lines and checked by immunocytochemistry for clones with a homogenous cell surface immunoreactivity for PSA or NCAM (45). The EGFP-positive SH-SY5Ycells are designated SH-SY5YEGFP, and the mock-transfected, NCAM- or NCAM-PSA-positive LS cells are designated LSmock, LSAM1, and LSAM1PST, respectively.
The relative amounts of PSA or NCAM on the cell surface were determined by the cellular ELISA procedure described previously (44, 45). Briefly, cells grown in 96-well plates were fixed with 3.8% paraformaldehyde-phosphate-buffered saline, blocked with 2% (wt/vol) bovine serum albumin-phosphate-buffered saline for 2 h, and incubated with PSA-specific MAb 735 (0.3 μg/ml) or NCAM-specific MAb 123C3 (1 μg/ml), followed by detection with alkaline phosphatase-labeled anti-mouse IgG and p-nitrophenylphosphate.
For immunoblotting, cells were washed briefly with ice-cold phosphate-buffered saline and harvested with a cell scraper in ice-cold lysis buffer consisting of 1% Brij 96, 150 mM NaCl, 10 mM Tris-HCl, pH 7.4, 10 mM NaF, 1 mM EDTA, 1 mM EGTA, 1 mM Na3VO4, 1 mM PMSF, 10 μg of leupeptin per ml, and 10 μg of aprotinin per ml. After 10 min of incubation on ice, the lysates were clarified by centrifugation at 20,000 × g for 15 min, and the protein concentration was determined by the Bio-Rad protein assay. If not specified otherwise, 20 μg of protein was separated on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (10% polyacrylamide gels) and transferred to nitrocellulose filters (Roth). Equal loading of proteins was confirmed by Ponceau S staining, and proteins of interest were visualized with specific antibodies and the enhanced chemiluminescence detection system (Amersham) with Kodak Biomax MS films. The intensity of enhanced chemiluminescence bands was analyzed as mean grey value by computerized densitometric scanning and ScionImage software (Scion Corporation). For reprobing of membranes, antibodies were stripped by incubation with 100 mM 2-mercaptoethanol-70 mM SDS in 62.5 mM Tris-HCl, pH 6.7, at 70°C for 45 min.
Crude membrane fractions from LSmock, LSAM1, and LSPSTAM1 were homogenized in 10 mM NaHCO3 (pH 8.0) with 1 mM CaCl2 and 1 mM MgCl2. Cells were passed 10 times through a 20-gauge needle, centrifuged at 30,000 × g for 10 min, and the resulting pellets were resuspended in serum-supplemented cell culture medium. This suspension was added to cultured cells in a 1:1 ratio of extracted to living cells.
Cells were fixed for 30 min with 3.8% paraformaldehyde in phosphate buffer, blocked with 2% bovine serum albumin-phosphate buffer for 2 h, and incubated with primary antibodies overnight at 4°C. For immunofluorescence detection of neurofilament, cells were solubilized with 0.1% Triton X-100-phosphate buffer for 15 min before blocking and a CY3-conjugated secondary antibody was used. PSA and NCAM immunostaining with alkaline phosphatase-coupled secondary antibodies and 5-bromo-4-chloro-3-indolylphosphate-nitoblue tetrazolium as well as confocal immunofluorescence with Texas Red and fluorescein isothiocyanate-conjugated secondary antibodies were carried out as described (19, 45).
For ultrastructural analysis, pre-embedding immunolabeling of polysialic acid on the cell surface was performed as for immunocytochemistry with 10-nm-gold-labeled secondary antibody. As control, primary antibodies were omitted. Cells were postfixed in the culture plate with 1% glutaraldehyde-phosphate buffer (5 min, room temperature) and 1% OsO4 (10 min, room temperature), washed, and subjected to a graded series of 30, 50, and 70% ethanol (19 min each), counterstained in 1% uranyl acetate-70% ethanol (15 min), dehydrated three times with 100% ethanol, incubated 30 min in Araldite (Plano)-ethanol, 1:1, and embedded in Araldite. Then 500-nm-thick horizontal sections were prepared with an Ultracut S (Reichert) and mounted on pioloform-coated copper grids. Imaging of thick horizontal sections with energy-filtering transmission electron microscopy, allowing the ultrastructural analysis of large areas of the cell surface, was performed as described (26) with the electron microscope CEM902 (Zeiss) equipped with a Henry Castaign energy filter (prism-mirror-prism spectrometer).
