INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Various types of growth factors bind to heparin and heparan sulfate. It has been suggested that fibroblast growth factors (FGFs) are stored and protected from proteolytic degradation in the extracellular matrix and on the cell surface by interaction with heparan sulfate proteoglycans, which thereby serve as a reservoir of the growth factors (9, 26, 30, 33, 44, 45). Heparan sulfate proteoglycans not only play such a modulatory role but are involved in the binding of FGF to its high-affinity receptors as an essential component. Thornton et al. (42) demonstrated that heparin potentiates the mitogenic activity of crude preparations of FGF-1. Further studies of the interaction of FGF-1 with heparin suggested that the potentiation of mitogenic activity is caused by the ability of heparin to increase the affinity of FGF-1 for its receptor (35). It has recently been shown that free heparin or heparan sulfate glycosaminoglycan (HSGAG) is required for high-affinity binding of FGF-2 to FGF receptor (FGFR) 1 expressed in heparan sulfate-deficient CHO cells (50). Based on these findings, a dual-receptor model has been proposed, in which a complex composed of a classical protein-type receptor and a low-affinity glycosaminoglycan-type receptor mediates the cellular responses to FGF (15).

The FGF family of growth factors currently consists of 15 members: FGF-1 (acidic FGF [aFGF]), FGF-2 (basic FGF), FGF-3 (INT-2), FGF-4 (HST; kaposi-FGF), FGF-5, FGF-6 (HST-2), FGF-7 (keratinocyte growth factor [KGF]), FGF-8 (AIGF), FGF-9 (GAF), FGF-10, FGF-11 (FHF-1), FGF-12 (FHF-2), FGF-13 (FHF-3), FGF-14 (FHF-4), and FGF-15 (1, 37, 47). FGFs have been shown to function as mitogens, angiogenic factors, neurotrophic factors, oncogenes, motogens, and key factors for several developmental processes. FGFs show differential binding to FGFRs (28). FGFRs are members of a transmembrane tyrosine kinase receptor superfamily. Four distinct FGFR genes, FGFR1, FGFR2, FGFR3, and FGFR4, have been identified. A variety of isoforms are generated by alternative splicing of these FGFR genes, which contributes to the functional diversity of this receptor family (13). While the extracellular domain of FGFRs determines ligand-binding specificity (24), the C-terminal domain regulates the basal activity of the receptor (20). The extracellular domain is composed of two or three immunoglobulin (Ig)-like loops. In FGFR1 and FGFR2, Ig-2 and Ig-3 loops have been shown to bind FGF-1 and FGF-2, respectively. The KGF receptor, an isoform of FGFR2 generated by alternative splicing, binds KGF/FGF-7 at its Ig-3 loop and FGF-1 at its Ig-2 loop, but not FGF-2. The KGF receptor is referred to here as FGFR2-K, while FGFR2, which binds FGF-1 and FGF-2 with high affinity but not KGF/FGF-7, is designated FGFR2-B to distinguish it from the KGF receptor. FGFR2-B and FGFR2-K differ only in the second half of their Ig-3 domains, which are encoded by alternative exons (13, 24). While both Ig-2 and Ig-3 loops play major roles in FGF binding, no growth factors are known to bind to the Ig-1 loop. In addition to these Ig domains, both the FGFR2-B and FGFR2-K isoforms have variants containing or lacking an acidic box, a domain consisting of a stretch of acidic amino acids, and surrounding sequences in their ligand-binding domains. The functional significance of the acidic box in FGFRs was not known.

We report here the identification of a unique isoform of FGFR2 which exhibited a diffuse band with a much larger molecular mass than other isoforms. This receptor was modified by glycosaminoglycans, and the modification was regulated by alternative splicing of the receptor mRNA that encodes the Ig-1 loop and the acidic domain. Moreover, we show that both the HSGAG moiety and the receptor core protein are required for high-affinity binding of FGF-1 and for enhanced and sustained signal transduction. Thus, this receptor exhibits characteristics of a dual-receptor system consisting of a cell surface heparin-like molecule and FGFR tyrosine kinase in a single molecule.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Engineering of FGFR2 cDNA constructs. A rat FGFR2-B isoform containing two Ig loops and an acidic domain has been designated WT (21) and used to construct FGFR2-B variants. A rat FGFR2-K isoform (41) was used to construct FGFR2-K variants. Deletions and mutations were created by three steps of PCR. The vector used for expression of all the constructs was pCEV27 (25). In the first step, a forward primer containing the sequence 5' to the BamHI site of pCEV27 (5'-GGATCATTTAGGTGACACTAT-3') and a reverse primer containing the desired deletion or mutation (primer 1R [described below]) were used. The second step utilized a forward primer containing the desired deletion or mutation (primer 2F [described below]) and a reverse primer containing the sequence downstream of the BamHI site of FGFR2 (5'-TTCAGCCATGACTACT-3'). The third step employed the forward primer of the first step and the reverse primer of the second step with equal amounts of the products from the first and second steps as a template. PCRs were 35 cycles of 1 min at 95°C, 2 min at 55°C, and 2 min at 72°C. The products of these three steps were digested with BamHI and ligated with the BamHI-digested original construct containing pCEV27 and a part of FGFR2. The orientations of FGFR2 variants in the vector were determined by several restriction enzyme digestions. The nucleotide sequences of the final products were confirmed by DNA sequencing. The following FGFR2-B and FGFR2-K constructs were generated and designated as shown in Fig. 1: 3 Loop, FGFR2 with three Ig domains and an acidic domain; WT, FGFR2 with two Ig domains (Ig-2 and -3) and the acidic domain with an acidic box (a short stretch of acidic amino acids [Fig. 1]) and a normal Ser-Ser-Gly amino acid sequence at the N terminus of the acidic box; A, FGFR2 with two Ig domains and the acidic domain lacking the acidic box; ASG, FGFR2 with two Ig domains and the acidic domain containing a mutated Ala-Ser-Gly sequence at the N terminus of the acidic box; SAG, FGFR2 with two Ig domains and the acidic domain, containing a mutated Ser-Ala-Gly sequence at the N-terminus of the acidic box; and AD, an FGFR2 isoform without the acidic domain. The primers 1R and 2F for each construct are listed below in a 5'-3' orientation. Primers 1R: 3 Loop, ATCTGTGTCGTCCTCGTCATC (template, three-loop FGFR2-K); A, GTCTTCGGAGCTTCCAGATGAGATGGCATC; ASG, CGTCATCTCCAGATGCGATGGCATCTTCTGG; SAG, CGTCATCTCCAGCTGAGATGGCATC; and AD, TACGGTGCTCTTTCTGGTTCTAAAGTGGT. Primers 2F: 3 Loop, CACAGATGCCATCTCATC; A, ATCTCATCTGGAAGCTCCGAAGACTTTGTC; ASG, CCAGAAGATGCCATCGCATCTGGAGATGACG; SAG, GATGCCATCTCAGCTGGAGATGACG; and AD, CACTTTAGAACCAGAAAGAGCACCGTACTGG.

