Fibroblast growth factor-2 can mediate cell attachment by linking receptors and heparan sulfate proteoglycans on neighboring cells.

The myeloid 32D cell line, which grows in suspension and does not express FGF receptors or heparan sulfate proteoglycans, was transfected with the cDNA encoding FGF receptor-1 (32D-flg cells). When co-cultured with glutaraldehyde-fixed Chinese hamster ovary (CHO) cells, the 32D-flg cells remained in suspension in the absence of FGF-2 but attached to the CHO monolayer in the presence of 10 ng/ml FGF-2. In contrast, 32D cells transfected with the vector alone did not attach to the CHO monolayer in the presence of FGF-2. FGF-2-dependent attachment of 32D-flg cells was prevented by inclusion of 10 μg/ml heparin in the incubation medium and was diminished when CHO mutants in glycosaminoglycan synthesis or wild-type CHO cells treated with heparinase were used, indicating that the attachment occurred through FGF-2 interactions with heparan sulfates on the CHO cells. Attachment of 32D-flg cells to wild-type CHO cells was half-maximal at 0.4 ng/ml FGF-2 and was also observed with FGF-1 but not FGF-4. 32D-flg cells also attached to living CHO cells in a FGF-2-dependent manner, but attachment was transient at 37°C. Induction of new proteins was not required for FGF-2-dependent attachment, since attachment occurred when the co-cultures were incubated at 4°C and when the 32D-flg cells were preincubated with cycloheximide. FGF-2-dependent attachment of 32D-flg cells was also observed with Balb/C 3T3, NIH 3T3, and bovine capillary endothelial cells. We conclude that attachment is due to FGF-2 binding simultaneously to receptors on the 32D-flg cells and heparan sulfates on the CHO monolayers; thus, the FGF-2 acts as a bridge between receptorexpressing cells and heparan sulfate-bearing cells. In addition, induction of DNA synthesis in 32D-flg cells in response to FGF-2 was potentiated by the CHO-associated heparan sulfates to the same extent as by soluble heparin, indicating that this interaction has functional significance.

The fibroblast growth factors (FGFs) 1 are a family of nine polypeptides that share sequence homology and a high affinity for heparin (1,2). The members of the family have a variety of activities in vivo, including stimulation of proliferation, migration, and differentiation (1,2), and the activities of the members of the family overlap to a considerable extent (1,2). The two prototypes of the family, acidic and basic FGFs (FGF-1 and FGF-2), were originally identified and purified as factors that induce an angiogenic response in cultured endothelial cells. In vivo, FGF-1 and FGF-2 act as potent angiogenic factors and stimulate the formation of new blood vessels (3). However, these growth factors also seem to have important roles in the development and maintenance of the nervous system, skeletal system, muscle, and blood cells.
FGF-2 interacts with both specific high affinity receptors and heparan sulfate proteoglycans on the cell surface (4). The FGF receptors also constitute a family of transmembrane tyrosine kinases with four known members that have overlapping affinities for the various members of the FGF family (5). At least two of the FGF receptors, FGF receptor-1 (the flg gene product) and FGF receptor-2 (the bek gene product), are high affinity receptors for FGF-2 (6 -8). Binding of FGF-2 to FGF receptor-1 or FGF receptor-2 results in autophosphorylation of the receptor and signaling to the cell (9). Several members of the receptor family also exist in alternatively spliced forms (5). Thus, FGF receptor-1 and FGF receptor-2 can exist in forms containing either two or three immunoglobulin-like domains in the extracellular portion of the molecule. The presence of 2 or 3 immunoglobulin-like domains may alter the affinity of the receptor for its ligands (10). Another splicing variation can occur in the second half of the third immunoglobulin-like domain. Variation in this region affects ligand specificity. Expression of the IIIb exon in this location in FGF receptor-2 generates a receptor that recognizes FGF-1 and keratinocyte growth factor but not FGF-2, whereas expression of the IIIc exon generates a receptor that recognizes FGF-1 and FGF-2 but not keratinocyte growth factor (11). In addition to these regions defined by splicing variations, other regions of the extracellular portion of the FGF receptors that might have functional importance have been identified. These include (i) the acidic box, a sequence of four to eight contiguous acidic amino acids between the first and second immunoglobulin-like domains, (ii) a proposed heparin-binding domain within the second immunoglobulin-like domain (12), and (iii) a region between the first and second immunoglobulin-like domains that bears homology to the cadherin cell adhesion recognition sequence (13).
