Expression and biological activity of mouse fibroblast growth factor-9.

Receptor specificity is an essential mechanism governing the activity of fibroblast growth factors (FGF). To begin to understand the developmental role of FGF-9/glial activating factor, we have cloned and sequenced the murine FGF-9 cDNA and expressed the protein in mammalian cells and in Escherichia coli. We demonstrate that the FGF-9 protein is highly conserved between mouse and human. Receptor specificity was determined by direct binding to soluble and cell surface forms of FGF receptor (FGFR) splice variants and by the mitogenic activity on cells, which express unique FGF receptor splice variants. Our data demonstrate that FGF-9 efficiently activates the “c” splice forms of FGFR2 and FGFR3, receptors expressed in potential target cells for FGF-9. Significantly, FGF-9 also binds to and activates the “b” splice form of FGFR3, thus becoming the first FGF ligand besides FGF-1 to activate this highly specific member of the FGF receptor family.

Receptor specificity is an essential mechanism governing the activity of fibroblast growth factors (FGF). To begin to understand the developmental role of FGF-9/glial activating factor, we have cloned and sequenced the murine FGF-9 cDNA and expressed the protein in mammalian cells and in Escherichia coli. We demonstrate that the FGF-9 protein is highly conserved between mouse and human. Receptor specificity was determined by direct binding to soluble and cell surface forms of FGF receptor (FGFR) splice variants and by the mitogenic activity on cells, which express unique FGF receptor splice variants. Our data demonstrate that FGF-9 efficiently activates the "c" splice forms of FGFR2 and FGFR3, receptors expressed in potential target cells for FGF-9. Significantly, FGF-9 also binds to and activates the "b" splice form of FGFR3, thus becoming the first FGF ligand besides FGF-1 to activate this highly specific member of the FGF receptor family.
Fibroblast growth factors (FGFs) 1 are a family of at least nine related polypeptides that can differentially activate a family of four related tyrosine kinase receptors (1)(2)(3). FGFs are involved in embryonic and fetal development, neovascularization, wound healing, and neoplastic transformation. Glial activating factor (GAF), a polypeptide secreted by the human glioma cell line, NMC-G1, was originally identified as a trophic factor for primary rat glial cells (4). Purified GAF is a potent mitogen for glial cells, rat primary cortical astrocytes, BALB/ c3T3 fibroblasts, and oligodendrocyte-type 2 astrocyte progenitor cells (4). GAF has weak mitogenic activity for rat adrenal pheochromocytoma cells  and no activity toward human umbilical vein endothelial cells (4). Similarities between GAF and FGFs are based on their affinity for heparin, heat stability, and chromatographic behavior. However, target cell specificity indicates significant differences in activity between GAF and FGF-2. The cloning and sequencing of the GAF cDNA revealed approximately 30% sequence homology with members of the FGF family and complete preservation of conserved amino acids that define the FGF family (1,2). Because of both the biochemical and sequence similarities, GAF was classified as the ninth member of the FGF family and will henceforth be referred to as FGF-9. Using a rat cDNA probe, FGF-9 expression was detected in rat brain and kidney but not in liver, lung, spleen, thymus, testis, heart, or adrenal gland (2). Like FGF-1 and FGF-2, FGF-9 lacks a signal peptide sequence. Nevertheless, FGF-9 is secreted by the glioma cell line, NMC-G1, and by transfected COS and Chinese hamster ovary cells (2).
FGFs differentially bind to and activate up to four related transmembrane receptors, which in turn mediate a biological response. FGF receptors (FGFRs) are members of the tyrosine kinase receptor superfamily (5). The extracellular region of the FGFR contains two or three immunoglobulin-like (Ig-like) domains that are differentially expressed as a result of alternative splicing (5). Additionally, another alternative splicing event can alter the sequence of the carboxyl-terminal half of Ig-like domain III without altering the reading frame of the remainder of the receptor. These two splice forms, referred to as "b" and "c," occur for FGFRs 1, 2, and 3 but not 4 (6 -9). The specificity of FGF ligand-receptor interaction involves the region of the FGFR ectodomain encompassing Ig-like domain II and III and is dependent on the alternative splicing event in Ig-like domain III (8,10,11). The proposed sequence of events involved in the activation of FGFRs includes the formation of a complex between FGF, a heparin-like molecule or a heparan sulfate proteoglycan and an FGFR (12)(13)(14)(15). The initial binding event is followed by receptor dimerization, autophosphorylation, and the subsequent activation of downstream signaling molecules (16 -20).
In this study, the biochemical characteristics of FGF-9 are further elucidated. We have cloned the murine FGF-9 cDNA and demonstrate that it can transform NIH3T3 fibroblasts. We demonstrate that recombinant murine FGF-9 requires heparin for optimal receptor activation and that FGF-9 preferentially binds to and activates FGFR2c and FGFR3c. Additionally, we demonstrate that unlike FGFs 2-8, FGF-9 is able to bind to and activate FGFR3b.