View larger version (19K): [in this window] [in a new window]   FIG. 1.   FGFR2 constructs used in the present studies. The transmembrane and intracellular domains are the same in all of the constructs and the same as those of the clones previously reported (41). FGFR2-B and FGFR2-K contain different sequences only in the second half of the Ig-3 loop, which is shown by thick arcs. SSG, ASG, or SAG is the amino acid sequence immediately N-terminal to the acidic box. SSG represents the wild type sequence, while the other two were generated by localized mutagenesis. The thick lines at the ends of the N termini represent the common sequence including the signal peptide. TM, transmembrane domain; A, acidic box; AD, acidic domain (an amino acid sequence that is encoded by an exon and contains both SSG and the acidic box).

Reporter assays. CHO-Ki cells in a 10-cm-diameter plate were transiently cotransfected with 5 µg of FGFR2-K cDNA (AD type or WT type) in pCEV27 and 3.5 µg of a reporter plasmid containing three tandem copies of SRE.mutL and the luciferase-coding sequence as described previously (23) by using 35 µl of Lipofectamine (Life Technologies)/plate. SRE.mutL is the minimal serum response factor (SRF)-binding site of the c-fos promoter lacking a ternary complex factor (TCF) binding site (12). The cells were plated in 24-well plates 24 h after transfection and kept in a medium containing 2% fetal bovine serum. Twenty-four hours after the cells were plated, they were exposed to 10 ng of FGF-1/ml for various periods, washed with phosphate-buffered saline, and lysed in cell culture lysis buffer (Promega) (200 µl/well). Luciferase activity was determined with a Lumat LB 9501 luminometer (Berthold) with automatic injection. For each sample, 20 µl of lysate and 100 µl of luciferase substrate (Promega) were used. The values were normalized for both protein content in each lysate and induction by 10% fetal calf serum, which was regarded as maximal stimulation, in each transfectant. The data were analyzed statistically by analysis of variance (ANOVA) with StatView software (Abacus Concepts, Inc.).

Detection of c-fos mRNA expression. A mutant CHO-Ki cell line carrying the pgsA-745 mutation, which is deficient in glycosaminoglycan modification initiation, was provided by J. D. Esko and designated CHO-745 (8). CHO-745 cells were stably transfected with pCEV27 expression vector alone or a construct containing the FGFR2-K isoform with two Ig domains (AD [Fig. 1]) as described previously (41). CHO-745 transfectants were maintained in Ham's F-12 medium supplemented with 10% fetal bovine serum and 450 µg of G418/ml. The cells were first plated in the same medium in six-well plates for time course studies of c-fos mRNA expression, and the medium was changed overnight to Ham's F-12 containing 3% fetal bovine serum. The cells were stimulated by FGF-1 (final concentration, 10 ng/ml) for the indicated period in the presence (final concentration, 0.1, 1, or 10 µg/ml) or absence of heparin. The stimulation by FGF-1 was terminated by lysing the cells, and total RNA was extracted by using RNA STAT-60 (TEL-TEST B, Inc., Friendswood, Tex.) according to the company's protocol. Total RNA (10 µg) was fractionated by 1% formaldehyde agarose gel, and Northern blotting analysis was performed by conventional methods (34). The DNA probes used for detecting c-fos and cyclophilin mRNAs were kind gifts from T. Curran (5) and P. E. Danielson (7), respectively. A PhosphorImager (Molecular Dynamics) was used for autoradiography, and the image was imported into NIH Image software for quantitative analysis of band intensities.

Affinity-labeling of FGFR2 with ¹²⁵I-FGF-1. Affinity labeling was performed as described previously (32). Cells were incubated at 4°C for 3.5 h with 10 ng of ¹²⁵I-FGF-1/ml in the presence or absence of an excess amount of unlabeled FGF-1 (R&D Systems Inc., Minneapolis, Minn.) in 1 ml of NIH Eagle's no. 2 medium (without bicarbonate) containing 20 mM HEPES, pH 7.3, 0.3% bovine serum albumin, 0.5 mM MgSO₄, and 1 mM CaCl₂. FGF receptors were then cross-linked with radiolabeled ligand by exposing the cells to 0.3 mM disuccinimidyl suberate (Pierce Chemical Co., Rockford, Ill.) in cold HEPES-buffered saline (HBS) for 20 min, and the reaction was terminated by adding 250 µl of quenching buffer containing 10 mM Tris-HCl, pH 7.5, 200 mM glycine, and 2 mM EDTA. After the reaction medium was aspirated, the cells were washed three times in cold HBS and harvested in HBS containing 1 mM phenylmethylsulfonyl fluoride (Pierce), 1 mM N-ethylmaleimide, 0.1 mM EDTA, 10 µg of antipain/ml, 10 µg of pepstatin/ml, 10 µg of leupeptin/ml, and 10 µg of aprotinin/ml (Calbiochem, San Diego, Calif.). The cells were separated from the buffer by centrifugation and solubilized in 60 µl of cold lysis buffer containing 10 mM Tris-HCl, pH 7.4, 0.25% Nonidet P-40 (Pierce), and the protease inhibitor mixture. Samples were boiled and fractionated under reducing conditions in a discontinuous Laemmli buffer system (18) by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) with an SDS-7% polyacrylamide resolving gel and a 3% stacking gel. The gels were fixed and stained with 0.13% Coomassie blue R-250 in 45% methanol, 45% water, and 10% acetic acid overnight, destained with several changes of the same solution, dried under vacuum, and then exposed to a PhosphorImager screen overnight for autoradiography.