Cell growth or survival was assessed by a colorimetric tetrazolium-formazan assay (Cell Proliferation Assay; Promega), and rates of apoptosis were assessed by the quantitative determination of intracellular mono- and oligonucleosomes with a sandwich-ELISA procedure with monoclonal antibodies directed against DNA and histones (Cell Death Detection ELISAplus; Roche). Detached cells were collected by centrifugation of the cell culture supernatant (200 g) and lysed together with the adherent cells. The specific enrichment of intracellular mono- and oligonucleosomes is given as the apoptotic index, calculated from the absorbance of the sample relative to untreated controls (see manufacturer's instructions for details). Both assays were applied to cells seeded in parallel on 96-well plates.
For each treatment, phase contrast images of three wells were captured with an Axiovert 135 microscope (Zeiss) and a charge-coupled device camera, and cells were counted to compare cell numbers with the results of the metabolic assay. Alternatively, cells were fixed as described for immunocytochemistry, washed with phosphate buffer, and mounted in Vectashield containing 4′,6′-diamidino-2-phenylindole (DAPI; Linaris) to analyze the morphology of the nuclei with DAPI fluorescence. In some of the experiments, the rate of proliferation was addressed by incorporation of bromodeoxyuridine (BrdU; Boehringer). Cells were incubated for 16 h with 10 μM BrdU prior to fixation. After incubating with 2 N HCl for 15 min at 37°C and then 0.1 M borate, pH 8.5, for 10 min, BrdU was detected with an anti-BrdU antibody diluted 1:100 (Chemicon).
Neurite outgrowth of SH-SY5YEGFP neuroblastoma cells was addressed in 12-well plates. For treatment with endo-NE, dimethyl sulfoxide, or PD98059, 30,000 SH-SY5YEGFP cells were seeded in the presence of the reagent. For incubation with crude membrane fractions, SH-SY5YEGFP cells were grown for 24 h before the membrane suspension was added. After 48 h of treatment, living SH-SY5YEGFP cells were imaged by EGFP fluorescence with an Axiovert 100 M microscope equipped with LSM (5 Pa), a helium-neon laser, and a Plan-Neofluar 10× objective (Zeiss). From each well, five randomly selected frames were scanned, SH-SY5YEGFP cells were counted, and the length of the longest process per cell was measured by computer-assisted image analysis with the software package Axio Vision 3.0 (Zeiss). Per frame, the percentage of cells with processes longer than 20 μm was calculated relative to the total number of cells.
Differences between two groups were evaluated with Student's t test. With more than two groups to compare, one-way analysis of variance (ANOVA) or ANOVA with repeated measures was applied with Prism (Graphpad Software). Pearson's χ2 test was used to compare BrdU incorporation.
Treatment of living cells with endoneuraminidase NE (endo-NE) (15) degrades PSA with high specificity and leads to a better accessibility of NCAM on the cell surface (19). In NCAM- and PSA-positive neuroblastoma (SH-SY5Y, Kelly, and LAN-5) and rhabdomyosarcoma (TE671) cells (44), endo-NE induced a similar inhibition of cell growth, while growth of the NCAM- and PSA-negative neuroblastoma cell line LS was unaffected (Fig. (Fig.1a).1a). Control incubations with heat-inactivated endo-NE, colominic acid (i.e., soluble PSA from bacteria), or N-acetylneuraminic acid had no effect on cell growth. The endo-NE-induced reduction of cell growth was confirmed by counting the cells in some of the experiments. Detached, dead, or damaged cells were never observed, and DAPI staining revealed no abnormal condensation or fragmentation of nuclei. For SH-SY5Y cells, the absence of apoptosis after endo-NE treatment was confirmed by a cell death ELISA (see Fig. Fig.6),6), and the analysis of BrdU incorporation revealed that the endo-NE-induced reduction of cell growth was associated with a significant decline in proliferation (Fig. (Fig.1b).1b). Thus, the observed growth inhibition was due to reduced rates of proliferation.