FGF-2 also binds with lower affinity to the heparan sulfate moieties of proteoglycans on the cell surface and in the extracellular matrix (4,14,15). The binding of FGF-2 to heparan sulfates confers several biological advantages to the growth factor: (i) FGF-2 bound to heparin or heparan sulfates is protected from proteolysis and thermal denaturation (16,17); (ii) the heparan sulfate-bound FGF-2 serves as a reserve of growth factor that can support long term responses to FGF-2 after a brief exposure to the growth factor (18,19); (iii) the heparan sulfates of the tissues may provide a means to localize FGF-2 to a particular site, limiting its diffusion (20); (iv) soluble heparan sulfates can act as carriers of FGF-2 and by preventing its interaction with fixed heparan sulfates in the tissues assure its dissemination away from its site of release (20); (v) FGF-2 can be internalized through its interaction with cell-surface heparan sulfates, clearing excess active molecules from the cell surface, perhaps helping to dampen the response to FGF-2 (21)(22)(23); and (vi) heparin or heparan sulfates can increase the affinity of FGF-2 for its receptors, by decreasing the dissociation rate of the FGF-2-receptor complex (24 -26). This final point suggests that trimolecular complexes of FGF-2, receptor, and heparan sulfate are formed and that these complexes are more stable than complexes of FGF-2 and receptor alone.
The interaction of FGF-2 with heparin or heparan sulfates is reported to be necessary for interaction of the growth factor with its tyrosine kinase receptor (27)(28)(29). However, several recent studies have found that heparin or heparan sulfates were not strictly required for binding of FGF-2 to its receptor but increased the affinity of the FGF-2-receptor interaction to a moderate degree (26, 30 -33). Some of these results are based on experiments with 32D cells (a myeloid cell line that does not normally express FGF receptors or heparan sulfates) that have been transfected with the cDNA encoding FGF receptor-1 (32Dflg cells). The 32D-flg cells bound FGF-2 both in the presence and in the absence of heparin, but heparin increased the affinity of binding about 4-fold (26).
Although the requirement for heparin or heparan sulfates for binding of FGF-2 to its receptor remains controversial, there seems to be a strong requirement for heparin or heparan sulfates for long term responses to FGF-2 (34). To determine whether the heparan sulfates from one cell were able to potentiate the binding of FGF-2 to its receptor on another cell type, the 32D cells expressing FGF receptor-1, which lack heparan sulfates and grow in suspension, were incubated with CHO cells, which express heparan sulfates but have very low levels of FGF receptors and grow attached to the culture dish. In the co-cultures, the normally suspended 32D-flg cells attached to the CHO monolayer in an FGF-dependent manner. Attachment appeared to be due to the simultaneous binding of FGF-2 to receptors on the 32D cells and to heparan sulfates on the CHO cells, providing a bridge between the two cell types. These results demonstrate that, under certain conditions, FGF-2 can act as a direct attachment factor and that this interaction potentiates the biological activity of FGF-2.

EXPERIMENTAL PROCEDURES
Cells-CHO K-1 cells and mutants in glycosaminoglycan synthesis derived from them (pgsA-745, pgsB-618, pgsB-650, pgsD-677, and pgsE-606) were a generous gift of Dr. Jeffrey Esko of the University of Alabama. The CHO cells were grown in Dulbecco's modified Eagle's medium (DMEM) with 7.5% fetal calf serum (Intergen Co., Purchase, NY). 32D cells, a myeloid-derived cell line requiring interleukin-3, transfected with cDNA encoding the mouse FGF receptor-1 (flg) form with two immunoglobulin-like domains were described previously (8,26). The transfected 32D-flg cells and 32D cells transfected with the vector alone (32D-neo) were maintained in suspension culture in Iscove minimal essential medium, 10% heat-inactivated fetal calf serum, and 10% medium conditioned by WEHI cells (an interleukin-3-expressing cell line). To prepare conditioned medium, WEHI cells were grown for 48 h, the cells were pelleted, and the medium was filtered, aliquoted, and frozen for future use. CHO cells transfected with cDNA encoding the mouse two-immunoglobulin-like domain form of FGF receptor-1 (CHO-flg, clone 3B) were also described previously (7). The 32D-flg cells express about 3,000 receptors/cell, whereas the CHO-flg cells express about 100,000 receptors/cell as determined by Scatchard analysis of binding of 125 I-FGF-2 (7,26).