Immunoblotting of FGFR2 and phosphotyrosine. NIH 3T3 cells were stably transfected with various FGFR2 cDNA constructs or the expression vector alone (pCEV27), and the transfected cells were selected with G-418 as described previously (25). For detection, NIH 3T3 or CHO-Ki cells transfected with the FGFR2-B or FGFR2-K cDNA construct in pCEV27 expression vector were plated on 100-mm-diameter plates, incubated with FGF-1 or the vehicle alone, and lysed in 300 µl of lysis buffer after the cells became 90 to 95% confluent. The lysis buffer contained 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 10 mM sodium pyrophosphate, 50 mM sodium fluoride, 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride, 10 µg of aprotinin/ml, and 10 µg of leupeptin/ml. The lysate was mixed well and centrifuged at 14,000 rpm for 10 min at 4°C in an Eppendorf S417C microcentrifuge. The supernatant was separated, and its protein concentration was determined. For immunoprecipitation, the supernatant containing 1 mg or 500 µg of protein was incubated with an anti-FGFR2 antibody (SA571; 1:100), a rabbit polyclonal antibody raised against the intracellular domain of FGFR2, or 10 ml of agarose-conjugated anti-phosphotyrosine monoclonal antibody (no. 16-101C; UBI) at 4°C for 2 h. For immunoprecipitation of FGFR2, the mixture incubated with an anti-FGFR2 antibody was then incubated with 40 µl of GammaBind G-plus (Pharmacia) at 4°C for 1 h with constant rotation. The GammaBind G-plus immune complex or agarose-conjugated immune complex was washed with 1 ml of lysis buffer three times by mixing them, centrifuging at 14,000 rpm for 2 to 3 min in an Eppendorf S417C microcentrifuge, and aspirating the supernatant. The final pellet was resuspended in 2× sample buffer, and the supernatant was boiled and fractionated by discontinuous SDS-7% PAGE under reducing conditions. The fractionated protein was electroblotted onto Immobilon P. FGFR2-B or FGFR2-K was detected by autoradiography after the membrane was sequentially incubated with a 1:500 dilution of the anti-FGFR2 antibody and a 1:1,000 dilution of ¹²⁵I-protein A as described previously (21) or by using ECL Western blotting detection reagents (Amersham) after incubating the membrane with a 1:500 dilution of the anti-FGFR2 antibody and a 1:2,000 dilution of horseradish peroxidase-linked anti-rabbit Ig antibody (NIF824; Amersham) according to the manufacturer's instructions. Phosphotyrosine was detected with ECL Western blotting detection reagents after incubating with a 1:2,000 dilution of anti-phosphotyrosine monoclonal antibody (no. 05-321; UBI) and a 1:2,000 dilution of horseradish peroxidase-linked anti-mouse Ig antibody (Amersham). Whenever necessary, the same blotted membrane was reprobed with other antibodies after the antibodies used for the first detection were removed with 0.5 M glycine buffer (pH 2.5) and 0.05% Tween 20.

Heparitinase and/or chondroitinase ABC digestion. Cell lysates from affinity-labeling samples were diluted at least 50-fold in a 0.1 M Tris acetate buffer and treated with heparitinase (code 100703; lot no. E95601; Seikagaku America, Rockville, Md.) and/or chondroitinase ABC (code 100332; lot no. KE95702; Seikagaku America) (final concentration, 10 mIU/ml) in 0.1 M Tris acetate buffer, pH 7.3, at 37°C for 1 h. In the case of treatment with heparitinase alone, shark cartilage chondroitin sulfate (catalog no. 2307; Calbiochem) was added to a final concentration of 50 µg/ml to protect the samples from digestion by chondroitinases possibly contaminating the heparitinase preparation. The enzyme-digested samples were dialyzed against water overnight at 4°C and concentrated under vacuum before being loaded onto a gel.

Analysis of MAP kinase activation. For detecting extracellular signal-regulated kinase (ERK) mitogen-activated protein (MAP) kinase, NIH 3T3 cells were cultured in 6-well plates 3 days before the experiments and incubated in serum-free medium for 24 h. The cells were exposed to 10 ng of FGF-1/ml for the periods indicated and lysed in the same lysis buffer used for immunoprecipitation of FGFR2. The cell lysates (50 µg of protein) were fractionated by discontinuous SDS-12.5% PAGE under reducing conditions and blotted onto Immobilon-P (Millipore) membrane. For protein, either rabbit anti-phosphorylated ERK MAP kinase polyclonal antibody (-phosMAPK), which recognizes tyrosine 204-phosphorylated MAP kinase, or rabbit anti-MAP kinase polyclonal antibody (-MAPK) was used as the first antibody, followed by an alkaline phosphatase-conjugated anti-rabbit IgG according to the manufacturer's instructions (PhosphoPlus MAPK antibody kit [catalog no. 9100]; New England Biolabs) Chemiluminescence was detected after the addition of a substrate of alkaline phosphatase (CDP-Star).

Induction of DNA synthesis. DNA synthesis was measured as reported previously (31). Ninety-six-well microtiter plates (no. 3596; Falcon) were precoated with human fibronectin (Collaborative Research) at 1 µg/cm² prior to being seeded with NIH 3T3 cells. The cells were incubated without serum for 48 h before samples were added. Incorporation of [³H]thymidine into DNA was monitored during a 6-h period beginning 16 h after the addition of the samples.

