Syndecan-2 Is Involved in the Mitogenic Activity and Signaling of Granulocyte-Macrophage Colony-stimulating Factor in Osteoblasts*

We previously showed that granulocyte-macrophage colony-stimulating factor (GM-CSF) binds to heparan sulfate proteoglycans expressed at the surface of osteoblastic cells and that the mitogenic activity of this cytokine is dependent on the presence of fully sulfated proteoglycans. In this study, we determined if GM-CSF interacts with syndecans, a family of cell surface heparan sulfate proteoglycans. Human primary osteoblasts were found to express syndecan-2 and -4 but few synde-can-1 transcripts and proteins. Recombinant human GM-CSF coupled to biotin was found to bind to synde-can-2. Immunocytochemical transmission electron microscope analysis showed co-localization of syndecan-2 and GM-CSF at the cell membrane surface. Syndecan-2 also co-localized at the cell surface and co-immunopre-cipitated with the GM-CSF receptor a chain, suggesting a strong interaction between the cytokine, its receptor, and syndecan-2. Phosphorylation of tyrosine residues in syndecan-2 associated with the a chain of the GM-CSF receptor was increased after cell stimulation by GM-CSF. Antisense oligonucleotides that reduced specifically the expression of syndecan-2 inhibited the mitogenic activity of GM-CSF and the activation of extracellular signal-regulated kinase-1 induced by the cytokine. Our results indicate functional interactions between syndecan-2 and GM-CSF in osteoblasts, and we propose that syndecan-2 plays a role as a co-receptor for this cytokine. Oligonucleotide syndecan-2 precise of

Matrix and cell surface heparan sulfate proteoglycans (HSPG) 1 are complex and versatile molecules that display various functions in the cellular environment. HSPGs can interact with many different ligands and thereby participate to cell adhesion, proliferation, and differentiation (1). Among HSPGs expressed on the surface of most types of cells, syndecans form a family that includes four members, three of them (syndecan-1, -2, and -4) being cloned in human cells (2)(3)(4)(5). Syndecans are expressed in cell-, tissue-, and development-specific pat-terns. In mineralized tissues, as in other mesenchymal tissues, syndecan-1 is only expressed transiently at particular stages of morphogenesis and cell differentiation (6). For example syndecan-1 is expressed during mesenchymal condensation in the developing tooth (7). In adult tissues, syndecan-1 is mainly expressed in epithelial cells (8). In contrast, the major source of syndecan-2 appears to be mesenchymal cells. Notably, syndecan-2 occurs during mouse bone development and osteoblast differentiation and persists in differentiated hard tissues (9). Syndecan-3 was first identified in neural tissues but also seems to play a role in limb development (10). Syndecan-4 displays a more ubiquitous distribution in different tissues (11).These transmembrane PGs appear to display two main functions. One is to bind extracellular ligands such as matrix adhesive proteins, cell-cell adhesion molecules, enzymes, and growth factors (11)(12)(13). All known ligand binding sites are localized on glycosaminoglycan (GAG) chains that syndecans bear on the extracellular domain of their core protein. These GAGs are mostly of the heparan sulfate type, although chondroitin sulfate chains could also be associated with syndecan-1 and -4 (11). The second function of syndecans may be to promote signaling events that are associated with the ligand-dependent activation of high affinity receptors. For example, syndecans were found to stimulate fibroblast growth factor receptor (FGFR)-1 occupancy and signaling by FGF (14). This may result from an increased availability of the ligand retained at the cell surface or from a more efficient presentation to the receptor (15,16). Moreover, syndecans display a highly conserved intracytoplasmic domain that includes several phosphorylatable residues. This may indicate an important function of the cytoplasmic domain that may be kinase substrates. Indeed, syndecan-4 can be phosphorylated on a serine residue, this phosphorylation being regulated by the protein kinase C and a FGFdependent serine/threonine phosphatase (17). On the other hand, the cytoplasmic tail of syndecans was shown to be associated with molecules involved in intracellular signal transduction. Thus, the cytosolic part of syndecan-3 can bind a protein complex including Src family kinases and their substrates (18). The COOH-terminal FYA sequence of syndecans was reported to interact with syntenin, a PDZ protein (19). Another PDZ domain-containing protein, the CASK/LIN-2, was found to bind the COOH-terminal motif of syndecan-2 (20,21). These proteins may connect syndecans to the cytoskeleton and different signaling pathways and allow them to participate to signaling events induced by PG ligands.