Cell Attachment Assays-CHO cells were plated at 5 ϫ 10 5 cells/ 35-mm dish. After 16 h at 37°C, the cells were either fixed with glutaraldehyde or used directly for experiments. For glutaraldehyde fixation, cells were washed two times with phosphate-buffered saline (PBS) and were incubated with 3% glutaraldehyde in PBS for 2 h at 4°C. Glycine was added to a final concentration of 0.1 M to stop the fixation. The fixed cells were washed twice with PBS and used for experiments. For a typical experiment, 5 ϫ 10 5 32D-flg cells in serumfree Iscove medium containing no addition, 10 ng/ml FGF-2, or 10 ng/ml FGF-2 and 10 g/ml heparin were added to washed monolayers of CHO cells. After incubation at 4 or 37°C, medium was removed, and the monolayers were washed twice gently with PBS to remove unattached cells. The medium and PBS washes were combined. The attached cells were removed with a brief wash with PBS containing 10 g/ml heparin. The unattached and attached cells were counted with a Coulter particle counter. Results are presented as (attached cells/(unattached cells ϩ attached cells)) ϫ 100.
Treatment of Cells with Heparinase or Chondroitinase-For some experiments, CHO monolayers were treated with heparinase or chondroitinase ABC. The CHO monolayers were incubated with 2.5 units/ml of Flavobacterium heparinum heparinase I (E.C. 4.2.2.7) (Sigma) or 1 unit/ml Proteus vulgaris chondroitinase ABC (E.C. 4.2.2.4) (Sigma) in PBS containing 0.1 g/ml bovine serum albumin for 2 h at room temperature. Control cultures were incubated for the same period in PBS with 0.1 g/ml bovine serum albumin alone. At the end of the incubation, the cells were washed twice with cold PBS. In some experiments, the cells were fixed with glutaraldehyde as described above and used in attachment assays. In other experiments, they were used directly for attachment assays without fixation.
CHO Attachment Assays-CHO-flg cells or nontransfected CHO-K1 cells were trypsinized and replated at subconfluent density. Twentyfour hours later they were washed twice with Ca 2ϩ -and Mg 2ϩ -free PBS and detached from their dishes after incubation in Ca 2ϩ -and Mg 2ϩ -free PBS containing 10 mM EDTA. Cells were collected by centrifugation; washed twice with PBS; resuspended in DMEM containing 0.15% gelatin and 25 mM HEPES, pH 7.5, with or without 3 units/ml F. heparinum heparinase I (E.C. 4.2.2.7) (Sigma) or 0.005 units/ml F. heparinum heparatinase (E.C. 4.2.2.8) (Seikagaku Kogyo Co., Tokyo, Japan); and incubated for 3 h at room temperature in an end over end mixer. Enzyme units used were defined by the manufacturers. The cells were washed twice with PBS to remove digested heparan sulfate fragments and suspended in DMEM containing 10 mM EDTA. Five hundred thousand CHO cells in suspension were added to each 35-mm dish containing glutaraldehyde-fixed monolayers of wild type CHO-K1 cells or CHO-pgsB-618 mutants in heparan sulfate synthesis prepared as described above. FGF-2 was added to the medium to a final concentration of 0 or 10 ng/ml, and heparin was added to a final concentration of 0 or 10 g/ml. After 2 h at 4°C, nonattached and attached cells were recovered as described above and counted. 125 I-FGF-2 Binding Assays-FGF-2 was radiolabeled with 125 I (DuPont NEN) as described previously (4). Assessment of the binding of 125 I-FGF-2 to low affinity sites (heparan sulfates) on CHO cells was performed as described previously (4). Briefly, cells were incubated for 2 h at 4°C with 10 ng/ml 125 I-FGF-2. Cells were washed twice with cold PBS, and radioactivity bound to low affinity sites was released with two washes with 2 M NaCl in 20 mM HEPES, pH 7.4. As a control, some cultures were incubated with 10 ng/ml 125 I-FGF-2 in the presence of 10 g/ml heparin and processed as above. Binding that could be competed by addition of soluble heparin was considered to be due to 125 I-FGF-2 bound to heparan sulfates. 