Saturation binding. Saturation binding studies with ¹²⁵I-FGF-1 were performed according to the methods described previously (2), and the binding data were analyzed with Ligand software (27). ¹²⁵I-FGF-1, which was prepared by the Bolton-Hunter method and purified by heparin-Sepharose affinity chromatography (specific activity, 20 to 50 µCi/µg), was purchased from DuPont NEN (North Billerica, Mass.). The biological activity of the ¹²⁵I-FGF-1 as measured by the thymidine incorporation assay described above always showed more than 90% of the potency of the unlabeled ligand.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Heparin or heparan sulfate glycosaminoglycan is required for high-affinity binding of FGF to FGFR2 and c-fos mRNA induction. The addition of heparin is required for high-affinity binding of FGF-2 to FGFR1 expressed in mutant CHO cells defective in HSGAG synthesis (50). To examine whether FGF-1 and FGFR2 show the same requirements in similar systems, we stably expressed an isoform of FGFR2 containing Ig-2 and Ig-3 loops and the K exon sequence in the Ig-3 domain (FGFR2-K; AD [Fig. 1]) in a mutant cell line, CHO-745, deficient in glycosaminoglycan chain initiation (8). Scatchard analysis of the FGF-1 saturation binding to the cells expressing FGFR2-K showed curvilinear plotting in the presence of heparin (Fig. 2A), suggesting that multiple forms of complex with different affinities can be formed between the ligand and the receptor by adding heparin, as reported previously (29). To assess the effect of heparin on the dissociation constant (K_d), we applied linear regression to the Scatchard plots. Heparin increased the binding affinity of FGF-1 to FGFR2-K in a dose-dependent fashion with K_ds of 2250, 450, and 200 pM at 0, 1, and 10 µg of heparin/ml, respectively. The increase of binding affinity by the addition of 10 µg of heparin/ml was more than 10-fold higher than the basal level (Fig. 2A). Addition of chondroitin sulfate (up to 10 µg/ml) in place of heparin did not affect the binding affinity. Cells expressing vector alone showed barely detectable levels of FGF-1 binding either with or without heparin (data not shown).

View larger version (35K): [in this window] [in a new window]   FIG. 2.   Effect of heparin on FGF-1 (aFGF) binding to its receptor and on c-fos mRNA induction in glycosaminoglycan-deficient CHO mutant cells (CHO-745). (A) Saturation binding studies of 125I-FGF-1 with or without heparin in the binding buffer (inset) and Scatchard analyses of the binding studies. CHO-745 was stably transfected with the AD isoform of FGFR2-K (CHO-745FGFR2-K) and used for binding studies. Heparin concentrations are as follows: , 10 µg/ml; , 1 µg/ml; , 0 µg/ml. (B and C) Northern blot analysis of c-fos and cyclophilin (cyclo) mRNA expression. Northern blotting was performed with total RNA (10 µg/lane) extracted from CHO-745 cells transfected with either vector alone (CHO-745-Vector) or FGFR2-K (CHO-745-FGFR2-K). The cells were stimulated by FGF-1 (final concentration, 10 ng/ml) for the periods indicated in the presence (final concentration, 0.1, 1, or 10 µg/ml) or absence of heparin. The same membrane was separately probed by c-fos cDNA and cyclophilin cDNA probes. (D) Densitometric analyses of the bands detected by Northern blot analysis of c-fos mRNA in CHO-745-Vector (left) and CHO-745-FGFR2-K (right). Heparin concentrations are as follows: , 10 µg/ml; , 1 µg/ml; , 0.1 µg/ml; , 0 µg/ml.

A high-molecular-mass FGFR2 species is created by glycosaminoglycan modification. During the course of expression cloning of oncogenes activated in osteosarcoma cells, we identified an FGFR2 variant with a large molecular mass generated by C-terminal fusion to a novel protein, FRAG1 (21). We have converted this constitutively activated FGFR2 to a wild type FGFR2 by replacing the rearranged C-terminal domain with the normal sequence and designated it FGFR2-WT. While the cDNA for the same FGFR2-WT isoform had already been cloned from stomach cancer and HeLa cells (3, 11), neither the receptor expression nor the glycosaminoglycan modification has been characterized for this FGFR2 isoform. When expressed in NIH 3T3 cells and analyzed by SDS-PAGE following cross-linking with ¹²⁵I-FGF-1, this wild type receptor still showed a diffuse band with a mass larger than that predicted from the primary amino acid sequence (Fig. 3A). Since the broad band was suggestive of glycosaminoglycan modification of the receptor, we digested the cell surface receptor with heparitinase and chondroitinase ABC, which degrade glycosaminoglycan moieties. The enzyme treatment shifted the molecular mass of the receptor to ~130 kDa, which after subtraction of the mass of the ligand, corresponded well to the reported size of FGFR2 with Ig-2 and Ig-3 loops (Fig. 3A). These results indicated that the diffuse high-molecular-mass band of FGFR2-WT in these transfectants arises from glycosaminoglycan modification.

View larger version (65K): [in this window] [in a new window]   FIG. 3.   Affinity labeling (A, B, and C) and immunoblotting (D and E) of the FGFR2 isoforms expressed in NIH 3T3 cells. (A) The FGFR2-B isoform, FGFR2-B-WT, was expressed in NIH 3T3 cells and cross-linked with 125I-FGF-1. The cell lysates after cross-linking were treated with heparitinase plus chondroitinase ABC (+) or vehicle alone () and fractionated by discontinuous SDS-7% PAGE under reducing conditions. Molecular mass markers are shown in kilodaltons on the left. pCEV27 was the vector used for expression of all of the constructs shown hereafter. (B) Various FGFR2 constructs (Fig. 1) were expressed in NIH 3T3 cells, cross-linked with 125I-FGF-1, and analyzed as described above. (C) Glycosaminoglycan-degrading enzyme treatments of the affinity-labeled FGFR2-B isoforms and mutants. Lanes: 1, no digestion; 2, digestion by heparitinase and chondroitinase ABC; 3, digestion by heparitinase alone; 4, digestion by chondroitinase ABC alone. (D) Various forms of FGFR2-B expressed in NIH 3T3 cells were immunoprecipitated by anti-FGFR2 polyclonal antibody, SA571, from the same amount of protein from cell lysates and examined by immunoblotting with the same antibody. (E) Treatment of the immunoprecipitated FGFR2-B molecules with heparitinase and/or chondroitinase ABC. Lanes: V, transfected with pCEV27 alone and no digestion; 1, no digestion; 2, digestion by heparitinase and chondroitinase ABC; 3, digestion by heparitinase alone; 4, digestion by chondroitinase ABC alone.