Granulocyte-macrophage colony-stimulating factor (GM-CSF) is an heparin binding factor (22,23) implicated in the control of hematopoietic cell proliferation and differentiation (24,25). The biological activities of GM-CSF are mediated by a transmembrane high affinity receptor that is a heterodimeric complex composed of an ␣ chain, which binds GM-CSF specifically with low affinity, and a ␤ chain, which does not bind the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  cytokine but is essential for its signal transduction (24). Both ␣ and ␤ chains are members of the hematopoietin receptor superfamily and lack intrinsic tyrosine kinase activity. Signaling by GM-CSF depends on tyrosine kinases associated with the receptor, such as JAK2, which is associated with the ␤ chain of the GM-CSF receptor (26). Stimulation of GM-CSF-dependent cell lines with the cytokine has been found to induce a variety of immediate cellular responses including rapid tyrosine phosphorylation of cellular substrates, activation of components of the Ras signaling pathway such as mitogen-activated protein kinase (MAPK), and induction of early genes transcription (26,27). In the skeleton, GM-CSF is a potential regulator of bone cells. This hematological factor was found to be produced by different osteoblastic cell lines in response to stimuli (28,29). Exogenous GM-CSF stimulates human osteoblast-like cell growth and antagonizes the induction by 1,25-dihydroxyvitamin D of osteocalcin and alkaline phosphatase activity (30). We previously showed that GM-CSF is an autocrine growth factor for normal human osteoblastic cells (31). We subsequently found that GM-CSF not only binds to HSPG expressed at the osteoblastic cell surface but that the mitogenic activity of GM-CSF is dependent on the presence of fully sulfated PGs in these cells (32). This suggested to us that GM-CSF could be a ligand for syndecans that may be involved in the control of the mitogenic activity of the cytokine. In this study we therefore determined if GM-CSF is able to interact with syndecans in osteoblasts. Our results provide evidence for functional interactions between syndecan-2 and GM-CSF, and we propose that syndecan-2 may play a role as a co-receptor for this cytokine in osteoblastic cells.
Human Cell Cultures-Primary cultures of normal human osteoblastic (hOB) cells and cloned immortalized hOB cells (AHTO-7) were obtained as described previously (33,34) and cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (FCS).
Polymerase Chain Reaction (PCR) Amplification-RNA extracted from hOB cells were used as template for reverse transcriptase reaction. The cDNA obtained was amplified using amplification primers that were synthesized based on published sequences of human syndecan-1 (2), human syndecan-2 (3), and human syndecan-4 (5) ( Table I).
Aliquots of the amplified cDNA were size-fractionated in 2% agarose gel, and the PCR products were then identified by Southern hybridization using specific internal oligonucleotide probes ( Table I) that were 5Ј end-labeled.
Immunocytochemistry-The hOB cells were grown in multiwell chambers and fixed in 4% paraformaldehyde. Nonspecific binding sites were blocked with phosphate-buffered saline (PBS) containing 3% bovine serum albumin and goat or sheep serum diluted 1:3. The following antibodies were used as primary antibodies: affinity-purified mouse monoclonal anti-human plasmacytes that recognize syndecan-1 (MCA681), mouse monoclonal antibodies 10H4 and 8G3 recognizing syndecan-2 and -4, respectively (5,9), mouse monoclonal antibody recognizing the human common ␤ chain receptor for interleukin-3, interleukin-5, and GM-CSF, and mouse monoclonal antibody anti-human GM-CSF ␣ chain receptor. The secondary antibodies were anti-mouse IgG linked to colloidal gold particles (Amersham Pharmacia Biotech), which were enlarged by precipitation of metallic silver before microscopic visualization using an inverse condenser (Olympus BH-2). For transmission electron microscopy analysis, cells were incubated with 10 ng/ml rhGM-CSF for 5 min at 37°C, then washed in PBS and fixed in 4% paraformaldehyde. A double immunostaining using mouse antisyndecan-2 and polyclonal rabbit anti-GM-CSF or anti-GMR␣ was then performed. Labeling was revealed using anti-mouse and anti-rabbit secondary antibodies linked to 10-nm and 5-nm beads, respectively. After the immunological reaction, cells were post-fixed with 0.1% glutaraldehyde, dehydrated in graded ethanol series, and embedded in epoxy resin. Ultrathin sections (60 -70 nm) were performed, treated with uranyl acetate and lead citrate, and observed in transmission electron microscopy (Jeol 100 CXII) at high magnification. Isolated and coupled gold particles (total Ͼ1,000 particles) localized at the outer surface of the cells were counted.