In primary cultures of cortical neurons, the expression of PSA has been shown to improve the sensitivity to BDNF, and in contrast to other neurotrophins, application of exogenous BDNF was able to rescue neurons from endo-NE-induced death (49). Because SH-SY5Y cells express low levels of BDNF and TrkB (23), we compared the influence of endo-NE on the mitogenic activities of BDNF, nerve growth factor, and fibroblast growth factor-2. The proliferative response to all three growth factors was drastically reduced in the absence of PSA (Fig. (Fig.2a).2a). This pleiotropic effect argues against a BDNF-specific interaction with PSA but suggests that growth inhibition is a direct result of PSA failure.
We then analyzed whether changes in NCAM binding abilities may be responsible for the effect of endo-NE. In a first experiment, the MAb 123C3 (29) was used to interfere with NCAM binding. The PSA-specific MAb 735 and MAb 14G2a (31), directed against the ganglioside GD2, served as controls. All three reagents produced an 80% growth reduction, indicating a strong but nonspecific cytotoxic effect. Incubation with normal mouse IgG as a nonbinding control antibody had no effect on cell growth (data not shown). However, the endo-NE-mediated growth inhibition was completely prevented if cells were treated with the dendrimeric C3d peptide (39), a synthetic ligand of the first immunoglobulinlike domain of NCAM (Fig. (Fig.2b).2b). In contrast, an inactive variant of C3d, C3d2Ala (39), was unable to protect the cells from the endo-NE-induced growth inhibition (not shown).
As controlled by cellular ELISA, incubation with the C3d peptide neither interfered with PSA degradation nor led to reduced cell surface expression of NCAM (cell surface immunoreactivities of untreated controls and of cultures treated with endo-NE in the presence and absence of C3d [arbitrary units ± standard error of the mean] were PSA, 0.77 ± 0.03, 0.02 ± 0.005, and 0.02 ± 0.002, respectively; NCAM, 0.54 ± 0.08, 0.76 ± 0.02, and 0.62 ± 0.05, respectively; n = 6). Also, C3d had no autonomous effect on proliferation (Fig. (Fig.2b).2b). The protective effect of C3d therefore strongly suggests that the inhibition of proliferation involves interactions of NCAM. Moreover, because the C3d peptide binds to the first immunoglobulinlike domain of NCAM, this module appears to be involved in mediating this interaction.
The next question was if trans-interacting NCAM is also able to cause growth inhibition. In order to mimic the situation in the cell culture system, experiments were carried out with NCAM-positive and NCAM-negative cell membranes isolated from differentially transfected LS cells. The phenotypes of transfected LS cells are shown in Fig. Fig.2c.2c. Parental LS are NCAM and PSA negative (Fig. (Fig.1).1). The same is true for cells transfected with an empty vector (NCAM negative, PSA negative; Fig. Fig.2c,2c, LSmock). Cells after transfection with a vector driving the expression of NCAM-140 stained positive in immunohistochemistry for NCAM but not for PSA (NCAM positive, PSA negative; Fig. Fig.2c,2c, LSAM1), while double transfectants containing vectors driving NCAM-140 and polysialyltransferase ST8SiaIV expression, stained positive for both epitopes (PSA positive, NCAM positive; Fig. Fig.2c,2c, LSAM1PST). Membranes from PSA-positive, NCAM-positive double transfectants were isolated before and after endo-NE digestion (compare Fig. Fig.2c,2c, LSAM1PST, untreated versus LSAM1PST + endo-NE).
In the experiment, SH-SY5Y (PSA positive, NCAM positive) and the control cells (LS, PSA negative, NCAM negative) were overlaid with membrane preparations of all phenotypes. Microscopic control revealed that the membrane particles settled rapidly and dispersed uniformly over cells and cell-free areas (Fig. (Fig.2c,2c, lower panels for an example). After a coculture period of 2 days, the growth rates were compared (Fig. (Fig.2d).2d). Membranes isolated from mock (PSA negative, NCAM negative) or double transfected LS cells (AM1PST; PSA positive, NCAM positive) did not noticeably affect cell growth in either of the cell systems. In contrast, the membrane fractions derived from LSAM1 or LSAM1PST after endo-NE treatment (both are PSA negative, NCAM positive) significantly inhibited the growth of SH-SY5Y cells (which are PSA positive, NCAM positive) and of LS cells (which are PSA negative, NCAM negative).
