Specificity for Fibroblast Growth Factors Determined by Heparan Sulfate in a Binary Complex with the Receptor Kinase*

A divalent cation-dependent association between heparin or heparan sulfate and the ectodomain of the FGF receptor kinase (FGFR) restricts FGF-independenttrans-phosphorylation and supports the binding of activating FGF to self-associated FGFR. Here we show that in contrast to heparin, cellular heparan sulfate forms a binary complex with FGFR that discriminates between FGF-1 and FGF-2. FGFR type 4 (FGFR4) in liver parenchymal cells binds only FGF-1, whereas FGFR1 binds FGF-1 and FGF-2 equally. Cell-free complexes of heparin and recombinant FGFR4 bound FGF-1 and FGF-2 equally. However, in contrast to FGFR1, when recombinant FGFR4 was expressed back in epithelial cells by transfection, it failed to bind FGF-2 unless heparan sulfate was depressed by chlorate or heparinase treatment. Isolated heparan sulfate proteoglycan (HSPG) from liver cells in cell-free complexes with FGFR4 restored the specificity for FGF-1 and supported the binding of both FGF-1 and FGF-2 when complexed with FGFR1. In contrast, FGF-2 bound equally well to complexes of both FGFR1 and FGFR4 formed with endothelial cell-derived HSPG, but the endothelial HSPG was deficient for the binding of FGF-1 to both FGFR complexes. These data suggest that a heparan sulfate subunit is a cell type- and FGFR-specific determinant of the selectivity of the FGFR signaling complex for FGF. In a physiological context, the heparan sulfate subunit may limit the redundancy among the current 18 FGF polypeptides for the 4 known FGFR.

The FGF 1 signal transduction system is ubiquitous and is a local mediator of developmental processes in the embryo and homeostasis in the adult (1). Through associations of both FGF and the FGFR ectodomain with heparan sulfate chains of pericellular matrix or transmembrane proteoglycans, the FGFR signaling complex senses perturbation and remodeling of the pericellular microenvironment (1). Heparan sulfate has been implicated in stability of FGF (2), access of FGF to the FGFR complex (3), oligomerization of FGF (4 -6) and FGF-FGFR complexes (1, 4 -8), and conformational repression and activation of FGFR oligomers (1,9,10). Heparin or heparan sulfate exhibits an FGF-independent, divalent cation-dependent high affinity interaction with immunoglobulin module II of selfassociated oligomers of the FGFR ectodomain to form a complex that will bind FGF (8 -10). Divalent cations and heparan sulfate cooperate to maintain the dependence of the FGFR complex on FGF (9), presumably by conformational restriction of the enzyme-substrate relationship between FGFR, which otherwise are self-activating by trans-phosphorylation (1). A large number of genetically distinct FGF polypeptides have emerged, currently 18, for which there are only four FGFR (1). The high ratio of ligands to receptors, the co-expression of multiple FGFs within the same tissue or cells, and the apparent ability of the four FGFR to bind multiple FGFs when heparin is used as the experimental mimic of heparan sulfate raise questions about the redundancy and functional relevance of the large number of FGF homologues. Here, we show that cellular heparan sulfate in a binary complex with the FGFR ectodomain is an FGFR isotype-and cell-specific determinant for selection of the FGF that interacts with the complex. In a physiological tissue context, this may restrict redundancy of the four FGFR for the large repertoire of FGF ligands.