125 I-Deoxyuridine Incorporation Assays-CHO-K1 cells or CHO-pgsB-618 glycosaminoglycan synthesis mutant cells were plated on 35-mm dishes at 5 ϫ 10 5 cells/dish. After overnight incubation at 37°C, the cells were fixed with glutaraldehyde. Serum-free medium containing 5 ϫ 10 5 32D-flg cells were added to the dishes containing the fixed cells or to dishes containing no cells. The serum-free culture medium contained varying concentrations of FGF-2 in the presence or absence of 10 g/ml heparin. After incubation at 37°C for 22 h, 0.5 Ci/ml 125 Ideoxyuridine (81.4 TBq/mmol, DuPont NEN) was added to each dish, and the cells were incubated for an additional 2 h at 37°C. The medium containing suspended 32D-flg cells was removed, and 32D-flg cells attached to the monolayer were harvested by washing the dishes with PBS containing heparin. The medium and PBS washes from each culture were combined and 32D-flg cells were collected by centrifugation. The cell pellet was resuspended in 10% trichloroacetic acid and filtered over GF/c glass fiber filters. The filters were washed twice with 10% trichloroacetic acid, and radioactivity precipitated on the filters was assayed in a Packard ␥ counter Materials-Recombinant human FGF-2 was a gift from Synergen Inc. (Boulder, CO). Recombinant human FGF-1 was a gift from Dr.

RESULTS
The ability of heparan sulfates from one cell to potentiate the binding of FGF-2 to its receptor on another cell type was examined by co-culturing 32D cells expressing FGF receptor-1 (32D-flg cells), which lack heparan sulfates and grow in suspension, and CHO cells, which express heparan sulfates but have very low levels of FGF receptors and grow attached to the culture dish. To avoid possible confounding effects caused by the metabolism of the test cells, the CHO cells were fixed with glutaraldehyde so that they were not metabolically active but their surface components were preserved. Preservation of heparan sulfates in the fixed cells was confirmed by the fact that the fixed cells bound 80% of the amount of 125 I-FGF-2 on low affinity binding sites as parallel cultures of non-fixed cells. In these co-cultures, the normally suspended 32D-flg cells attached to the CHO monolayer in an FGF-dependent manner.  (Table I).
The attachment was dependent on the presence of FGF receptors on the 32D cells, as only low numbers of 32D cells transfected with the vector alone (32D-neo), which lack FGF receptors, attached to CHO cells either in the absence or in the presence of FGF-2 (Fig. 1A). Attachment of the 32D-flg cells in the presence of FGF-2 could be inhibited by the addition of soluble heparin (Fig. 1A), which prevents the binding of FGF-2 to heparan sulfates (4). These results demonstrate that FGF-2 can promote the attachment of cells expressing FGF receptors to cells expressing heparan sulfates.
To confirm these results, FGF-2-dependent attachment of 32D-flg cells was assessed using a series of well-defined CHO mutants in the synthesis of glycosaminoglycans (35). As shown above, 32D-flg cells attached to wild type CHO-K1 cells in the presence of FGF-2, but little attachment was observed in the absence of FGF-2 (Fig. 1B). Furthermore, heparin alone did not promote attachment and heparin inhibited the attachment normally observed in the presence of FGF-2. The mutant CHO cell line pgsA-745, which lacks the enzyme xylosyltransferase that initiates glycosaminoglycan synthesis, and pgsB-618, which lacks the enzyme galactose transferase I, catalyzing the second step in glycosaminoglycan synthesis, make no glycosaminoglycans and did not support attachment of 32D-flg cells either in the presence or absence of FGF-2 (Fig. 1B). The mutant CHO line pgsD-677, which is deficient in heparan sulfate synthesis but makes supernormal levels of chondroitin sulfate, also did not support FGF-2-dependent attachment of 32D-flg cells. The CHO mutant pgsB-650, which has a 3-fold reduction in glycosaminoglycan synthesis, and pgsE-606, which displays diminished sulfation of heparan sulfate, supported lower levels of FGF-2-dependent attachment of 32D-flg cells. The attachment of 32D-flg cells to these CHO cell mutants in the presence of FGF-2 reflected their capacity to bind FGF-2 (Table II). Thus, the ability of CHO cells to support the FGF-2-mediated attachment of 32D-flg cells depends on their expression of normally sulfated heparan sulfate proteoglycans. This conclusion is further supported by the observation that FGF-2-dependent attachment of 32D-flg cells to wild type CHO-K1 cells was eliminated by pretreatment of the CHO cells with heparinase but not by pretreatment with chondroitinase ABC (data not shown).