Although unexpected, this observation indicates that the growth inhibition induced by endo-NE or by NCAM itself results from trans binding of nonpolysialylated NCAM to heterophilic target structures. These findings received further support by the observation that growth of LSAM1 cells and LS AM1PST plus endo-NE (both PSA negative, NCAM positive, metabolic rates [day 2/day 0 ± standard error of the mean]: 1.62 ± 0.06 and 1.76 ± 0.09) was significantly reduced compared to nontransfected LS cells, LSmock (both PSA negative, NCAM negative), or LSAM1PST (PSA positive, NCAM positive; metabolic rates [day 2/day 0 ± standard error of the mean]: 2.21 ± 0.07, 2.19 ± 0.07, and 2.22 ± 0.08; n = 10 to 12; t test, P < 0.01).
The above data imply that PSA regulates NCAM-dependent cell-cell interactions, which requires its localization at cellular contact sites. To check this, the cell surface distribution of PSA and NCAM was studied. As visualized by immunogold detection with energy-filtering transmission electron microscopy (26), PSA appears in clusters on SH-SY5Y cells and is preferably localized at sites of tight cell-cell contacts (Fig. (Fig.3a).3a). Immunofluorescence staining and confocal microscopy clearly demonstrate that NCAM and PSA are mainly colocalized at cellular contact sites (Fig. (Fig.3b,3b, left panel). Removal of PSA did not change the strong NCAM-immunoreactivity at contact sites, however, the number of NCAM-positive contact zones increased significantly after 2-day of endo-NE treatment (Fig. (Fig.3b,3b, right panel). Even if seeded as single cell suspension at low densities, the tumor cells used in this study never grew without any contacts between each other. Time lapse observations performed with SH-SY5Y cells stably transfected to express the enhanced green fluorescent protein (EGFP) in the cytosol (SH-SY5YEGFP) indicate that, besides the broad and rather stable contacts shown in Fig. Fig.3b,3b, even those cells that have no contact at a given time point, readily will form new contacts or just broke up existing contacts due to extensive cellular motility (Fig. (Fig.3c3c).
Taken together, these analyses of cell growth indicate that downregulation of PSA enables heterophilic NCAM interactions at contact sites between tumor cells leading to reduced proliferation in the recipient cell.
Because previous studies of NCAM signaling demonstrated the involvement of the p44/p42 MAPK ERK1/2 pathway (24, 27, 43), this system was surveyed during endo-NE treatment of SH-SY5Y cells. Western blotting with anti-NCAM MAb 123C3 was used to monitor the conversion of the hardly detectable high-molecular weight smear typical for highly sialylated NCAM into PSA-free isoforms seen as sharp bands of 140 and 180-kDa (Fig. (Fig.4a).4a). Already after 10 min of endo-NE digest, significant amounts of nonpolysialylated NCAM were generated, but only after 1 h the reaction was completed. Endo-NE was constantly present in the cultures and no PSA reappearance was detected during the 2 days of the experiment. Changes in ERK activity were examined by Western blotting with antibodies directed against dually phosphorylated ERK1/2, indicative for the activated MAP kinase (28), and compared to the total amount of ERK protein (Fig. (Fig.4b4b).