Binding Assay of Biotin-rhGM-CSF to PG-Biotin was linked to rhGM-CSF using a previously described method (36). To do that, 10 g of rhGM-CSF were reacted with 1 g of NS-biotin in PBS at pH 8.3. The reaction was performed in the presence of 100 g of heparin to protect potential heparin-binding sites on GM-CSF. At the end of the reaction, PBS containing 2 M NaCl was added to dissociate heparin from proteins, and rhGM-CSF linked to biotin (biotin-rhGM-CSF) was separated from free reagents on Sephadex G-25. Proteoglycans (0.2 mg/well) obtained as described above were subjected to 4 -15% SDS-PAGE under nonreducing conditions and were electrotransferred onto a polyvinylidene fluoride membrane (Amersham Pharmacia Biotech). The membrane was cut in different strips; one part was used for binding assay, and the other part for immunoblotting as described below, except that diaminobenzamidine was used as peroxidase substrate. Precise marks on the membrane allowed its reconstitution at the end of the experiments. Before the binding assay, strips of the membrane were digested at 37°C with 10 milliunits/ml heparinase III or 33 milliunits /ml chondroitinase ABC diluted in Tris-HCl (pH 7.2) containing 1 mM CaCl 2 , 0.5 mg/ml bovine serum albumin, and a mixture of protease inhibitors (35). Digested and nondigested strips of the membrane were blocked for nonspecific binding in 10 mM Tris-HCl (pH 7.5) and 150 mM NaCl (Tris-buffered saline) containing 0.5% gelatin for 1 h at room temperature and then exposed to biotin-rhGM-CSF for 90 min at 37°C. At the end of this incubation, after rapid washes, strips were reacted with avidin-peroxidase for 30 min. Recombinant hGM-CSF bound to proteoglycans on the membrane was revealed using diaminobenzamidine as peroxidase substrate.
Immunoprecipitation and Immunoblotting-Confluent AHTO-7 cells were serum-starved for 24 h. The medium was then changed for Dulbecco's-modified Eagle's medium containing 1% bovine serum albumin, and after a 1-h incubation at 37°C, the cells were treated with rhGM-CSF for 0, 5, 10, or 30 min. The cells were then transferred onto ice, washed with PBS, and lysed in 10 mM Tris (pH 7.5) with 5 mM EDTA, 200 mM NaCl, 30 mM sodium pyrophosphate, 50 mM sodium fluoride, 1 mM activated sodium orthovanadate, 10% glycerol, and 0.5% Triton X-100. The cell lysates were clarified by incubation with 10 l of protein A-Sepharose for 1 h and centrifugation at 12,000 ϫ g for 10 min. Protein content was then quantified. One mg of each protein extract was incubated overnight with polyclonal anti-GMR␣ or anti-syndecan-2 (10H4) and 10 l of protein A-Sepharose. Immune complexes were washed 3 times with lysis buffer, boiled 10 min in Tris containing 2.3% SDS, 10% glycerol, and 5% 2-mercaptoethanol, then resolved on SDS-polyacrylamide gel and electrotransferred onto a polyvinylidene fluoride membrane. Immunoblotting was performed by incubating the membrane with primary antibodies at 4°C overnight and then with horseradish peroxidase-conjugated anti-immunoglobulin (Amersham Pharmacia Biotech) as the secondary antibody. Signals were detected using a chemiluminescent peroxidase substrate (Amersham Pharmacia Biotech).
Antisense Oligonucleotide Experiments-The antisense oligomer (AS) used was complementary to the first 25 bases of the translation start site of rat syndecan-2 mRNA (37) and had the following sequence: CCGGGGACGTGGCTCGTACCC. A random (R) oligonucleotide composed of the 25 randomly arranged bases of the AS was used as control and had the following sequence: AGTGCGGCGCCGCGTCTAGCC. The R sequence was checked on a computed data base not to be complementary to any known sequence. The oligonucleotides were protected from endonucleases by phosphorothioate modification at the 3Ј end. Before oligonucleotide treatment, hOB cells were plated at 5,000 cells/cm 2 in multiwell chambers and cultured for 2 days in the presence of 10% FCS to allow homogenous cell adhesion and with 30 mM sodium chlorate to reduce the presence of sulfated glycosaminoglycans at the cell surface (32). The medium was then changed for serum-free Dulbecco's modified Eagle's medium containing 2 M AS or R oligonucleotides or the diluent. After 2 days of culture in the presence of oligonucleotides, the medium was replaced, and the culture was continued for 3 days in the presence or absence of 10 ng/ml rhGM-CSF or 1% FCS and in the presence or absence of 10 ng/ml heparin. At the end of the culture, the cells were washed, fixed in 2.5% glutaraldehyde, stained with hematoxylin, and the number of cells/cm 2 was counted using an ocular integrator mounted on a microscope (Olympus, BH2). To ensure that AS affected only syndecan-2, the effect of AS and R oligonucleotides on the expression of syndecans was determined by immunocytochemistry in the same culture conditions as the cell proliferation assay. Cell cultures were cultured for 2 days in the presence of AS or R oligonucleotides and stopped before rhGM-CSF treatment, and immunocytochemistry of syndecan-1, -2, and -4 was performed as described above. The effect of syndecan-2 inhibition on the expression of GM-CSF receptor was also examined by immunochemistry in the same conditions. To assess that AS effectively reduced syndecan-2 levels, we used a quantitative method previously shown to be accurate in determining precise changes in protein levels under AS treatment (31,38). The number of cells expressing syndecan-2, GMR␣, and GMR␤ in the presence of AS or R oligonucleotides was counted in 3-4 wells, and the results were reported as % of total number of cells.