Glycosaminoglycan modification is determined by an alternative exon encoding the acidic domain. The extracellular domain of the wild type FGFR2 modified by glycosaminoglycan contained the acidic domain and Ig-2 and Ig-3 loops, as shown in Fig. 1. Two FGFR2 isoforms, FGFR2-B and FGFR2-K, have different amino acid sequences in the second half of the Ig-3 loop which are encoded by alternative exons, B and K, respectively (24) (Fig. 1). We have found that the FGFR2-K isoforms containing Ig-2 and Ig-3 loops, but lacking the acidic domain, did not exhibit these glycosaminoglycan modifications (24, 25). Since the glycosaminoglycan-modified FGFR2 contained the sequence derived from the B exon in its Ig-3 loop, the receptor is the FGFR2-B isoform. Therefore, we reasoned that the B-exon-derived sequence, or the acidic domain, determines the modification. To test these possibilities, we constructed a set of FGFR2-K isoforms containing (WT) or lacking (AD) the acidic domain (Fig. 1) and expressed these constructs in NIH 3T3 cells. As shown in Fig. 3B, FGFR2-K-WT showed mobility similar to that of FGFR2-B-WT by SDS-PAGE, suggesting that the B-exon-derived sequence is not responsible for the modification. Although we previously found that the Ig-3 loop of FGFR2-K (K-exon sequence) is weakly modified by glycosaminoglycans in a parathyroid cell line (41), this sequence was not involved in the major glycosaminoglycan modification detected here. In contrast, an FGFR2-K isoform lacking the acidic domain (FGFR2-K-AD) was identified as a discrete band with a much smaller molecular mass. These results indicated that the acidic domain is involved in the glycosaminoglycan modification.

The Ser-Gly-acidic box motif is essential for glycosaminoglycan modification of FGFR2. Glycosaminoglycans modify core proteins at the Ser residue of a putative attachment site of Ser-Gly (8, 49). Close inspection of the amino acid sequence of the wild type two-loop FGFR2-B (FGFR2-B-WT) revealed a Ser-Ser-Gly sequence in the acidic domain. To examine whether this sequence is involved in the modification, we mutated the cDNA sequences of the amino acids in this site of FGFR2-B, as well as FGFR2-K (Fig. 1). The mutant receptors were stably expressed in NIH 3T3 cells, affinity-labeled with ¹²⁵I-FGF-1, and subjected to treatments with heparitinase, an HSGAG-degrading enzyme, and/or chondroitinase ABC, a chondroitin sulfate glycosaminoglycan (CSGAG)-degrading enzyme. As shown in Fig. 3C, the broad band of FGFR2-B-WT was shifted to a smaller discrete band by the double digestion with heparitinase and chondroitinase ABC. Single digestions with either of the enzymes resulted in similar but incomplete shifts of the band, indicating that each FGFR2-B-WT molecule was modified by either HSGAG or CSGAG. Mutation of the first Ser residue in the Ser-Ser-Gly sequence did not alter the level of glycosylation by HSGAG or CSGAG (Fig. 3C and D). In contrast, mutation of the second Ser residue to Ala in the Ser-Ser-Gly sequence abolished the modification by glycosaminoglycans (Fig. 3C), indicating that this Ser residue is essential for the modification. Thus, an FGFR2 molecule can be modified by either HSGAG or CSGAG at the single Ser residue adjacent to Gly. The population of FGFR2 molecules modified by CSGAG was predominant over that modified by HSGAG by about twofold.

The first Ig domain of FGFR2 inhibits glycosaminoglycan modification. The common Ig-2 loops of FGFR2-B and FGFR2-K are known to bind FGF-1, and their distinct Ig-3 loops bind FGF-2 and KGF/FGF-7, respectively (24). However, no ligand has been identified that binds to the Ig-1 loop. The FGFR2-B and FGFR2-K isoforms with three Ig domains contain the acidic domain between the Ig-1 and Ig-2 domains. To examine the effect of Ig-1 on glycosaminoglycan modification of the receptor, we performed similar enzyme digestion experiments after cross-linking the receptor with ¹²⁵I-FGF-1. Interestingly, the FGFR2-B isoform with three Ig domains was not modified by either HSGAG or CSGAG (Fig. 3C), despite the presence of the modification site in its acidic domain. Immunoprecipitation and immunoblotting studies with anti-FGFR2 antibodies before and after enzyme digestion also showed essentially the same results (Fig. 3E). These results indicate that the presence of the Ig-1 loop inhibits the glycosaminoglycan modification of FGFR2-B at the site within the acidic domain. Similar results were obtained with the FGFR2-K isoform containing three Ig domains (data not shown). Introduction of a mutation in a site responsible for the disulfide bond of the Ig-1 loop essentially abolished this negative effect of the modification (data not shown). Therefore, the Ig-1 domain appears to act as an inhibitor of glycosaminoglycan modification of both FGFR2-B and FGFR2-K.

Glycosaminoglycan-modified FGFR2 increases and sustains ligand-induced phosphorylation of the receptor and its substrates. The presence of covalently linked glycosaminoglycan chains in a two-loop FGFR2 containing the acidic domain suggested that this receptor species might function as a high-affinity receptor without the requirement of other cell surface heparin-like molecules. We reasoned that, in wild type CHO cells or in NIH 3T3 cells, the glycosaminoglycan-modified FGFR2 may be more potently stimulated by FGFs than unmodified receptor species, which require cell surface heparin-like molecules to function as high-affinity receptors.

View larger version (44K): [in this window] [in a new window]   FIG. 4.   Effect of HSGAG modification on FGF-1-induced phosphorylation of FGFR2 and FGFR substrates. (A and B) Receptor autophosphorylation. Cells stably transfected with vector alone (pCEV27) or various forms of FGFR2-B (WT, SAG, A, or 3 Loop) were cultured and exposed to 5 ng of FGF-1/ml. Cell lysates containing 500 µg of protein were immunoprecipitated with anti-FGFR2 antibody and detected by either anti-phosphotyrosine (PY) (A) or anti-FGFR2 (FGFR2) (B) antibody as noted on the right. For experiments involving immunoprecipitation of FGFR2, treatments with heparitinase and chondroitinase ABC were performed after immunoprecipitation. The membrane was first used for detection of phosphotyrosine followed by detection of FGFR2 after the antibodies were stripped off. (C) Phosphorylation of receptor kinase substrates. Cell lysates containing 500 µg of protein were immunoprecipitated with PY and detected by the same antibody. The locations of the molecular mass markers are shown on the left. The arrow denotes the major FGFR substrate of 95 kDa.