The ability of soluble glycosaminoglycans to inhibit attachment of 32D-flg cells was compared. Half-maximal inhibition of the attachment of 32D-flg cells to wild type CHO-K1 cells was obtained with 10 ng/ml heparin and with approximately 200 ng/ml fucoidin or dermatan sulfate (Fig. 2). Chondroitin 4-sulfate, chondroitin 6-sulfate, and keratan sulfate had no effect on FGF-2-dependent 32D-flg cell attachment to CHO cells (Fig. 2). The ability of glycosaminoglycans to block attachment of 32Dflg cells to CHO cells reflected their ability to block FGF-2 binding to heparan sulfates (4).
The ability of FGFs to promote attachment of 32D-flg cells was investigated in more detail. Half-maximal attachment of 32D-flg cells to fixed wild type CHO-K1 cells was obtained with approximately 0.4 ng/ml FGF-2 (Fig. 3A). To determine if other members of the FGF family would support attachment of 32Dflg cells, the cells were incubated at 37°C with fixed CHO-K1 cells in the presence of 10 ng/ml FGF-1, FGF-2, or FGF-4. Significant attachment was obtained in the presence of either FGF-1 or FGF-2, but not FGF-4 (Fig. 3B). Attachment of 32Dflg cells to the CHO cells in the presence of either FGF-1 or FGF-2 could be inhibited by the addition of heparin. The ability of FGF family members to support attachment of 32D-flg cells is consistent with their affinity for FGF receptor-1 (7).
To determine if the relative number of 32D-flg and CHO cells would affect the FGF-2-dependent attachment, varying numbers of 32D-flg cells were incubated with 10 ng/ml in co-cultures with CHO-K1 cells fixed at different densities. Fig. 4A shows that at high densities of CHO cells, a high percentage of the added 32D-flg cells attached, approaching 100% at the highest densities except when very high numbers of 32D-flg cells were added. At low densities of CHO cells, only a low percentage of the 32D-flg cells attached. This data has been replotted in Fig. 4B to demonstrate that when there were three or fewer 32D-flg cells per CHO cell, a high percentage of the added 32D-flg cells attached. At higher ratios, attachment decreased proportionately. This may indicate a saturation of attachment sites on the CHO cells or a physical hindrance between 32D-flg cells crowded over a few CHO cells.
To examine the effect of temperature on 32D-flg cell attachment, the 32D-flg cells were incubated at 4°C with fixed CHO-K1 cells in the presence or absence of 10 ng/ml FGF-2. The same number of 32D-flg cells attached to CHO-K1 cells in the presence of FGF-2 if the co-cultures were incubated at 37 or 4°C (Fig. 5), indicating that cell metabolism is not required for attachment. There was little attachment to the CHO 618 heparan sulfate mutants at either temperature. In addition, treatment of the 32D-flg cells with the protein synthesis inhibitor cycloheximide for 30 min prior to exposure to FGF-2 and throughout the attachment assay had no effect on their ability to attach to CHO-K1 cells, demonstrating that expression of new proteins is not required for attachment. Some cytokines can cause a rapid increase in integrin activity on the cell surface (36). Since attachment through integrins and cadherins is Ca 2ϩ -dependent (36), the ability of the divalent ion chelators EDTA and EGTA to inhibit attachment was investigated. Addition of 10 mM EDTA or EGTA during the 2-h assay had no effect on FGF-2-dependent 32D-flg cell attachment (data not shown). Furthermore, addition of the protein-tyrosine kinase inhibitor genistein did not inhibit the FGF-2-dependent attachment of 32D-flg cells to CHO-K1 cells, suggesting that signaling through the receptor is not involved (data not shown). Finally, addition of antibodies to FGF-2 to attached cells resulted in a rapid detachment of the 32D-flg cells (data not shown), suggesting that attachment directly involves FGF-2 and does not require the induction of other attachment molecules.