The effect of syndecan-2 suppression on GM-CSF signaling was determined by examining the activation of MAPK induced by rhGM-CSF in the presence of oligonucleotides. AHTO-7 cells were treated for 2 days with 2 M AS or R oligonucleotides in medium containing 1% FCS then stimulated for 10 min with 10 ng/ml rhGM-CSF and lysed onto ice. One-hundred g of total protein extract were separated on SDS-PAGE, and activation of MAPK was determined by Western blot as described above using an affinity-purified rabbit IgG that recognizes the phosphorylated active form of the MAPK enzymes, ERK1 and ERK2.

RESULTS
Osteoblasts Express Syndecans-The expression of syndecans in osteoblastic cells was examined by reverse transcriptase-PCR, immunocytochemistry, and immunoblotting in two types of osteoblastic cells, primary cultures of cells derived from the trabecular bone surface (hOB cells) and clonal immortalized hOB cells (AHTO-7 cells). Except for the proliferation rate, these two cell types display similar characteristics of the mature osteoblast phenotype (33,34). As expected in mature mesenchymal cells, hOB (Fig. 1A) and AHTO-7 cells (not shown) expressed low levels of syndecan-1. In contrast, reverse transcriptase-PCR analysis revealed high levels of syndecan-2 and -4 mRNA (Fig. 1A). These results were confirmed by Northern blot hybridization (not shown). Syndecan-2 and -4 synthe-sis was found to correlate with the mRNA levels as shown by immunocytochemistry in hOB cells (Fig. 1B) and Western blot analysis in AHTO-7 cells (Fig. 2B). These results demonstrate that human osteoblastic cells mainly express syndecan-2 and -4.
Biotin-rhGM-CSF Binds to Heparan Sulfate Chains of Syndecan-2-To determine the interactions between GM-CSF and syndecans in human osteoblastic cells, a PG-enriched fraction was obtained from AHTO-7 cells. Separation of these proteins was based on their high negative charge, which makes them bind strongly to an anion exchange column. We initially checked that the fraction eluted by buffer with 1 M NaCl contained most of the proteins that were metabolically labeled with 35 S in AHTO-7 culture. PGs isolated from osteoblastic cells were separated on SDS-PAGE in nonreducing conditions and electrotransferred on a membrane to analyze their capacity to bind biotin-conjugated rhGM-CSF. Different species of PGs (20, 45, and 70 kDa) displayed affinity for biotin-rhGM-CSF ( Fig. 2A, lane 1). Binding of GM-CSF was strongly decreased when PGs on the membrane were digested by heparinase III before incubation with the cytokine (Fig. 2A, lane 2), whereas chondroitinase ABC digestion did not modify PG affinity ( Fig. 2A, lane 3), indicating that GM-CSF-PG interaction mostly depends on heparan sulfate side chains. To identify some of the PGs that bind GM-CSF, part of the membrane with blotted PG was incubated with anti-syndecan antibodies instead of the cytokine (Fig. 2B, lane 4 and 5). This revealed heparan sulfate chains of syndecan-2 bind rhGM-CSF.
Co-localization of Syndecan-2 and GM-CSF at the Cell Surface-To determine if the interaction between syndecan-2 and GM-CSF occurred at the surface of hOB cells, a double immunostaining of these proteins was performed using a rabbit polyclonal anti-GM-CSF and a mouse monoclonal anti-synde-

FIG. 1. Expression of syndecans in human osteoblastic cells. A,
Southern blot analysis of reverse transcriptase-PCR products obtained from total RNA extracted from hOB cells. Ethidium bromide staining showed that the amplified cDNAs have the predicted sizes for the syndecans. These products hybridized with the specific probes. hOB cells expressed high levels of mRNA for syndecan-2 and -4 and lower mRNA levels for syndecan-1. B, hOB cells were stained with purified mouse anti-human syndecan-1 (a), anti-syndecan-2 (b), or anti-syndecan-4 (c). Negative control cells were reacted with mouse Ig to detect nonspecific staining (d). Staining was revealed using anti mouse-IG link to gold particles that were enlarged by precipitation of metallic silver. A low level of syndecan-1 was detected compared with syndecan-2 and -4. bp, base pair.