Glycosaminoglycan modification of FGFR2 results in sustained MAP kinase activation. Two parallel signaling pathways converge at the level of the serum-responsive element (SRE) to induce c-fos expression (12). Signals through Ras activate ERK MAP kinase pathways, which in turn activate TCF, which binds the SRE. Another signal through the Rho family of small GTPases results in activation of SRF, which binds the 3' half of SRE, designated SRE.mutL. Therefore, we first investigated the effect of FGF-1 on MAP kinase activation in NIH 3T3 cells expressing FGFR2-B-WT, its glycosaminoglycan modification site mutant (SAG), or the three-loop form with anti-phosphorylated ERK MAP kinase antibody (Fig. 5A and B). All of these FGFR2-B transfectants showed increased ERK activity compared to cells transfected with vector alone. Interestingly, NIH 3T3 cells expressing the FGFR2-B isoform with glycosaminoglycan modification (WT) showed sustained ERK MAP kinase phosphorylation of both p42 and p44 over a 4-h period, compared to those expressing mutant (SAG) or natural (3 Loop) receptors lacking HSGAG modification. The protein expression levels of all three FGFR2-B constructs were comparable (Fig. 3D). It is noteworthy that the cells expressing FGFR2-B with three Ig domains (3 Loop), which is not modified by glycosaminoglycan, exhibited the same time course as those expressing the mutant two-loop FGFR2-B (SAG) lacking the glycosaminoglycan modification site. These findings indicate that the activation of the MAP kinase pathway is sustained in the FGFR2 isoforms with glycosaminoglycan modification.

View larger version (53K): [in this window] [in a new window]   FIG. 5.   Effect of glycosaminoglycan modification on signal transduction through FGFR2. (A) Activation of ERK MAP kinase by FGF-1 in NIH 3T3 cells expressing different FGFR2 forms. The cells were exposed to FGF-1 (10 ng/ml) for the periods indicated, and the cell lysate at each time point was examined by immunoblotting with either rabbit anti-phosphorylated ERK MAP kinase polyclonal antibody (-phosMAPK) or rabbit anti-MAP kinase polyclonal antibody (-MAPK). -phosMAPK can detect the active MAPK phosphorylated at tyrosine 204, while -MAPK detects the total MAP kinase. pCEV27 was the expression vector used for expressing all the FGFR constructs in NIH 3T3 cells. (B) Quantitative analysis of the intensities of bands detected in panel A. The band intensities were measured by using NIH Image software after the image was scanned with Adobe Photoshop. The same experiments were repeated twice with similar results. , WT; , SAG; , 3 Loop; , pCEV27. (C) Induction of TCF-independent transcription. CHO-Ki cells were transiently cotransfected with the FGFR2-K expression vector (AD-type or WT) and the SRE.mutL reporter construct and exposed to FGF-1 for the periods indicated, and then the luciferase activity was examined as described in the text. The mean value at time 0 was subtracted from all the subsequent values at each time point in both transfectants. The value at each time point represents the mean + standard deviation of hexaplicate measurements. Responses at 60 and 90 min were significantly different in the two transfectants (P < 0.01 by ANOVA). (D) Western blotting analysis of FGFR2-K isoforms immunoprecipitated from the same transfectants used for panel C. The same amount of protein (1 mg) of cell lysates was used for immunoprecipitation from both transfectants. Gel fractionation and blotting were performed with the samples before (lane 1) and after (lane 2) enzyme treatment (heparitinase plus chondroitinase ABC). The arrowheads indicate comparable expression levels of FGFR2-K isoforms in the two transfectants.

TCF-independent transcription is also potentiated by glycosaminoglycan-modified FGFR2. Although it is known that FGFR2 can activate SRE through phosphorylation of TCF by ERK MAP kinase, whether FGFRs can activate TCF-independent transcription is still unknown. To examine TCF-independent activation of SRE, we transfected CHO-Ki cells with FGFR2-K isoforms either containing or lacking the acidic box together with a luciferase reporter DNA construct containing SRE.mutL (12), which can be activated by SRF but not by TCF. Exposure of the transfectants to FGF-1 resulted in a three- to fivefold increase of luciferase activity over the basal level (data not shown), indicating that FGFR2-K can activate TCF-independent transcription upon FGF-1 stimulation. Cells expressing the FGFR2-K isoform with the glycosaminoglycan modification domain (FGFR2-K-WT) showed a higher peak and a more sustained time course than those expressing an FGFR2-K isoform without the glycosaminoglycan modification site (FGFR2-K-AD) (Fig. 5C). In separate experiments, cells expressing FGFR2-K-WT showed a luciferase activity increased over the basal level for more than 4 h, while luciferase activity decreased to the basal level within 3 h in FGFR2-K-AD or FGFR2-K-3Loop (data not shown). Similar amounts of receptor protein were expressed in these two transfectants, as assessed by Western blotting after immunoprecipitation from equal amounts of cell protein (Fig. 5D). Cells transfected with a vector lacking SRE.mutL did not show luciferase induction upon exposure to FGF-1 (data not shown). Cells expressing FGFR2-K with three Ig domains (3 Loop), which is not modified by glycosaminoglycan, exhibited a time course of luciferase induction similar to that of the cells expressing the AD isoform of FGFR2-K. Essentially the same results were obtained with the corresponding FGFR2-B constructs (data not shown). These results indicate that glycosaminoglycan modification of FGFR2 promotes sustained activation of TCF-independent transcription as well as the ERK-mediated pathway.