To apply equal amounts of endo-NE, the enzyme was diluted in culture medium and medium were changed in all experimental groups at the beginning of the experiment. In the control group, the supply with fresh serum resulted in a moderate increase of ERK phosphorylation, relapsing to baseline within 16 h (Fig. 4b and c). In endo-NE treated cultures a strong increase of ERK phosphorylation was observed (Fig. 4b and c), while the total amount of ERK1/2 protein remained constant throughout the experiment (Fig. (Fig.4b,4b, lower panel). ERK phosphorylation was maximal as soon as 10 min after the beginning of the treatment and the peak level was maintained up to 1 h before it slowly returned to the level of the time-matched controls (Fig. 4b and c). In contrast to the effect of medium replacement, the strong MAPK activation during PSA digest was not caused by serum factors, because in the absence of serum endo-NE treatment carried out for 10, 30, or 60 min resulted in the same activation of MAPK as in serum-supplemented cultures (Table (Table1).1). However, in contrast to the MAPK activation after supply with fresh serum, ERK phosphorylation was not increased by medium replacement and these experiments were limited to 1 h, because longer periods of serum withdrawal led to a marked increase of cell death. The parallel analysis of TE671 rhabdomyosarcoma cells produced an identical time course of ERK phosphorylation after endo-NE treatment (Table (Table11).
If the endo-NE-induced MAPK activation relates to signaling via the NCAM molecule, factors that prevented growth retardation as shown in Fig. Fig.2b,2b, should be able to interfere with this induction. In fact, a complete inhibition of ERK phosphorylation was observed, if PSA removal was carried out in the presence of 1 μM C3d peptide (Fig. (Fig.5a).5a). As shown in Fig. Fig.5b,5b, similar results were obtained by antibody inhibition. Here it is important to note that, in contrast to the long-term incubations described above, short periods of antibody application had no apparent cytotoxic effect. After endo-NE treatment for 1 h, SH-SY5Y cells were incubated for 10 min with antibodies directed against NCAM (MAb 123C3), ganglioside GD2 (MAb 14G2a) or PSA (MAb 735). Although no complete reversal could be achieved by this specific protocol, application of MAb 123C3, but not 14G2a or 735, clearly reduced the ERK phosphorylation induced by endo-NE (Fig. (Fig.5b,5b, endo-NE-induced pERK intensities relative to the untreated control: 2.29 without additional antibody application, 1.38 after MAb 123C3, 1.96 and 2.53 after MAb 14G2a and 735, respectively). The same experiment performed with TE671 cells yielded identical results (endo-NE-induced pERK intensities relative to the untreated control: 1.77 without antibody, 1.30 after MAb 123C3, 2.23 and 1.95 after MAb 14G2a and 735, respectively).
Next, NCAM-positive membranes were tested for their capability to activate MAPK. No MAP kinase activity could be detected in the membrane fractions themselves (not shown). NCAM-positive membranes (AM1, AM1PST +Endo) were able to induce ERK phosphorylation in PSA- and NCAM-positive SH-SY5Y cells, but membranes containing polysialylated NCAM (AM1PST) were ineffective (Fig. (Fig.5c).5c). Likewise, ERK phosphorylation in TE671 cells was stimulated only by NCAM-positive, PSA-negative membranes (data not shown). Exposure of the PSA- and NCAM-negative LS cells to NCAM-positive membranes resulted in an ERK activation with a time-course similar to that, seen after removing PSA from the surface of the PSA-NCAM-positive neuroblastoma cells (Fig. (Fig.5d).5d). As for the endo-NE treatment of the PSA- and NCAM-positive SH-SY5Y cells, this induction of MAPK signaling was clearly inhibited by concomitant application of C3d (Fig. (Fig.5e,5e, left panel) and completely blocked in the presence of PSA (Fig. (Fig.5e,5e, right panel, AM1PST). In accordance with the growth inhibition experiment (see Fig. Fig.2d),2d), these data are most consistent with the assumption that the MAPK activation after removal of PSA is induced by heterophilic NCAM interactions.