The 32D-flg cells attached to living CHO cells as well as glutaraldehyde fixed CHO cells, but the cell-cell association was transient when measured at 37°C. Fig. 5 shows that at 4°C approximately equal numbers of 32D-flg cells attached to fixed or living CHO-K1 cells in the presence of 10 ng/ml FGF-2. However, at 37°C substantially fewer 32D-flg cells attached to living CHO cells than fixed CHO cells. The number of 32D-flg cells attached to living CHO cells in the presence of FGF-2 varied with time. At 37°C, 32D-flg cell attachment to living CHO-K1 cells reached a peak at 2 h, approaching 70% of the level of attachment observed with fixed CHO-K1 cells (Fig. 6). The number of 32D-flg cells attached to living CHO-K1 cells declined after that, reaching values only slightly above control by 24 h. Low levels of attachment to the CHO-pgsB-618 heparan sulfate-deficient mutants were observed independent of whether the cells were fixed or living.
To determine whether this FGF-2-dependent attachment of 32D-flg cells was limited to CHO cells, attachment to bovine capillary endothelial cells, NIH 3T3 cells, and Balb/C 3T3 cells  was investigated. At 4°C in the absence of FGF-2, only small numbers of 32D-flg cells attached to either glutaraldehydefixed or living bovine capillary endothelial cells (Table III). However, in the presence of 10 ng/ml FGF-2, approximately 80% of the added 32D-flg cells bound to the endothelial cells (Table III). As with the CHO cells, attachment of the 32D-flg cells could be blocked by the addition of soluble heparin. Pretreatment of the bovine capillary endothelial cells with heparinase prevented attachment of the 32D-flg cells, showing that the attachment is due to endothelial cell heparan sulfates (data not shown). Similar results were obtained with NIH 3T3 and Balb/C 3T3 cells (data not shown).
The ability of cells expressing both FGF receptors and heparan sulfates to participate in FGF-2-dependent attachment was examined. CHO cells expressing transfected FGF receptor-1 containing two immunoglobulin-like domains (CHO-flg) were detached from their culture dishes with EDTA and were incubated at 4°C in suspension over a glutaraldehyde-fixed monolayer of CHO-K1 cells in the presence of EDTA. Untreated CHO-flg cells did not attach to the fixed CHO cells either in the presence or absence of FGF-2 (Fig. 7, column a). However, if the CHO-flg cells were treated with heparinase prior to their incubation with the glutaraldehyde-fixed CHO-K1 cells, they attached to the monolayer in the presence of 10 ng/ml FGF-2 (Fig.  7, column b). No attachment of heparinase-treated CHO-flg cells was detected in the absence of FGF-2 or if FGF-2 and soluble heparin were added together. Nontransfected CHO cells treated with heparinase did not attach in the presence or absence of FGF-2 (Fig. 7A, column d), showing that attachment was dependent on the presence of FGF receptors on the CHO cells. No attachment of heparinase-treated CHO-flg cells to glutaraldehyde-fixed CHO pgsB-618 glycosaminoglycan mutant cells was observed either in the presence or absence of FGF-2, showing that attachment was dependent on the presence of heparan sulfates (Fig. 7A, column f). The percentage of cells that attached was variable in these experiments, perhaps because of incomplete digestion of heparan sulfates. However, treatment of CHO-flg cells with heparatinase rather than heparinase did not improve their ability to participate in FGF-2dependent attachment (Fig. 7B). Thus, expression of heparan sulfates by the same cell type that expresses FGF receptors limits the interaction of the FGF-2-receptor complex with heparan sulfates on neighboring cells.
To determine whether the interaction of FGF-2 with heparan sulfates on the CHO cells could potentiate its activity, the effect of co-culture on incorporation of 125 I-deoxyuridine into DNA was assessed. When 32D-flg cells were cultivated alone, the addition of FGF-2 at concentrations up to 20 ng/ml stimulated incorporation of 125 I-deoxyuridine into DNA to a minor extent (Fig. 8, open squares). Addition of 10 g/ml heparin along with the FGF-2 increased the stimulation significantly, resulting in a dose-dependent increase in 125 I-deoxyuridine incorporation, with a maximal stimulation about 3.5-fold above control levels with 10 to 20 ng/ml (Fig. 8, filled squares). When 32D-flg cells were cultivated in co-culture with glutaraldehyde-fixed CHO pgsB-618 glycosaminoglycan mutant, they responded to FGF-2 in a manner similar to the cells cultivated alone. Less than 2-fold stimulation of 125 I-deoxyuridine incorporation was obtained with concentrations of FGF-2 up to 20 ng/ml (Fig. 8, open triangles). Addition of heparin along with the FGF-2 resulted in a dose-dependent response to FGF-2 with maximal stimulation at 10 to 20 ng/ml (Fig. 8, filled triangles). However, addition of FGF-2 to 32D-flg cells co-cultured with glutaraldehyde-fixed wild type CHO-K1 cells resulted in a dose-dependent stimulation DNA synthesis in the absence of added heparin with a maximum 3.5-fold increase at 10 -20 ng/ml (Fig. 8, open circles). Addition of soluble heparin along with the FGF-2 did not significantly increase this stimulation (Fig. 8, filled circles). Thus, the heparan sulfates of the wild type CHO-K1 cells were able to substitute for soluble heparin in the potentiation of FGF-2 bioactivity.