Glycosaminoglycan modification potentiates induction of DNA synthesis by FGF-1. Since glycosaminoglycan-modified FGFR2 exhibited sustained activation of ERK activity and TCF-independent transcription, we examined the effects of the glycosaminoglycan modification of FGFR2 on a long-term biological effect of FGF by measuring DNA synthesis. NIH 3T3 cells expressing FGFR2-B with glycosaminoglycan modification (WT) showed a dose-dependent response of DNA synthesis upon FGF-1 stimulation between 0.1 and 20 ng/ml (Fig. 6A). This response was almost completely abolished in cells expressing FGFR2-B without glycosaminoglycan modification (SAG), especially at low concentrations of FGF-1, indicating that the receptor-linked glycosaminoglycan potentiates the long-term response upon FGF-1 stimulation. In cells expressing the FGFR2-B modified by CSGAG alone (A), DNA synthesis was not induced by FGF-1 at concentrations lower than 1 ng/ml, and only moderate induction was observed at higher concentrations. Cells transfected with vector alone exhibited a response similar to that by SAG-transfected cells (data not shown). In contrast, platelet-derived growth factor (PDGF) induced DNA synthesis similarly in all of these transfectants. When the DNA synthesis stimulation by FGF-1 was compared with that by 3 ng of PDGF/ml in the same transfectants, the cells expressing FGFR2-B-WT showed significantly higher values than did those expressing FGFR2-B-SAG or FGFR2-B-A at 1 and 1.5 ng of FGF-1/ml (Fig. 6B). The expression levels of all receptor species used in these studies were similar, as detected by immunoblotting with anti-FGFR2 antibodies following immunoprecipitation of the receptors from equal amounts of cell protein (Fig. 6C). All of these results indicate that HSGAG modification of FGFR2 promotes more potent FGF-1 stimulation of DNA synthesis and that cells expressing the modified receptor are stimulated even at low concentrations of the ligand.

View larger version (40K): [in this window] [in a new window]   FIG. 6.   Effects of modification site mutations in FGFR2 on DNA synthesis. (A) DNA synthesis in NIH 3T3 cells transfected with different constructs of FGFR2-B. DNA synthesis was examined by thymidine incorporation in response to FGF-1 (aFGF) or PDGF ranging from 0.1 to 30 ng/ml. Each point represents a mean of duplicate measurements. The same experiment was repeated three times with similar results. The cells transfected with vector alone exhibited a response similar to that exhibited by SAG-transfected cells. The ligand concentrations used for these experiments were 0.1, 0.3, 1, 3, 10, and 30 ng/ml. (B) The enhancement of DNA synthesis was further examined by hexaplicate measurements at 1.0 and 1.5 ng of FGF-1/ml. Each value was divided by the mean response value (of hexaplicate measurements) to 3 ng of PDGF/ml in the corresponding transfectant. The responses by the glycosaminoglycan-modified form of FGFR2 (WT) were significantly higher (P < 0.01) than those by other forms when analyzed by ANOVA. (C) Expression of the FGFR2-B variants. FGFR2 was immunoprecipitated from the same amount of cell protein (1 mg) after enzyme digestion (right) or without digestion (left) and detected by immunoblotting. HS'ase, heparitinase; Ch'ase, chondroitinase ABC.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

A number of FGFR2 isoforms generated by alternative splicing have been reported (13). These include FGFR2-B and FGFR2-K, with three Ig loops, and those with two Ig loops containing or lacking the acidic domain. However, the functions of the Ig-1 loop and the acidic domain were not known. We have identified a major glycosaminoglycan modification site in both FGFR2-B and FGFR2-K molecules. The modification site has been localized to the Ser residue of the Ser-Gly sequence immediately N-terminal to the acidic box in these FGFR2 isoforms. The sequence requirement of core proteins for glycosaminoglycan modification, especially by HSGAG, has been studied in several proteoglycans (16, 36, 51, 52). A Ser-Gly motif flanked by acidic residues appears to be required for HSGAG modification (51, 52). The localization of the modification site in FGFR2 that we report here is consistent with these results. In FGFR2, the Ser-Gly sequence and the acidic box are encoded by a single exon (13, 48). The acidic box in FGFR2 appears to be a prerequisite for HSGAG modification of the Ser residue in this Ser-Gly motif immediately N-terminal to the acidic box, while it is not absolutely required for CSGAG modification. These results suggest that the Ser-Gly-acidic box motif is crucial in promoting glycosaminoglycan attachment to FGFR2.

We have shown that only the two-loop isoforms containing the acidic domain can be modified by glycosaminoglycan. It was surprising that FGFR2-B and FGFR2-K, with three Ig loops, showed no modification by either HSGAG or CSGAG, even though they contain the required amino acid structural motif. This finding strongly suggests that the Ig-1 domain has an inhibitory effect on the glycosaminoglycan modification. The Ig-1 loop appears to be encoded by a single exon (13). An isoform containing the Ig-1 loop but not the acidic domain is not known. Based upon these results, we have concluded that the glycosaminoglycan modification of FGFR2 at the Ser residue N-terminal to the acidic box is regulated by alternative splicing: positively and negatively by exons encoding the acidic domain and the Ig-1 domain, respectively.

Since the four known FGFRs are closely related, they may also contain possible glycosaminoglycan modification sites. FGFR3 is likely to be subject to glycosaminoglycan modification, since it has a sequence similar to that of FGFR2, including the Ser-Gly-acidic box motif (14). FGFR1 is unlikely to be modified by glycosaminoglycan, since the Ser-Gly motif is located more than 15 amino acid residues to the N-terminal side of the acidic box and is encoded by the same exon as that for the Ig-1 loop (46). FGFR4 does not have any Ser-Gly motifs in the region flanking the acidic box. These features of FGFRs are conserved in mice, rats, and humans.