To elucidate the contribution of MAPK activity to changes of cell growth induced by endo-NE, the effects of MAPK inhibition were compared to those of PSA removal. In serum-supplemented cultures of SH-SY5Y cells, the inhibition of ERK phosphorylation by the MEK inhibitor PD98059 (Fig. (Fig.6a)6a) reduced cell growth, and growth inhibition induced by endo-NE was significantly enhanced in the presence of PD98059 (Fig. (Fig.6b,6b, upper graph). The semiquantitative evaluation of intracellular mono- and oligonucleosomes by ELISA revealed that EndoNE treatment did not induce apoptotic cell death, whereas the inhibition of MAPK led to a small but clear-cut increase of apoptosis, which occurred independent of endo-NE treatment (Fig. (Fig.6b,6b, lower graph). As evident from the intracellular fragmentation of nuclei, a strong induction of apoptosis was observed after 2-day of serum withdrawal (Fig. (Fig.6c).6c). Only 40% of the cells survived the 2-day starvation in otherwise untreated or in dimethyl sulfoxide-treated controls (Fig. (Fig.6d,6d, upper graph), and the high rate of apoptosis was confirmed by the semiquantitative evaluation of intracellular mono- and oligonucleosomes (Fig. (Fig.6d,6d, lower graph).
By endo-NE treatment, the survival rate of the serum-starved cells was significantly improved (Fig. (Fig.6d,6d, upper graph), while apoptosis was reduced (Fig. (Fig.6d,6d, lower graph). In the presence and in the absence of endo-NE, MAPK inhibition with PD98059 reduced the survival and increased apoptosis (Fig. (Fig.6d).6d). In a second set of experiments, a clear increase of apoptosis was detected as soon as 4 h after the onset of serum withdrawal. In accordance with the results shown in Fig. Fig.6d6d for the 2-day incubation period, the incidence of apoptosis was enhanced by the inhibition of MAPK with PD98059, reduced by PSA removal with endo-NE and the anti-apoptotic effect of endo-NE was reversed by MAPK inhibition (apoptotic indices for the 4 h treatments: solvent-control [dimethyl sulfoxide]: 0.291, PD98059: 0.334, endo-NE: 0.213, endo-NE+PD98059: 0.302). Despite the significantly higher number of cells resulting from the endo-NE treatment, the percentage of BrdU-positive cells was decreased (Fig. (Fig.6e).6e). Thus, removal of PSA from serum-deprived cells induced a similar inhibition of proliferation as under serum-supplemented conditions (see Fig. Fig.1b).1b). Together, these data indicate that the activation of MAPK after PSA removal or NCAM application exerts a survival promoting, anti-apoptotic effect but did not cause the growth inhibition observed after these treatments.
The activation of MAPK observed in our experiments is highly reminiscent to the time course of ERK activity underlying the induction of neuronal differentiation in PC12 cells (28). The induction of neuronal differentiation was therefore investigated by counting neuritic extensions in NCAM- and endo-NE treated SH-SY5Y cells. SH-SY5YEGFP were used in these experiments because all processes can be reliably detected by fluorescence microscopy of living cells (Fig. 7a to d). In line with the well-characterized potential of SH-SY5Y cells to differentiate into a neuron-like phenotype (35), process-bearing SH-SY5YEGFP cells can be stained with an antibody against phosphorylated neurofilament, a marker of neuronal differentiation (46) (Fig. 7a and b, insets). Accordingly, long processes were referred to as neurites and cells with neurites longer than 20 μm were evaluated to assess the degree of neuronal differentiation. In untreated cultures grown for 48 h, an average of 21% of the SH-SY5YEGFP cells developed processes longer than 20 μm, ranging up to 120 μm (mean length ± standard error of the mean, 36 ± 0.8 μm, n = 401).
Incubation with solvent (dimethyl sulfoxide) or MAPK inhibition by PD98059 caused no significant changes in neurite lengths or in the amount of neurite-bearing cells, while in the presence of endo-NE the neurites grew slightly longer and the number of neurite-bearing cells was clearly increased (Fig. 7e and f). This increase was completely prevented by the concomitant application of PD98059 (Fig. (Fig.7f).7f). Similar to the effect of endo-NE, coculture of SH-SY5YEGFP with NCAM-positive membrane fractions from LSAM1 or endo-NE treated LSAM1PST induced a strong increase in the relative amounts of neurite-bearing cells together with a slight enhancement of neurite length (Fig. 7g and h). Consistent with the results on MAPK activation (see Fig. Fig.5c),5c), NCAM negative (mock) or PSA-NCAM-positive (AM1PST) membrane fractions had no effect (Fig. 7g, h). Despite the prolonged activation of MAPK after NCAM-exposure of untransfected LS cells (see Fig. Fig.5d),5d), no neurite outgrowth could be induced in this cell line (data not shown). Since protocols such as retinoic acid or growth factor treatment, which induce neuronal differentiation in other neuroblastoma cell lines, were also ineffective with LS cells (Seidenfaden, unpublished observation), these cells appear unable to differentiate morphologically. Thus, downregulation of PSA or exposure to nonpolysialylated NCAM induces a MAPK-dependent change towards a neuron-like phenotype in cells that exhibit the ability for such a differentiation.