DISCUSSION
These results suggest that FGF-2, a potent growth factor, can also act as an attachment factor for suspension cells that express FGF receptors. There have been previous reports that FGF-2 can promote cell attachment (37,38). In these experiments, the attachment of PC-12 cells or endothelial cells to FGF-2 coated on a plastic surface was measured. In addition, PC-12 cells plated on heparin-coated dishes aggregated in the presence of FGF-2 (37). Both adhesion and aggregation could be inhibited by FGF-2 antagonists, suggesting that the receptor was involved in these processes. Furthermore, the ability of cells to bind to FGF-2-coated plastic dishes has also been used as an assay for the cloning of FGF-2-binding molecules (39). In these experiments, a cDNA library from baby hamster kidney cells was introduced into the human lymphoblastoid cell line W1-L2-729 HF 2 . The parental cells did not bind to FGF-2coated dishes, and transfected cells that gained the capacity to bind to FGF-2-coated dishes were selected. Transfected cells that gained the ability to bind were found to express the heparan sulfate proteoglycan, syndecan. Together, these earlier studies showed that both FGF receptors and heparan sulfates could participate in FGF-2-mediated adhesion events. Our observations provide one mechanism by which both FGF receptors and heparan sulfates are involved directly in cell to cell attachment interactions.
Other growth factors, including macrophage colony-stimulating factor, kit ligand, and transforming growth factor-␣, have also been proposed to act as attachment factors (40 -42). The primary translation products of these growth factors are anchored in the plasma membrane by hydrophobic transmembrane sequences. It is proposed that a membrane-anchored growth factor on one cell type can interact with its transmembrane receptor on a second cell type, promoting cell-to-cell interactions. Indeed, the transmembrane forms of transforming growth factor ␣ and macrophage colony-stimulating factor can mediate the attachment of cells bearing specific receptors for those growth factors (40,42). Thus, with these growth factors, there is a two-component linkage, in which a growth factor that is a membrane constituent of one cell binds to a receptor expressed in the membrane of a second cell. The model proposed here for FGF-2 is novel in that it is composed of three components: a binding molecule on one cell (heparan sulfate), a nominally soluble growth factor, and a transmembrane receptor on a second cell type. The model we propose is shown in Fig. 9.
The use of heparan sulfate synthesis mutants of CHO cells and digestion of cell surface heparan sulfates on wild-type CHO cells demonstrated that the 32D-flg cell attachment is heparan sulfate-dependent. The addition of small amounts of soluble heparin inhibited attachment of the 32D-flg cells to wild-type CHO cells, presumably by competing with the cell surface heparan sulfates for binding of FGF-2, thereby preventing FGF-2-mediated bridging between the cells. These results suggest that this type of attachment may be limited to cells like the 32D-flg cells that express FGF receptors but do not produce heparan sulfates. If cell surface heparan sulfates are present on the same cells that are expressing the receptors, their relatively high concentration in the vicinity of the receptors may effectively displace heparan sulfates on other cells from interactions between FGF-2 and receptors. Indeed, when the same receptors were expressed in wild type CHO cells, which do produce heparan sulfates, FGF-2-dependent attachment could not be observed unless the heparan sulfates were removed by heparinase or heparatinase treatment.