EXPERIMENTAL PROCEDURES
Cell Culture, Preparation of Recombinant Receptors, and Radioreceptor Assays-The culture of primary rat hepatocytes collected by perfusion (11), HepG2 cells (12), and human endothelial cells has been described (13). Preparation and expression of recombinant FGFR isoforms in the baculoviral-Sf9 system (10), preparation of FGFR-specific antibodies (14), and procedures for preparation and iodination, determination of specific binding, and covalent affinity cross-linking of radiolabeled FGF to infected Sf9 cells, mammalian cells, and immobilized purified recombinant FGFR have also been previously described (9,10,15,16). Recombinant FGFR4 was constructed as follows: P1 (5Ј-TTGG-GAATTCCAGCTTGGGTCCCT-3Ј) and P3 (5Ј-GCCAGGTATACGGA-CATCATC-3Ј) were used as sense primers beginning at 42 base pairs upstream of the translational initiation site and 15 base pairs upstream of the coding sequence for the transmembrane domain, respectively, using the polymerase chain reaction. P2 (5Ј-GGATGGAATTCACTTGC-CGGAAGAGCC-3Ј) and P4 (5Ј-CACAGCCTCGAGCCTTGCTCATGT-3Ј) were used as antisense primers ending 124 base pairs downstream of coding sequence for the transmembrane region and 18 base pairs downstream of the stop codon. Paired primers P1P2 and P3P4 were separately applied using HepG2 cell cDNA as template to amplify 1.3-kilobase extracellular and intracellular fragments, respectively. The P1P2 fragment was ligated into pBSK cloning vector at an EcoRI site. The P3P4 fragment coding for the intracellular domain was then ligated into the pBSK vector with the extracellular domain fragment at AccI and XhoI sites to obtain the full-length FGFR4 cDNA. Sequence was verified and then the FGFR4 cDNA was cloned either into mammalian expression vector pcDNA1neo (Invitrogen Co., SanDiego, CA) at BamHI and XhoI sites for expression on A431 cells or into transfer plasmid pVL1393 (Invitrogen Co., San Diego, CA) at SmaI sites for preparation of the baculovirus. * This work was supported by United States Public Health Service Grants DK40739 and DK35310 from the NIDDK, National Institutes of Health, and Grant CA59971 from the NCI, National Institutes of Health. 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  Assay for Activity of Heparan Sulfate by FGF Receptor Kinase Complementation-Heparin or heparan sulfate activity was assessed by binding to recombinant FGFR receptor kinase ectodomain, and then the ability of the binary complex to support the binding of radiolabeled FGF was assessed (8,9,16). About 5 ϫ 10 5 Sf9 cells expressing recombinant FGFR were extracted with 200 l of 1% Triton X-100 in PBS. Lysates containing approximately 5 g of FGFR were mixed with either 4 g of anti-FGFR1 (17A3) monoclonal antibody or 16 l of anti-FGFR4 rabbit antiserum (17) and 80 l of protein A-Sepharose beads. The immobilized FGFR was washed with 1 M NaCl in PBS and then extensively with PBS. After division into 8 -12 portions, beads were incubated with standard heparin (Sigma, 174 USP units/mg) or extracts or partially purified fractions containing heparan sulfate for 1 h at room temperature in 25 mM Tris-HCl (pH 7.4), 0.15 M NaCl, 10 mM MgCl 2 , and 0.01% CHAPS (binding buffer). After washing three times with PBS, 125 I-FGF (4 ng/ml in 250 l) was added for 1 h. Beads were then washed with PBS, 1 mM disuccinimidyl suberate was added for 15 min, and extracts from the beads were counted and analyzed by SDS-PAGE and autoradiography.
Scatchard Analysis-Dissociation constants (K d ) were determined by Scatchard analyses. Assays using isolated FGFR were performed under conditions that promote maximum dependence on external heparin/ heparan sulfate for FGF binding (9). Procedures to determine specific binding and to distinguish binding to complexes of FGFR and heparin/ heparan sulfate from FGF bound to heparin/heparan sulfate sites that are not in sufficient proximity to FGFR to yield covalent FGF-FGFR complexes have been described in detail (9,15,16). Immobilized binary FGFR complexes were prepared by incubation of immobilized FGFR and 1 g/ml heparin or 2 g/ml cellular heparan sulfate followed by removal of unbound carbohydrate. Analysis was performed in a range of 1.5-10 ng/ml radiolabeled FGF. Nonspecific binding was less than 10% of total at all concentrations determined in the presence of 30-fold unlabeled FGF. K d and receptor site number were estimated by linear least squares analysis. Except where indicated, the K d values cited in the text are the mean Ϯ S.D. of three separate determinations.