Numerous studies indicate a major contribution of PSA to cellular plasticity in neural development, remodeling, and repair (9, 41), and PSA is increasingly recognized as a positive modulator of tumor malignancy (7, 8, 14). The present study establishes for the first time that expression of PSA affects NCAM-dependent signaling implicated in the regulation of tumor cell proliferation, survival and differentiation. This function of PSA appears clearly distinct from its role as a positive regulator of chain migration and axon fasciculation (9) and from the anti-adhesive properties of PSA, which seems to be dominated by the modulation of NCAM-independent cell adhesion (13).
Endo-NE treatment of PSA-NCAM-positive neuroblastoma or rhabdomyosarcoma cells removed PSA without altering NCAM expression and induced growth inhibition, MAPK activation, as well as MAPK-dependent survival and differentiation. Our data provide substantial evidence that these effects are due to the release of heterophilic NCAM interactions at cell-cell contact sites and neither due to endo-NE effects other than PSA removal, nor due to interactions of PSA itself. (i) Unspecific responses to endo-NE can be excluded, since heat-inactivated endo-NE or endo-NE treatment of PSA-negative cells had no effect. (ii) In contrast to the experiments suggesting a role of PSA in BDNF signaling (49), the effects of endo-NE on tumor cells could not be mimicked by soluble PSA. The PSA-dependent growth of SH-SY5Y cells cannot be assigned to a specific interaction with BDNF, since the mitogenic responses to different growth factors were uniformly reduced by endo-NE treatment. (iii) Together with the appearance of nonpolysialylated NCAM, ERK phosphorylation was increased after 10 min of endo-NE treatment. As the complete removal of PSA takes 1 h, the occurrence of some nonpolysialylated NCAM appears sufficient to stimulate MAPK. Interactions of PSA or nonpolysialylated NCAM with serum factors cannot account for this effect, since MAPK is activated in the presence and absence of serum. Similarly, the survival promoting activity of PSA removal under conditions of serum-withdrawal appears not compatible with the idea that growth inhibition could be due to a depletion of soluble factors by nonpolysialylated NCAM. With or without PSA, NCAM is highly concentrated at cell-cell contact sites, and the number of NCAM-positive cell-cell contacts increased after PSA removal. Even at low densities, the tumor cells never grow without any contact between each other. This supports the view that removal of PSA enables NCAM interactions at cell-cell contact sites. (v) Growth inhibition and MAPK activation after PSA removal were abolished by the NCAM-specific ligand C3d, which previously has been shown to prevent the neuritogenic effect of trans-interacting NCAM (39). Despite the nonspecific cytotoxicity of antibodies that bind to the cell surface, short-term incubations with an antibody against NCAM specifically interfered with the MAPK activation induced by endo-NE. Thus, two distinct NCAM-specific ligands were able to inhibit effects induced by endo-NE. (vi) Incubations with membrane fractions containing PSA-NCAM, nonpolysialylated NCAM or no NCAM at all, was chosen as an approach to imitate cell-cell contacts as close to the in vivo situation as possible. For a functional assay of PSA-NCAM versus nonpolysialylated NCAM, the membrane association appears particularly important, as differences in polysialic acid content do not alter the binding properties of solubilized NCAM (18). In line with the assumption of a contact-dependent effect of NCAM, growth inhibition, MAPK activation and MAPK-dependent differentiation were induced by NCAM-positive but not by PSA-NCAM-positive or NCAM-negative membranes. (vii) Membranes from PSA-NCAM-positive cells treated with endo-NE induced the same effects as membranes derived from NCAM-positive, PSA-negative cells. Thus, nonpolysialylated NCAM synthesized de novo or produced by endo-NE treatment was equally effective, i.e., PSA expression either has no effect or a reversible effect on the membrane representation of NCAM. (viii) NCAM-positive membrane fractions induced growth inhibition and MAPK activation of the PSA- and NCAM-negative LS cells. PSA-NCAM positive membranes never induced any effect, indicating that PSA prevents the NCAM interactions in question. Nevertheless, the NCAM-positive membranes were effective, if applied on PSA-NCAM-positive cells and were able to mimic all the effects observed after endo-NE treatment. The data therefore provide direct evidence for heterophilic NCAM interactions and strongly imply that heterophilic NCAM interactions underlie the changes of cell growth and differentiation induced by PSA removal.