However, since many leukemia-derived cells do not produce heparan sulfates (43), natural equivalents of the transfected 32D-flg cells may exist in the primitive blood cell population. Recent evidence that FGF-2 can promote hematopoiesis in 32D-flg cells were added to 35-mm dishes containing no cells (squares), 5 ϫ 10 5 glutaraldehyde-fixed CHO-K1 cells (circles), or 5 ϫ 10 5 glutaraldehyde-fixed CHO-pgsB-618 mutants in heparan sulfate synthesis (triangles). The serum-free culture medium contained the indicated concentrations of FGF-2 (bFGF) in the presence (filled symbols) or absence (open symbols) of 10 g/ml heparin. After incubation at 37°C for 22 h, 125 I-deoxyuridine was added to each dish, and the cells were incubated for an additional 2 h at 37°C. 32D-flg cells were collected by washing the dishes, and 125 I-deoxyuridine incorporated into macromolecules was precipitated with 10% cold trichloroacetic acid, collected on filters, and counted. culture (44 -50) along with evidence that blood cells express FGF receptors (46, 51-53) 2 make this an intriguing possibility. The maturation of hematopoietic cells occurs when the cells are in intimate contact with stromal cells (55). It has been proposed that a specific interaction of primitive hematopoietic cells with bone marrow stromal cells producing the appropriate cytokines might be obtained if the growth factor itself were involved in the binding (56). As noted above, some growth factors are produced as transmembrane proteins. Stem cell factor, also known as the kit ligand, has been shown to be much more potent in stimulating the growth of hematopoietic stem cells when it is expressed as a transmembrane form rather than as a soluble growth factor (41). Thus, it is possible that the attachment of primitive hematopoietic cells to cytokine-producing stromal cells is mediated by anchored growth factors produced by the stromal cells interacting with their specific receptors on the primitive hematopoietic cells. Such an interaction has recently been demonstrated for the transmembrane form of kit ligand (57). Rather than being anchored by transmembrane sequences, some nominally soluble cytokines, such as FGF-2, may be anchored by their association with heparan sulfates in the pericellular matrix. Indeed, heparan sulfatemediated cell attachment may not be limited to FGF-2 and may be a property of a number of heparin-binding growth factors. These observations together with recent demonstrations of signaling through adhesion molecules suggest that the distinction between growth factors and attachment factors may be an arbitrary one.
With a soluble factor, growth factor-mediated interactions might be expected to be less specific than with a membrane anchored factor. However, two properties of FGF-2 might limit attachment to cells producing the growth factor. First, the interaction of FGF-2 with fixed binding sites on cells decreases its diffusibility (20). Thus, the majority of FGF-2 may not diffuse far from the cell that produced it. Second, FGF-2 is rapidly taken up by cells through cell surface heparan sulfates, clearing the surface of active molecules (21)(22)(23). The uptake and metabolism of FGF-2 bound to cell surface heparan sulfates may limit the amount of FGF-2 available for cell attachment interactions. In the experiments presented here, attachment of 32D-flg cells to CHO cells was transient when living CHO cells were used, perhaps as the result of metabolism of the added FGF-2 by the CHO cells. Thus, attachment mediated by an exogenous source of FGF-2 is likely to be transient in vivo too. Stable attachment mediated by FGF-2 may only occur in the vicinity of cells producing FGF-2, so that there is a constant source of growth factor.
In addition to the potential role of FGF-2 in cell attachment, these observations also provide some information on the biochemistry of FGF-2 interactions with FGF receptors. First, they provide additional evidence that a trimolecular complex is formed among FGF-2, heparan sulfate, and FGF receptor. Second, they show that heparan sulfate proteoglycans do not have to be expressed on the same cell as FGF receptors to potentiate the biological activity of FGF-2. Thus, the glycosaminoglycan chains alone and not their location are important for FGF-2 bioactivity. These results are predicted from the previous observations that soluble heparin can substitute for cell-associated heparan sulfates in potentiating the interaction of FGF-2 with its receptor (28) and the recent report that an extracellular matrix proteoglycan can potentiate binding of FGF-2 to its receptor (54). Third, although the FGF receptor is reported to bind to heparin (12,33), the association does not seem to be strong enough to promote attachment of FGF receptor-bearing cells to heparan sulfate-producing cells in the absence of FGF-2. Indeed, studies with purified extracellular domain of FGF receptor-1 have shown that the affinity of FGF receptor-1 for heparin is 200 times lower than the affinity of FGF-2 for heparin (33). Thus, the interaction of heparan sulfates with the FGF-2-receptor complex may be attributed primarily to the affinity of the heparan sulfates for FGF-2.