Syndecan-2 Modulates GM-CSF Activity
can-2. Clear co-localization of the PG and the cytokine was found by transmission electron microscopy using secondary antibodies linked to 10-nm and 5-nm beads, respectively (Fig.  3A). In these conditions, the cells were post-fixed only after the immunocytochemical reaction was performed, explaining the low contrast of the cell membrane (Fig. 3). Beads present at the cell surface were counted and were considered as associated when the distance in between was less than 15 nm. About 30% of the beads that labeled syndecan-2 and 25% of the beads that labeled GM-CSF were found associated with beads of the other size. These results show that syndecan-2 and GM-CSF colocalize at the cell surface and further indicate that syndecan-2 is one of the PGs of the cell surface that interacts with GM-CSF in osteoblasts.
Interaction between Syndecan-2 and GMR␣-All functions of GM-CSF were shown to depend on its binding to a transmembranous receptor (GMR) (24). We thus used the method of double-staining analyzed by electron microscopy to examine the localization of syndecan-2 and GMR␣ at the surface of hOB cells. About 30% of the beads that labeled syndecan-2 or GMR␣ were found to be associated, showing that molecules of syndecan-2 were co-localized with GMR␣ (Fig. 3B). To confirm that GMR␣ and syndecan-2 associate at the cell surface, we performed a co-immunoprecipitation assay in AHTO-7 cells stimulated or not with rhGM-CSF. GMR␣ was immunoprecipitated with a polyclonal antibody, and the immunoprecipitates were resolved by SDS-PAGE and blotted onto a membrane that was probed with anti-syndecan-2. As shown in Fig. 4, syndecan-2 co-immunoprecipitated with GMR␣. The amount of syndecan-2 associated with GMR␣ did not seem to depend on stimulation by exogenous rhGM-CSF, at least after 5 and 10 min of treatment with the cytokine (Fig. 4). These results demonstrate that syndecan-2 is associated with GMR␣ at the surface of the osteoblasts and strengthen the hypothesis of functional interactions between syndecan-2 and GM-CSF.
Protein phosphorylation is one of the major events that allows propagation of intracellular signals. To address the question of the possible implication of syndecan-2 in the intracellular signaling events induced by GM-CSF, we examined tyrosine residue phosphorylation of syndecan-2 associated with GMR␣ after ligation of rhGM-CSF. We tested the effect of rhGM-CSF stimulation on tyrosine phosphorylation in syndecan-2 co-immunoprecipitated with GMR␣. To do that, the membrane with blotted immunoprecipitates was probed again with an antiphosphotyrosine. The addition of rhGM-CSF for 5-30 min induced increased phosphorylation of tyrosine residues in syndecan-2 (Fig. 4). Together, these results indicate that syndecan-2 is involved not only in the binding of the cytokine to its high affinity receptor but also in GM-CSF signaling.
Effect of Syndecan-2 Suppression on Cell Growth-To further document the role of syndecan-2 in the GM-CSF mitogenic activity, we used an antisense strategy (31,38) to inhibit the synthesis of syndecan-2 in hOB cells. We first examined by immunocytochemistry the level of syndecan-2 expression after After incubation with avidin-peroxidase, staining was revealed using diaminobenzamidine. Different PGs bound rhGM-CSF (arrowheads), and one of these is recognized by anti-syndecan-2 (double arrow).

FIG. 3. GM-CSF and GMR␣ colocalize with syndecan-2 at the cell surface.
Human osteoblastic cells were grown until confluence, fixed with paraformaldehyde, double-immunostained with polyclonal rabbit anti-GM-CSF, and monoclonal mouse anti-syndecan-2 (A) or polyclonal rabbit anti-GMR␣ and monoclonal mouse anti-syndecan-2 (B). Stainings were revealed with anti-rabbit Ig bound to 5-nm beads and anti-mouse Ig bound to 10-nm beads, respectively. Cells were then post-fixed, and 60 -70-nm sections were performed and observed with a transmission electron microscope at high magnification (ϫ60,000). About 25% of the beads present at the cell surface were associated with beads of different size (arrows).

FIG. 4. Co-immunoprecipitation of GMR␣ and syndecan-2.