Preparation and Partial Purification of Cellular Heparan Sulfate Proteoglycan Fragments-Subconfluent cell cultures described in the text in 75-cm 2 flasks in batches of 10 were treated with 25 g/ml trypsin and 0.02% EDTA in PBS (5 ml/flask) for 15 min at 4°C. The solution containing material released from the monolayers was acidified with 0.1% trifluoroacetic acid, dialyzed against water, and freeze-dried. One flask of cells per batch was labeled by the addition of 20 Ci/ml 35 SO 4 overnight prior to extraction for use as a tracer. The freeze-dried powder was reconstituted in PBS and applied to a Biosil SEL-400 HPLC gel filtration column (exclusion limit, 10 6 kDa, 300 ϫ 7.8 mm, Bio-Rad). The radioactivity ( 35 S) of each fraction was monitored by liquid scintillation, and the activity of each fraction was screened by ability to complement recombinant insect cell-derived FGFR1 for binding of 125 I-FGF-1. In some cases, fractions were prescreened for binding activity using intact Sf9 cells expressing the recombinant FGFR prior to extraction and immobilization. Activity was expressed as a stimulation index, which was the binding in the presence of fraction divided by that in absence of the test fraction. Fractions of coincident radioactivity and FGF binding activity were pooled and applied directly to a Bio-Gel TSK-DEAE-5-PW anion-exchange HPLC column. Application of a gradient of NaCl from 0.15 to 1.0 M revealed a peak of coincident binding activity and 35 S radioactivity at 0.5 M NaCl, which was dialyzed and freeze-dried prior to reconstitution. Carbohydrate content was measured by the carbozole method and standardized to the dry weight of heparin. Carbohydrate equivalent to about 50 g of heparin containing 10 g of protein was obtained from 10 8 HepG2 cells.
A portion of the reconstituted partially purified material with uronic acid content equivalent to 2.5 g of standard heparin was diluted with 2 ml of 200 mM borate buffer (pH 9.0), mixed with 100 l of a fresh solution of 5 mg/ml Bolton-Hunter reagent in Me 2 SO and incubated for 3 h on ice. The fraction was then dialyzed against PBS and then iodinated to a specific activity of 1 ϫ 10 5 cpm/ng with 1 mCi of Na 125 I using the Chloramine T method. Free iodine was eliminated by centrifugal filtration (Millipore Ultra free-MC, molecular weight 30,000 cutoff, Millipore Products Division, Bedford, MA) with a recovery of TCAinsoluble material of 40%.
Assay and Partial Purification of Radiolabeled Heparan Sulfate Proteoglycan by FGFR or FGF Affinity-An FGFR-affinity matrix for heparan sulfate was prepared using the membrane-anchored recombinant FGFR ectodomain expressed in Sf9 cells following extraction and immobilization on antibody-protein A beads or immobilization of constructs of the FGFR ectodomain fused to glutathione S-transferase downstream of the transmembrane domain to GSH-beads (9). Heparan sulfate was applied in conditions described above for preparation of FGF-binding binary complexes of FGFR and heparin/heparan sulfate. Elution was carried out with increasing concentrations of NaCl or 10 mM EDTA (9).
Noncovalent affinity matrices of FGF-1 and FGF-2 (5 g) bound to Cu-chelated beads (500 l) (Chelating-Sepharose Fast Flow, Amersham Pharmacia Biotech, Uppsala, Sweden) were prepared in the presence of 100 g of heparin and 10 mM MgCl 2 in PBS containing 1% Triton X-100 for 2 h at 4°C. The beads were washed with 2 M NaCl in PBS and incubated with another 5 g of FGF, and the procedure was repeated. About 85% of the applied FGF could be immobilized under these conditions; this percentage was unaffected by the presence or absence of heparin and the subsequent 2 M NaCl wash. Separate analyses revealed that elution of the column with 10 mM imidazole removed about 75% of the bound FGF after multiple runs and that the activity of the recovered FGF was equal to unbound FGF in receptor binding and cell growth assays. The binding of radiolabeled heparin or 125 I-labeled heparan sulfate proteoglycan to the affinity matrix was dependent on the presence of FGF. Nearly 100% of bound heparin/heparan sulfate was eluted by salt or imidazole. Bound heparin or heparan sulfate was routinely eluted from FGF-1 columns with 1.0 M NaCl and from FGF-2 columns with 1.5 M NaCl.