Although the nature of these heterophilic NCAM interactions remains to be resolved, the known cell surface binding partners of NCAM can either be excluded or appear highly unlikely to account for the effects of PSA removal. Interactions of NCAM with L1 (21) cannot underlie the effects of endo-NE treatment on TE671, as these cells are negative for L1 (Hildebrandt, unpublished data). Substantial evidence points towards activating cis-interactions of NCAM with fibroblast growth factor receptors (3, 42). However, fibroblast growth factor, such as BDNF and nerve growth factor, was mitogenic for SH-SY5Y cells and thus induced the opposite effect of PSA removal or NCAM exposure. NCAM binds to heparan sulfate proteoglycans (4) but this binding is promoted by the presence of PSA (47), i.e., in contrast to the results of the current study removal of PSA and NCAM exposure should have contrary effects. In addition, NCAM binding to heparan sulfate proteoglycans is engaged in cell-substrate rather than cell-cell interactions (3, 37) affecting the NCAM-bearing cell and not an NCAM-negative recipient cell as in the current study.
The first indication for a potent heterophilic effect of NCAM came from mutant mice, which produce only secreted forms of NCAM and die during embryonic development (38). Trans-interacting NCAM has been repeatedly reported to induce growth inhibition, but the significance of PSA for this process has never been addressed (5). Notably, heterophilic interactions of NCAM inhibit proliferation and promote differentiation of hippocampal progenitor cells (1). Thus, the effects of PSA removal or trans-interacting NCAM on neuroblastoma cells were highly reminiscent to the stimulation of neural progenitors with NCAM. In contrast to these trans-interactions with a heterophilic NCAM binding partner on the recipient cell, the neurite elongation and MAPK activation of PC12 cells and cultured neurons after NCAM activation were assigned to homophilic NCAM-NCAM binding with NCAM as the neuritogenic receptor (24, 25, 34) and the NCAM-dependent formation of a signaling complex in pancreatic tumor cells is thought to induce neurite outgrowth independent of cell-cell contacts (3).
In conclusion, our data strongly suggest that PSA acts as a negative regulator of heterophilic NCAM signals at sites of cell-cell contact, which after downregulation of PSA trigger the cell to cease proliferation and to differentiate. The dynamic regulation of PSA therefore provides the control over an instructive signal for tumor cell growth. The PSA-positive neuroblastoma and rhabdomyosarcoma cell lines will be important to unravel the molecular mechanisms underlying the changes in cell growth and differentiation after downregulation of PSA, while the PSA- and NCAM-negative LS neuroblastoma cells appear suited for searching heterophilic NCAM receptors involved. Further unravelling the impact of PSA on NCAM signals will allow new insights into cell contact dependent growth control and opens up new therapeutic options for PSA-positive tumors.
Ralph Seidenfaden and Andrea Krauter contributed equally to this work.
We thank U. Paulus for electron microscopy, S. Kustermann, K. Marquart, and I. Röckle for cell culture work, K.-H- Herzog and A. Schulz for BrdU immunostaining, E. Bock, N. Pedersen, V. Matranga, R. Handgretinger, M. Fukuda, A. Münster, and R. Michaelidis for cells and reagents, and M. Mühlenhoff for critical comments on the manuscript.
This work was supported by grants from the Deutsche Forschungsgemeinschaft and Fonds der Chemischen Industrie to H.H. and R.G.-S.