AHTO-7 cells were grown until confluence and treated with rhGM-CSF for 0, 5, 10, and 30 min. The cells were lysed and immunoprecipitated (IP) with mAb 10H4, a purified polyclonal anti-GMR␣, or a rabbit Ig fraction to serve as control (NI). Immunoprecipitates were separated on SDS-PAGE, transferred onto a membrane, and immunoblotted sequentially with anti-syndecan-2 and with antiphosphotyrosine. A protein corresponding to syndecan-2 was detected in the fraction immunoprecipitated by anti-GMR␣. Syndecan-2 was phosphorylated on tyrosine residues, and this phosphorylation was enhanced by rhGM-CSF. WB, Western blot. 2 days of culture in the presence of the AS or R oligonucleotides and prior to treatment with rhGM-CSF. Syndecan-2 levels were reduced in the presence of AS but not R oligonucleotides (Fig. 5A). The quantitative determination of syndecan-2-positive cells showed that AS significantly decreased the expression of syndecan-2 in osteoblasts, whereas R oligonucleotides had no effect (Fig. 5B). AS oligonucleotides did not affect the expression of GMR␣ and GMR␤ (Fig. 5, A and B). We also checked that the AS used did not affect the expression of syndecan-1 and -4 in these cells (not shown). Having shown that AS oligonucleotides selectively reduced the level of syndecan-2 in osteoblastic cells, we examined the biological consequences of the reduced syndecan-2 expression in a proliferation assay in the presence or absence of rhGM-CSF. The cells were cultured in serum-free medium in the presence or absence of AS or R oligonucleotides alone for 2 days, and then rhGM-CSF and/or heparin or FCS were added for 3 days. Consistent with our previous findings (31,32), hOB cells proliferated in the absence of exogenous mitogenic factors, and the basal proliferation rate was enhanced in the presence of rhGM-CSF. The rate of proliferation was similar in R-treated cells and in cells cultured in the absence of oligonucleotides. Moreover, rhGM-CSF similarly enhanced the number of control and R-treated cells (Fig. 6). In contrast, AS-treated hOB cells with reduced syndecan-2 expression displayed a lower basal proliferation rate than control cells. Furthermore, the mitogenic activity of rhGM-CSF was abolished in AS-treated cells (Fig. 6). This effect was not due to a toxic effect of oligonucleotides since AS-treated cells were still able to respond to 1% FCS to the same level as control cells (Fig. 6). Moreover, we checked that the number of AS-treated cells was higher after 3 days of culture compared with the number of cells present when GM-CSF treatment was initiated (not shown), indicating that the lower number of AS-treated cells at the end of the culture resulted from a reduced proliferation rate and not from cell death. These results indicate that syndecan-2 is involved in the mitogenic activity of GM-CSF in osteoblastic cells.
We previously showed that low doses of heparin are able to recover the GM-CSF-induced proliferation in cells that express PGs with under-sulfated GAGs (32). In the present study, we found that the addition of heparin did not rescue the inhibition of GM-CSF mitogenic activity induced by syndecan-2 AS (Table  II). The inability of heparin to rescue the inhibitory effect of AS on cell growth further suggests that, in addition to its heparan sulfate compounds, the core protein of syndecan-2 participates to the control of GM-CSF-induced cell growth.
Effect of Syndecan-2 Suppression on GM-CSF Signaling-Binding of GM-CSF to its high affinity receptor leads to the activation of the MAPK signaling pathway (27) and rapid phosphorylation of p44 and p42 MAPK (also referred as extracellular signal-regulated kinase-1 (ERK1) and ERK2, respectively).
To determine the effect of syndecan-2 suppression on GM-CSF signaling, we examined the effect of AS oligonucleotides on activation of ERK1 and ERK2 by rhGM-CSF. AHTO-7 cells were treated with AS or R oligonucleotides and stimulated with rhGM-CSF. Total protein extract was then separated on SDS-PAGE, and the level of activated ERK1 and ERK2 was determined by Western blot. We found that rhGM-CSF enhanced the level of activated ERK1 and that suppression of syndecan-2 inhibited ERK1 activation by rhGM-CSF (Fig. 7). These results together with our data on syndecan-2 phosphorylation and heparin effect strongly suggest a role for syndecan-2 in intracellular signaling induced by GM-CSF binding.  DISCUSSION We recently showed that the mitogenic activity of GM-CSF in osteoblasts depends on interactions with cellular sulfated heparan sulfate chains (32). The results presented in this paper provide the first evidence for functional interactions between GM-CSF and syndecan-2 and identify this transmembrane HSPG as a potential co-receptor for this cytokine.