RESULTS
Cell Type-dependent Specificity of FGFR4 for FGF-1-Cultured rat hepatocytes, which express exclusively FGFR4 (Fig.  1A), exhibit a mitogenic response to specifically FGF-1 (11,18). Hepatocytes exhibit about 5000 FGFR sites/cell that specifically bind FGF-1 with K d of about 60 pM (11,18). The mitogenic response to or binding of FGF-2 in hepatocytes was undetectable (11,18) (Fig. 1C). In contrast, hepatocyte-like human hepatoblastoma (HepG2) cells ectopically express isoforms of FGFR1, FGFR2, and FGFR3 in addition to FGFR4 (Fig. 1B), bind FGF-2, and exhibit 180,000 FGFR sites/cell that bind FGF-1 (12). However, an analysis with antiserum specific for FGFR1 and FGFR4 revealed that the FGFR1 fraction in HepG2 cells bound both FGF-1 and FGF-2, but the FGFR4 fraction bound only FGF-1 (Fig. 1D). This absolute specificity of FGFR4 for FGF-1 in liver parenchymal cells was similar to the specificity of recombinant FGFR4 for FGF-1 reported in monkey kidney epithelial cells (19). However, it differed from FGFR4 in mouse fibroblasts (20), muscle cells (21), or leukemia cells (22), which bound both FGF-1 and FGF-2. Similar to the latter three mammalian cells, a reconstituted complex of heparin and isolated recombinant FGFR4 derived from insect cells bound FGF-1 and FGF-2 equally (Fig. 2). Separate Scatchard analyses of the binding of FGF-1 and FGF-2 to binary complexes prepared with 1 g/ml heparin yielded apparent K d values of 243 Ϯ 30 and 247 Ϯ 28 pM, respectively. This was similar to the recombinant FGFR1-heparin complex, which bound both FGF-1 and FGF-2 with K d values of 253 Ϯ 18 pM and 197 Ϯ 61 pM, respectively. Thus both FGFR1 and FGFR4 have an intrinsic ability to bind both FGF-1 and FGF-2, but a factor that either supports or prevents the binding of FGF-2 to specifically FGFR4 differs among different types of mammalian cells.

Selectivity of Recombinant FGFR4 for FGF-1 Is Restored in Reconstituted Epithelial Cells and Abrogated by Depression of
Heparan Sulfate-To test the hypothesis that a cell type-dependent factor confers selectivity of FGFR4 for FGF-1, we reconstituted FGFR-deficient epithelial cells (A431 from a squamous cell carcinoma) with 1 ϫ 10 5 FGFR1 or 5 ϫ 10 4 FGFR4 and then examined the binding of FGF-1 and FGF-2 (Fig. 3). Scatchard analysis indicated that FGF-1 and FGF-2 bound to cells expressing FGFR1 with K d values of 104 Ϯ 13 pM and 142 Ϯ 9 pM, respectively, whereas cells expressing FGFR4 bound only FGF-1 with a K d of 243 Ϯ 24 pM. The binding of FGF-2 to FGFR4 was below the limits of detection by both Scatchard analysis and covalent affinity cross-linking (Fig. 3A).
To determine whether cellular heparan sulfate played a role in specificity of the binding of FGF in the reconstituted A431 cells, we applied sodium chlorate, which inhibits sulfation of heparan sulfate, or heparinase, which modifies heparan sulfate chains. The treatment of the A431 cells expressing FGFR4 with sodium chlorate reduced the binding of FGF to cellular heparan sulfate sites for FGF-1 and FGF-2 to 15 and 35%, respectively, of untreated controls. Surprisingly, the treatment resulted in the binding of FGF-2 to FGFR4 (Fig. 3B). Similar to chlorate, treatment of cells with heparinase also resulted in the gain in the binding of FGF-2 in the cells reconstituted with FGFR4 (Fig. 3B) (8 -10). Total RNA was extracted with RNAZol (Biotech Labs Inc., Houston, TX). First strand cDNA was synthesized from the RNA using a random hexamer (Promega Corp., Madison, WI) and reverse transcriptase (SuperscriptII, Life Technologies, Inc.). Polymerase chain reaction was carried out in 100 l containing 2 l of 0.1 g/ml cDNA, 0.5 l of Tag DNA polymerase (5 units/ml, Promega), and 1 l each of 5Ј and 3Ј primers (0.1 g/ml) of total volume for 94°C for 1 min, 