We have shown that glycosaminoglycan-modified FGFR2 exhibited an increased and sustained autophosphorylation and substrate phosphorylation and sustained responses to the two pathways leading to c-fos induction. In addition to the sustained activation of ERK MAP kinase, which mediates the TCF-dependent pathway for SRE activation, we found TCF-independent activation of SRE by FGFR2. The activity of this TCF-independent pathway was also increased and sustained in the cells expressing glycosaminoglycan-modified FGFR2. These findings suggest that the two pathways are simultaneously activated by FGFR2. Whereas both the ERK-mediated and the TCF-independent pathways lead to c-fos induction, they may synergize to augment the mitogenic activity of FGFR2. FGFR2 may regulate pathways controlled by Rho proteins, such as actin reorganization, since Rho family GTPases are known to regulate TCF-independent transcription (12). There appear to be some differences in the degree of phosphorylation of FGFR2 and its substrates and MAP kinase activation between the HSGAG-modified and unmodified receptors. However, it is known that phosphorylation of different sites of FGFR transmits different downstream signals (40). We previously found that the ligand-independent transforming activity of FGFR2 mutants is not well correlated either with the overall phosphorylation level of the receptor or with their effect on MAP kinase activation (20). The glycosaminoglycan modification of FGFR2 might have differential effects on its downstream pathways. Furthermore, MAP kinases, receptors, and receptor substrates should have different feedback regulation mechanisms for reversing phosphorylation. This might explain the apparent discrepancy in phosphorylation levels along the signal transduction pathway.

In accordance with the sustained activation of ERK MAP kinase and TCF-independent transcription, glycosaminoglycan-modified FGFR2 greatly potentiated induction of DNA synthesis, a long-term response by FGF stimulation. The glycosaminoglycan-modified FGFR2 responded to FGF-1 at concentrations as low as 0.1 ng/ml, while the receptors without glycosaminoglycan modification did not show any increase in DNA synthesis in this concentration range. The mutant receptor with CSGAG modification alone showed no response at low FGF-1 concentrations, although a small increase in DNA synthesis was observed at higher concentrations. Therefore, it is confirmed that HSGAG, rather than CSGAG, is important to enhance and sustain the signal transduction of FGFR2.

The presence of a low-affinity receptor, HSGAG, covalently bound to a high-affinity receptor, FGFR2, appears to increase the binding affinity of FGF to FGFR and thereby to enhance downstream signal transduction. An alternative explanation for the enhancement of FGFR2 signal transduction might be differences in receptor turnover caused by glycosaminoglycan modification of FGFR2. However, this is much less likely, though not excluded completely, since the expression level of the modified receptor was similar to those of the unmodified receptors during the course of 12 h of ligand exposure. The high-affinity FGF binding induced by HSGAG does not exclude other reported functions of the glycosaminoglycan. The glycosaminoglycan-modified receptor appears to be subject to oligomerization through the ligand bound to the glycosaminoglycan (39). Since the glycosaminoglycan is covalently attached to the high-affinity receptor, ligand accessibility to the receptor and subsequent receptor oligomerization would probably be elevated compared to the case where glycosaminoglycan is attached to other cell surface proteoglycans. Furthermore, the HSGAG on the receptor might also protect the ligand from degradation by proteases and thereby prolong the activation of FGFR, as reported for the cell surface HSGAG (6, 10, 19, 30, 33, 38).

Based on the data described here, we propose a model for an FGF dual-receptor system (Fig. 7). In cells where proteoglycans are not present, FGF-1 binds marginally to an isoform of FGFR2 that contains two Ig loops but no acidic box and transmits weak downstream signals (Fig. 7A). In cells which have proteoglycans on their surfaces, FGF-1 binds to FGFR2 with a high affinity through the help of the glycosaminoglycan moiety (Fig. 7B). The effect of proteoglycans to increase the FGF-1 binding affinity to FGFR2 can be mimicked by exogenously added heparin (Fig. 7C). A two-loop form of FGFR2 containing the acidic box can be modified by HSGAG, and the modified receptor can act as a high-affinity receptor without the assistance of cell surface proteoglycans or exogenously added heparin. This receptor can transmit potent signals, particularly at physiological or subphysiological ligand concentrations (Fig. 7D). The three-loop form of FGFR2 cannot be modified by glycosaminoglycans even though it contains the modification site, since the Ig-1 domain inhibits the modification (Fig. 7E). Therefore, it behaves the same way as the two-loop receptor without any glycosaminoglycan modification in ligand binding and in signal transduction and requires cell surface proteoglycan or exogenous heparin for potent ligand signaling.

View larger version (39K): [in this window] [in a new window]   FIG. 7.   Models of dual-receptor systems for FGF-1 binding to FGFR2. (A) The two-loop form of FGFR2, AD, binds to FGF-1 with marginal affinity and transmits weak downstream signals. (B) In cells where cell surface proteoglycans are present, FGF-1 binds to FGFR2 with a high affinity through interaction with the heparan sulfate glycosaminoglycan moiety and transmits moderate downstream signals. (C) When heparin is added to the system where no cell surface proteoglycans are available, heparin promotes high-affinity binding of FGF-1 to FGFR2. The strength of the signal transduction may be modulated by heparin concentration. (D) The two-loop form of FGFR2 containing the acidic domain can be modified by HSGAG and thus functions as a dual-receptor system in a single molecule. It binds to FGF-1 with a high affinity and transmits strong downstream signals. (E) A three-loop form cannot be modified by glycosaminoglycan despite the presence of a modification site and therefore cannot function as a high-affinity receptor unless heparin or proteoglycans are present.

The concentration of FGF-1 in vivo has not been extensively studied. The soluble concentrations of FGF-2 reported in the aqueous humor of the eye are on the order of 1 ng/ml (43). If the physiological concentration of the growth factor is typically less than 1 ng/ml, our results would predict that the FGFR2 isoform containing the acidic domain might be required for the stimulation of cells by FGF-1. These results suggest that the HSGAG-modified FGFR2 plays a critical role where responses to low concentrations of FGFs are critical. Is the sustained effect on intracellular signaling critical in cellular responses? In PC12 cells, many signal transduction events seem to be shared between differentiation and proliferation (4). EGF stimulation of PC12 cells results in transient ERK activation and leads to cell proliferation, while NGF stimulation of the same cells results in sustained ERK activation and leads to differentiation to sympathetic neurons. Marshall (22) proposed a model in which quantitative differences in ERK activation are translated into qualitative differences in transcription factor activation. FGFR is known to be expressed in various stages of embryonic development (40). At certain stages, the two-loop form containing the acidic domain, which can be modified by glycosaminoglycans, might be selectively expressed and might thus play critical roles in decisions of differentiation. We propose a novel regulation of FGFR2 glycosaminoglycan modification by alternative splicing and thereby of the signal transduction via FGFR2.