Syndecan-2 Modulates GM-CSF Activity
We first examined the binding capacity of PGs isolated from osteoblastic cells and showed that different HSPGs display affinity for GM-CSF. Heparinase treatment reduced homogenously the binding of the cytokine to PGs, showing that binding of GM-CSF to the cellular osteoblastic PGs depends on heparan sulfate GAGs. Identification of PGs that bear GAGs responsible for the interactions with GM-CSF is an important step to understanding how these interactions are involved in the regulation of GM-CSF activity. Syndecans appeared to be good candidates for the interactions with GM-CSF. Indeed, this protein family is the major form of membrane-associated HSPG expressed by many cells (12,13) and is able to bind growth factors including FGF (39) and heparin binding growth-associated molecule (40). We therefore examined the expression of syndecans in human osteoblasts. Our findings demonstrate that osteoblastic cells derived from adult human bone express low levels of syndecan-1 and high levels of syndecan-2 and -4. The low pattern of syndecan-1 expression is consistent with reports showing that syndecan-1 is expressed in mesenchymal tissues only at particular stages of the development and is almost restricted to epithelial cells in adult tissues (6,7). In contrast, syndecan-2 was found to be expressed exclusively by mesenchymal cells in different tissues, whereas syndecan-4 is known as the most ubiquitous distribution in different tissues (9,11). The distinct expression of syndecan-1, -2, and -4 in normal human osteoblastic cells is in accordance with their expression in osteogenic cells during mouse development and rat neonatal ossification (9,41).
Immunoblotting analysis performed in parallel with the binding assay on PGs extracted from osteoblastic cells allowed identification of syndecan-2 as a major HSPG that binds GM-CSF. To determine if this interaction between blotted synde-can-2 and GM-CSF also occurs at the cell surface, we performed a double immunological staining of syndecan-2 and the cytokine in hOB cells. Our data showed that labeled molecules of syndecan-2 and GM-CSF were associated and co-localized on the cell surface, which supports our results showing binding of GM-CSF to syndecan-2. We therefore examined this interaction to determine if it is functional and may play a role in the mitogenic activity of the cytokine. We used an antisense strategy to reduce syndecan-2 expression. AS oligonucleotides reduced specifically immunoreactive syndecan-2 levels, whereas syndecan-1 and -4 levels were not modified. Although not striking, the efficiency of AS to reduce syndecan-2 expression was comparable with that previously observed in human osteoblastic cell cultures (31). Interestingly, the reduction in syndecan-2 levels induced by AS oligonucleotides was associated with decreased basal cell proliferation. This may be related in part to reduction of endogenous GM-CSF activity, since we previously found that this cytokine is an autocrine growth factor in human osteoblasts (31). Moreover, the mitogenic activity of rhGM-CSF was abolished by syndecan-2 AS, and this effect was associated with inhibition of ERK1 activation. The inhibition of GM-CSF mitogenic activity in AS-treated cells was not due to a reduction of GM-CSF receptor expression. These data indicate that syndecan-2 plays a role in the control of GM-CSF signaling and activity in osteoblasts.
We previously showed that reduction of GAGs sulfation by chlorate treatment, which results in lower binding capacity of GAGs at the cell surface, also inhibits the mitogenic activity of GM-CSF (31). Thus, a major role of syndecan-2 in GM-CSF sequestration at the cell surface may account for the involvement of this HSPG in the cytokine activity. Interaction with GAGs has been shown to protect soluble factors from degradation (42). Such interactions are also thought to increase the concentration of the ligand at the proximity of high affinity receptors (15,16,43). The number of GM-CSF receptors expressed by responding cells is very low, typically 800 to 1000 sites/cell in hematopoietic cells and 120 sites/cell in endothelial cells (24). Binding of GM-CSF to GAGs of syndecan-2 may increase the probability for this ligand to interact with its high affinity receptors. Interaction with HSPG may also promote ligand-receptor binding. Such a role was demonstrated for perlecan, a HSPG present in the basement membrane that binds FGF-2 and allows FGF-2-mediated mitogenic response by promoting the binding of this factor to its high affinity receptor (44). Syndecans were also found to increase FGF-2 binding to its receptor when transfected in cells that express low levels of cell-surface HSPG (14). In the present study, the co-localization of syndecan-2 and GMR␣ at the surface of osteoblastic cells suggests that syndecan-2 may play a role in the presentation of GM-CSF to its receptor. In addition, syndecan-2 co-immunoprecipitated with GMR␣, indicating a strong association between syndecan-2 and the receptor. Based on these results, direct interactions between this HSPG and GMR␣ can be postulated similar to those observed between heparin and other TABLE II Heparin does not rescue the inhibition of GM-CSF mitogenic activity induced by syndecan-2 antisense Human osteoblastic cells treated with or without AS oligonucleotides (2 M) were cultured for 3 days in the presence or absence of 10 ng/ml rhGM-CSF and/or 10 ng/ml heparin and then counted. The initial cell number/cm 2 at day 0 was 4709 Ϯ 159. For comparison, random oligonucleotides had no effect in this assay (see Fig. 6). Data are the mean Ϯ S.E. of 3-4 replicates.  7. Inhibition of syndecan-2 inhibits MAPK activation induced by rhGM-CSF. AHTO-7 cells were grown until confluence and treated for 2 days with R or AS syndecan-2 oligonucleotides. The cells were then incubated in the presence or absence of 10 ng/ml rhGM-CSF for 5 min and lysed. 100 g of total protein extract were Western-blotted with anti-activated MAPK that recognizes both ERK1 and -2. rhGM-CSF stimulation increased ERK1 in R-treated cells but not in AS -treated cells. dependent growth factor receptors such as FGFR. Indeed, a ternary functional interaction between heparin, FGF, and FGFR has been demonstrated (45). A heparin binding domain was found in the FGF receptor, and suppression of this site by mutation was shown to inhibit heparin and FGF binding to the receptor and to result in inhibition of the tyrosine kinase activity of this receptor (45). Interestingly, the ␣ and ␤ chains of the GM-CSF receptor belong to the hematopoietin receptor superfamily (46). These glycoproteins comprise structures related to fibronectin type III domain, and heparin-binding sites have been localized in these fibronectin domains (46,47). Thus, functional interactions between these potential heparan-binding sites in GM-CSF receptor and heparan sulfate chains in syndecan-2 may participate to the control of GM-CSF signaling.