55°C for 2 min, and 72°C for 3 min for 40 cycles. Primers were human and rat FGFR1 (5Ј-GGCAAGGAATTCAAACCT-GAC-3Ј and 5Ј-CATCACGGCTGGTCTCTC TTC-3Ј); human FGFR2 (5Ј-AACGGGAAGGAGTTTAAGGAG and 5ЈGGAGCTATTTATCCCCGAG-TG-3Ј (IIIb) or 5Ј-TGGGAGAACTGTCAACCATGC-3Ј (IIIc)); rat FGFR2 (5Ј-TTCATCTGCCTGGTCTTGGTC-3Ј and 5Ј ATGGAGCCGC-TTCTCCATCTT-3Ј (IIIb) or 5Ј-ATGGAGCCACCTCACTATCGT-3Ј (IIIc)); human FGFR3 (5Ј-AACGGCAGGGAGTTCCGCGGC-3Ј and 5Ј-CTTGGGGCCCGTGAACACGCAGCC-3Ј (IIIb) or 5Ј-GCAGCACCACC-AGCCACGGAGAGT-3Ј (IIIc)); rat FGFR3 (5Ј-AATGGCAAAGAATTC-CGAGGGCAG-3Ј and 5Ј-CTTGGGGCCCGTGAACACGCAGCC-3Ј (IIIb) or 5Ј-GCAGCACCACCAGCCACGCAGAGT-3Ј (IIIc)); human FGFR4 (5Ј-GATGGACAGGCCTTTCATGGG-3Ј and 5Ј-TGCTGCGGTC-CATGTGGGGTCCTC-3Ј); and rat FGFR4 (5Ј-GATGGACAGGCCTTG-CACGGG-3Ј and 5Ј-GGTTGTTGTTGTCCACGTGAG-3Ј). Products were analyzed with 1% agarose gel electrophoresis in the presence of ethidium bromide and visualized under UV light. C and D, FGF-1 and FGF-2 binding to hepatocytes (C) and HepG2 cells (D). Subconfluent cells in 6-cm 2 dishes were incubated with 125 I-FGF-1 or 125 I-FGF-2 at 4 ng/ml in 1.5 ml of binding buffer (NaHCO 3 -free RITC80 -7 medium containing 100 g/ml bovine serum albumin) for 1 h at room temperature. After incubation, the cells were washed stepwise with PBS, heparin (250 g/ml) in PBS, and PBS and exposed to 1 mM disuccinimidyl suberate for 15 min at room temperature. Cells were extracted with 0.5 ml of 1% Triton X-100 in PBS, and the lysate, after clarification, was mixed with 5 l of the indicated FGFR-specific antisera (8 -10) and 20 l of protein A beads. The beads were washed sequentially with PBS, 1 M NaCl in PBS, and then PBS followed by extraction with SDS-PAGE buffer and analysis by SDS-PAGE and autoradiography. The radiolabeled band at lower apparent molecular mass is a result of proteolysis of the full-length FGFR at a site near the transmembrane domain (14). absence of cells and the pericellular matrix. Cell-derived heparan sulfate proteoglycan (HSPG) was released into the medium by treatment of cells with trypsin and partially purified by molecular filtration and anion exchange chromatography as described under "Experimental Procedures." The activity of HSPG fractions was assessed by their ability to form a complex with the immobilized FGFR ectodomain after removal of unbound material and then by the ability of the resultant binary complex to bind radiolabeled FGF-1 or FGF-2 (8,9,16). The activity co-purified with macromolecules metabolically labeled with sodium [ 35 S]sulfate, which ran ahead of the bulk of the protein on gel permeation chromatography (Fig. 4A). Activity and [ 35 S]sulfate also co-purified during ion exchange chromatography slightly behind the main absorption peak at A280 (Fig. 4B). The partially purified fraction was radiolabeled with iodine as described under "Experimental Procedures" and analyzed by SDS-PAGE and autoradiography (Fig. 4C). On 7.5% gels, about 90% of radiolabeled material failed to penetrate the gel, whereas the rest of the material migrated in a band with mean apparent mass of about 75 (Ϯ 10) kDa (Fig. 4C, left  panel). Treatment with heparinase resulted in a broad band of labeled material from 45 to 130 kDa. Treatment with Pronase destroyed 100% of the radiolabeled material. Analysis on 15% gels after treatment with heparinase (Fig. 4C, right panel) revealed that about 70% of the treated material migrated at apparent mass of 35 kDa that ran off the 7.5% gels. Separate experiments not shown here revealed that the FGF complementation activity survived treatment with both Pronase and chondroitinase A, B, C but was destroyed by treatment with heparinase or nitrous acid. Inclusion of 8 M urea in the SDS-PAGE gels did not change the electrophoretic pattern of the indicated bands.
The iodinated partially purified heparan sulfate proteoglycan ( 125 I-HSPG) was employed to further characterize direct binding to FGFR. Under FGF binding assay conditions, about 200 pg of the radiolabeled HSPG bound per ng of immobilized FGFR1. One hundred % of the 125 I-HSPG that bound to immobilized FGFR1 was displaced by a 100-fold excess of heparin or treatment with 1 M NaCl. As reported previously, introduction of FGF into complexes of FGFR and heparin/heparan sulfate increased resistance of the bound heparin/heparan sulfate to displacement with external heparin, but not with 0.5 M NaCl (16). Displacement of 125 I-HSPG bound to FGFR1 or FGFR4 was undetectable by either FGF-1 or FGF-2 at concentrations below 10 ng/ml. This was 2.5 times the standard amount of 125 I-FGF used in binding assays and the highest concentration employed in the Scatchard analyses. Concentrations above 50 ng/ml of either FGF-1 or FGF-2 displaced 50% of the bound HSPG. Preliminary affinity chromatography experiments using immobilized FGFR1 columns (see under "Experimental Procedures") indicated that a maximum of 20% of the 125 I-HSPG can be retained on FGFR1 or FGFR4 at 0.15 M NaCl and that over 90% of the retained material, which elutes at 0.65 M NaCl, is in the high molecular weight fraction shown in Fig. 4A.
Additional experiments using the complementation assay with purified FGFR1 indicated that over 80% of the activity of the partially purified HSPG from HepG2 cells that bound to either FGFR1 or FGFR4 eluted from the immobilized FGFR between 0.25 and 0.65 M NaCl. This represented 7 and 5% of total material bound at 0.15 M NaCl to FGFR1 and FGFR4, respectively. About 65% of the FGFR1-or FGFR4-bound material that eluted between 0. 25  Fractions under the solid bar from the molecular filtration chromatography in A were pooled and applied to a TSK-DEAE5PW column. The pool contained 5.9% of the A280 and 70% of the 35 S. After loading and washing with PBS, a NaCl gradient from 0.15 to 1.0 M, indicated by the solid line, was applied at a flow rate of 0.5 ml/min. Fractions of 1.0 ml were collected, a 50-l portion was counted by liquid scintillation, and a 10-l portion was assayed for binding activity in the FGFR complementation assay. Activity was expressed as a stimulation index (SI), which was the amount of 125 I-FGF-1 bound to FGFR preincubated with a fraction divided by binding to an FGF incubated with a blank fraction. Fractions under the solid bar were pooled, de-salted, and used for FGFR complementation assays. The pool contained 2.6% of the A280 and 30% of the 35 S that was applied to the column. C, iodination and analysis of the partially purified HSPG fractions. A portion of the pooled fraction that was equivalent to 2.5 g of heparin based on uronic acid content from A and B above was radiolabeled with 125 I as described under "Experimental Procedures" and analyzed by SDS-PAGE on 7.5 and 15% gels followed by autoradiography. A 6.25-ng sample of the radiolabeled material (10 5 cpm) was subjected to treatment with 1. 25  HepG2 cells, 2 and 14%, respectively, bound to affinity columns of FGF-1 and FGF-2. Of the fraction that bound to immobilized FGF-1 that was eluted by 1.0 M NaCl, 18 and 14% bound to FGFR1 and FGFR4, respectively, at 0.25 M NaCl. Of the fraction that bound to immobilized FGF-2, 18 and 9% subsequently bound to FGFR1 and FGFR4, respectively. The results of purification of the HSPG fraction by combinatorial affinity chromatography using immobilized FGFR, FGF, and anti-thrombin 2 will be reported elsewhere.
Scatchard analysis indicated that FGFR1 incubated with 0.2 g/ml of the HSPG fraction supported FGF-1 and FGF-2 binding to isolated FGFR1 with a K d of 60 pM (mean of two experiments) and 50 pM (mean of two experiments), respectively. The same fraction supported FGF-1 binding to isolated FGFR4 with a K d of 140 pM (mean of two experiments). However, the binding of FGF-2 was insufficient to assign a K d by Scatchard analysis. No radiolabeled FGF-2 could be detected by covalent affinity cross-linking, even when FGFR4 was incubated with more than 2 g/ml of the HSPG fraction (Fig. 5, A and B). These results indicated that unlike heparin, cell-derived heparan sulfate restores the specificity of isolated FGFR4 for FGF-1 to that observed in the intact cell.