The interacting capacity of heparan sulfate chains of syndecans may not be alone responsible for all syndecan functions. In a previous study, we found that the inhibitory effect of chlorate treatment on GM-CSF mitogenic activity was reversed by low concentrations of heparin (32). In contrast, the inhibitory effect of AS oligonucleotides on cell growth was not rescued by the addition of heparin. This suggests that the core protein is also required and may play a role in the control of GM-CSF activity. The cytoplasmic tail of syndecans contain four conserved tyrosine residues (11). Recently, tyrosine residues of syndecan-1 and -4 were shown to be phosphorylated in adherent fibroblasts (48). This phosphorylation seems to depend on the cell type and to be tightly controlled by both kinases and phosphatases, suggesting that tyrosine phosphorylations may be a key event in syndecan activity. We show here that phosphorylation of tyrosine residues of syndecan-2 associated with GMR␣ in osteoblastic cells is increased in the presence of rhGM-CSF. This indicates that syndecan-2 is involved in intracellular signaling events that are activated by the cytokine. These results are consistent with recent findings showing that the cytoplasmic domain of syndecans are associated with signaling molecules. For example, syndecan-3, a cell surface receptor for heparin binding growth-associated molecule, binds tyrosine kinases of the Src family and substrates such as cortactin (18). Moreover, this signaling pathway appears to be activated during heparin binding growth-associated molecule-dependent neurite outgrowth (18). Highly conserved cytoplasmic domains of syndecans were also shown to interact with PDZ domains of CASK, a membrane-associated guanylate kinase (20,21). This type of multidomain scaffold proteins interacting with the cytoskeleton is thought to organize specific signaling networks and to connect extracellular signals to downstream signaling pathways. These observations together with the results presented in this paper suggest that syndecan-2 core protein may participate in the modulation of GM-CSF mitogenic activity by activation of proper signaling pathways run in parallel or connected to intracellular signaling events that depend on the GM-CSF high affinity receptor.
Interactions between syndecan-2 and GM-CSF may have functional biological implications in osteoblast P. It may result in fine tuning of osteoblastic cell growth and differentiation. Indeed GM-CSF was shown to be an autocrine/paracrine regulator of osteoblastic cell growth in vitro (28 -31) and in vivo in mouse calvaria osteoblastic cells. 2 This cytokine was also found to modulate alkaline phosphatase activity and osteocalcin (30) and reduces type I collagen expression. 2 It can also be postulated that syndecan-2 may be involved in the control of other heparan sulfate binding factors than GM-CSF in osteoblasts. For example, FGF-2 that binds to and is modulated by synde-cans in various cell types (6) is known to affect osteoblast proliferation and differentiation through interaction with FG-FRs (49). We recently found that syndecan-2 is co-expressed with FGFRs in osteoblasts during rat calvaria osteogenesis (41). Syndecan-2 may therefore play an important regulatory role, controlling the biological activity of local heparan sulfate binding growth factors in osteoblasts.
In summary, our results provide evidence that syndecan-2 interacts with GM-CSF and its receptor at the surface of human osteoblastic cells. Both the heparan sulfate chains that are responsible of GM-CSF binding and the core protein that seems involved in intracellular signaling events form a co-receptor for the cytokine. Thus, different domains in syndecan-2 are involved in the control of GM-CSF mitogenic activity and signaling in human osteoblastic cells.