Finally, we isolated HSPG fraction by the same procedure from human umbilical vein-derived endothelial cells in which FGFR1 is a resident isoform. In the absence of exogenous heparin, the cell line binds and responds mitogenically to FGF-2, but only poorly to FGF-1 (13,18). In contrast to the HSPG fraction from HepG2 cells, the fraction from the endothelial cells supported the binding of FGF-2 to both FGFR1 and FGFR4 when tested in the FGFR complementation assay (Fig.  5, C and D). In contrast to heparin (Fig. 2) and liver cell HSPG (Fig. 5, A and B), the endothelial cell HSPG was deficient in support of the binding of FGF-1 to binary complexes of both FGFR1 and FGFR4.

DISCUSSION
The FGF signal transduction system is a ubiquitous and local mediator of embryonic development and homeostasis in adult tissues. It is a tripartite oligomeric complex composed of the transmembrane tyrosine kinase and heparan sulfate chains from pericellular matrix proteoglycan (1, 8 -10). Pericellular matrix heparan sulfate plays multiple roles in assembly and activation of the FGFR complex. These roles include stability of FGF (2), access of FGF to the FGFR complex (3), and oligomerization of the FGFR complex through both the FGFR ectodomain and FGF (4 -10). Heparin and heparan sulfate exhibit a divalent cation-dependent interaction with apparent K d of about 10 nM to a specific sequence domain in Ig module II of the FGFR ectodomain (8 -10). This is thought to restrict the trans-phosphorylation between self-associated FGFR kinases in the absence of FGF (8 -10). In addition, the bound heparin/ heparan sulfate is required and sufficient for the high affinity interaction of FGF with the binary complex in the absence of soluble heparin/heparan sulfate (8,9,23). These results support a model in which an activating FGF docks into inactive oligomeric complexes of FGFR and heparan sulfate chains from the pericellular matrix. In this report, we utilized this model of the stepwise assembly of the ternary complex of heparan sulfate-FGF-FGFR to determine whether the heparan sulfate subunit of the complex contributes to discrimination between FGFs.
Because there were conflicting reports concerning the discrimination of FGFR4 relative to FGFR1 for FGF-1 and FGF-2 among cell types (19 -22), we employed FGFR4 and FGFR1 as experimental prototypes. FGFR4 is the sole isotype of the four known FGFR genes expressed in normal liver parenchymal cells and in that context, FGFR4 exhibits an absolute specificity for FGF-1 relative to FGF-2. In contrast, ectopically expressed FGFR1, which is co-expressed with FGFR4 in differentiated hepatoma cells, bound FGF-1 and FGF-2 equally. Yet reconstituted complexes of heparin and recombinant FGFR4, which were removed from cell membranes, pericellular matrix, and other cellular factors, bound both FGF-1 and FGF-2 in a manner similar to FGFR1. Expression of recombinant FGFR4 back into epithelial cells by transfection restored the specificity 2 W. L. McKeehan, X. Wu, and M. Kan, manuscript in preparation.
FIG. 5. Cellular heparan sulfate proteoglycan complexed with FGFR determines specificity for FGF. The ability of FGFR-HSPG complexes to bind FGF-1 and FGF-2 was compared using both total binding determined by gamma-counter (A and C) and covalent crosslinking (B and D). A and B, specific binding of FGF-1 to complexes of isolated liver cell heparan sulfate and recombinant FGFR4. Recombinant FGFR1 and FGFR2 were expressed in insect cells, isolated, and immobilized on antibody-protein A beads, and analyses were performed as described in Fig. 2, except that FGFR complexes were prepared with the indicated amount of liver cell HSPG prepared as described under "Experimental Procedures" instead of heparin. Lanes 1-3 in each set in B show the binding of the indicated FGF to FGFR with no added HSPG or heparin (N), to complexes of FGFR and liver cell HSPG and complexes of FGFR and heparin (H), respectively. B and D, specific binding of FGF-2 to complexes of isolated endothelial cell heparan sulfate proteoglycan and recombinant FGFR1 and FGFR4. C shows the binding of FGF-1 and FGF-2 to FGFR1 (R1) or FGFR4 (R4) after incubation and removal of complexes from unbound material with no carbohydrate (N), with 2 g/ml HSPG isolated from human umbilical vein endothelial cells (EC), or with 1 g/ml heparin (H). D shows the corresponding covalent cross-linking analysis of the immobilized complexes. Data sets in A and B and those in C and D are representative of seven and five reproductions, respectively, in which at least three independent preparations of HSPG and radiolabeled FGF-1 and FGF-2 were employed. The specific experiments indicated were performed with the same radiolabeled preparations of FGF-1 and FGF-2.
of FGFR4 for FGF-1. This suggested that the specificity was conferred by a cellular co-factor that was present in parenchymal cell lineages represented by the hepatocytes and HepG2 hepatoma cells, kidney epithelial cells (19), and the A431 squamous epithelial cells. The fact that the sulfation-inhibitor chlorate or treatment with heparinase abrogated the selectivity of FGFR4 for FGF-1 in intact cells suggested that cellular heparan sulfate was responsible.
Isolated and partially purified extracts containing cellular HSPG from the surface of the liver parenchymal tumor cell HepG2 substituted for heparin in formation of binary complexes with isolated recombinant FGFR1 and FGFR4 from insect cells. Similar to those formed with heparin, the binary complexes were then capable of binding FGF. Moreover, the isolated liver cell HSPG fraction restored the specificity of FGFR4 for FGF-1 to that observed in the liver parenchymal cells while continuing to support the binding of both FGF-1 and FGF-2 to FGFR1. This confirmed that the liver cells expressed a distinct HSPG that, when complexed with FGFR4, permits the high affinity docking of only FGF-1 into the complex. This explained the specificity of hepatocytes for FGF-1.
Finally, we demonstrated that the discrimination for FGF conferred by HSPG when complexed to FGFR4 was cell-specific and, therefore, not a consequence of a general interference of FGFR4-bound HSPG with access of FGF-2 to the active site or general interference of FGF-2 with the binding of HSPG to FGFR4. In contrast to the liver cell HSPG, but similar to heparin, partially purified HSPG from endothelial cells supported the binding of FGF-2 to reconstituted complexes of both FGFR1 and FGFR4. Unlike liver cell HSPG and heparin, binary complexes of both FGFR1 and FGFR4 that formed with endothelial cell HSPG failed to efficiently bind FGF-1. This was consistent with the selective requirement for exogenous heparin for the binding and mitogenic response of endothelial cells to FGF-1 (13,18). The results suggest that a deficiency in composition of an HSPG subunit of the FGFR complex may limit the response of some endothelial cells to FGF-2.
Cell or tissue specificity of heparan sulfates in respect to affinity for and differential effects on activity of FGF-1, FGF-2, and FGF-7 has been demonstrated (24 -30). We conclude that FGFR-and cell type-specific pericellular matrix heparan sulfates complexed to the FGFR kinase ectodomain as an integral subunit may limit binding and activation of the FGFR signaling complex to one or a subset of the FGF ligands. In tissue and in a physiological context, this may limit the redundancy of the current 18 FGF polypeptide activators for the four FGF receptor kinases. Recently, we have demonstrated that only the fraction of commercial heparin and cellular heparan sulfate that binds anti-thrombin and inhibits Factor Xa activity can form a binary complex with the ectodomain of the FGFR kinase, and only the fraction enriched by recombinant FGFR affinity matrix exhibits anti-coagulant activities. 2 Thus, the requirement for anticoagulant heparan sulfate and divalent cations (9) distinguishes the requirement for formation of the binary complex with all four FGFR from the simpler requirements for the FGFR-independent binding to FGF. Currently, it is unclear whether FGFR1 and FGFR4 select heparan sulfate chains that discriminate between FGFs, or the selectivity for FGF is mediated by the same chain when bound to different FGFR. Although there appears to be a small fraction selected by both FGFR1 and FGFR4 that cannot bind the other, our preliminary results indicate that most of the liver cell-derived HSPG that binds either FGFR1 or FGFR4 can bind the other FGFR. On the one hand, the cross-reactivity may be due to multiple chains on the same HSPG fragment or multiple, but FGFR-specific, binding sites within a single chain. On the other hand, the composite complex formed by the same binding site within the same heparan sulfate chain in the two different FGFR may form the basis of the specificity for FGF. Clarification of the structural basis for the results requires purification of the specific HSPG underlying the activities through combinatorial affinity chromatography using FGFR, FGF and antithrombin columns, identification of the core protein sequence, and structural characterization of the heparan sulfate chains.