EXPERIMENTAL PROCEDURES
Materials-FGF-1 was a gift from K. Thomas (Merck), FGF-2 was a gift from J. Abraham (Scios Nova, Inc.), FGF-7 was a gift from Amgen Inc., and FGF-8b was purified from Escherichia coli as described (21).
Cloning Murine FGF-9-The FGF-9 cDNA was cloned from mouse brain RNA by reverse transcription-polymerase chain reaction (PCR) methods. Brain cDNA was generated from FVB mouse brain RNA using random hexamers and Moloney murine leukemia virus reverse transcriptase. The oligonucleotides used for PCR were derived from conserved sequences in the rat and human cDNA (4). The forward primer, 5Ј-CAAGCTTGGATTGAAGAAAAGAACC-3Ј, generated a HindIII site (in bold type) at the 5Ј-end of the amplified fragment. The reverse primer, 5Ј-CGAATTCAATAAGAACCCACC-3Ј, generated an EcoRI site (in bold type) at the 3Ј-end of the amplified fragment. PCR was performed for 30 cycles at 94°C for 1 min, 55°C for 2 min, and 72°C for 2 min, followed by a 5-min extension at the end of cycling at 72°C. The amplified fragment was cloned into a pGEM5 T-vector (Promega) and designated pmFGF9. The cloned PCR fragment was sequenced on an Applied Biosystems Fluorescent Sequencer and using Sequenase 2.0 (U. S. Biochemical Corp.).
Expression and Isolation of Recombinant FGF-9-The FGF-9 cDNA was cloned into the bacterial expression vector pV1 and was expressed in E. coli strain W3110 (Envirogen Inc.) according to methods previously described (22,23). Cell paste from a 5-liter fermentation was suspended in 400 ml of 35 mM Tris-HCl, pH 7.6, 1 mM phenylmethylsulfonyl fluoride, plus 1 mM EDTA and lysed by continuous sonication for 12 min at 4°C. Sodium chloride was added to the lysate to a concentration of 0.3 M, and the solution was cleared by low speed centrifugation. The supernatant was decanted and loaded onto a 5-ml heparin-Sepharose column equilibrated to 35 mM Tris-HCl, pH 7.6, plus 0.2 M NaCl. The column was washed with 20 column volumes of the same buffer plus 0.5 M NaCl. Murine FGF-9 was eluted with 5 column volumes of the same buffer plus 1.0 M NaCl. The eluted material was concentrated in an Amicon stirred cell, using a YM5K membrane, to a protein concentration of approximately 1.0 mg/ml, diluted with 35 mM Tris-HCl, pH 7.6, to achieve 0.2 M NaCl, and then reconcentrated to approximately 1.0 mg/ml protein. Final purification was achieved by fast protein liquid chromatography using a Mono-Q column (Pharmacia Biotech Inc.) and the following buffers: mobile phase, solvent A ϭ 35 mM Tris-HCl, pH 7.6, 20% acetonitrile; solvent B ϭ 35 mM Tris-HCl, pH 7.6, 1.0 mM NaCl, 20% acetonitrile; gradient 0 -100% B in 13 min. The purified murine FGF-9 eluted as a single peak at approximately 50% solvent B. The eluted protein was then repeatedly concentrated and rediluted with a solution of 35 mM Tris-HCl, pH 7.6, 0.3 M NaCl, using an Amicon stirred cell and a YM5K membrane to remove the acetonitrile and to adjust the salt concentration to 0.3 M. The purified protein was lyophilized from 35 mM Tris-HCl, pH 7.6, 0.3 M NaCl at a protein concentration of 2.0 mg/ml and then reconstituted in the appropriate cell culture media or binding media. Amino-terminal sequencing of the purified recombinant FGF-9 (Tektagen Inc.) showed that the first five amino acids were PLGEV, which corresponds to amino acids 3-7 of the predicted translation from the murine FGF-9 cDNA sequence.
Binding Assays-FGF-1 and FGF-9 were labeled by the chloramine-T method as described previously (24). Briefly, 1-2 g of FGF was incubated with 1 mCi of Na 125 I (Amersham) in the presence of 43 g/ml chloramine T (Eastman Kodak Inc.), 143 mM HEPES, pH 7.4, in a volume of 70 l for 2 min at 23°C. 100 l of 20 mM dithiothreitol was then added, and the mixture was then incubated for an additional 10 min at 23°C. The labeled growth factor was then applied to a heparinagarose column (200 l, bed volume), which had been prewashed with 20 mM HEPES, pH 7.4, 0.2% bovine serum albumin, and 0.4 M NaCl. Labeled growth factor was eluted with 20 mM HEPES, pH 7.4, 0.2% bovine serum albumin, and 3 M NaCl and stored frozen at Ϫ70°C for up to 14 days.
Soluble FGFR (1c, 2b, 3b, and 3c)-alkaline phosphatase fusion proteins were made in COS cells as described previously (8,18). Binding components were added at 4°C in the following order: Dulbecco's modified Eagle's medium (Life Technologies, Inc.) with 0.1% bovine serum albumin, 50 l of a 2ϫ slurry of anti-alkaline phosphatase monoclonal antibodies coupled to Sepharose (25), 10 l of 25 g/ml heparin, 50 l of FGFR-alkaline phosphatase-conditioned media containing specific soluble FGFRs (0.3 optical density units/min) (18), non-iodinated FGF as a competitor, and 125 I-FGF (20,000 cpm) in a total volume of 250 l. The reaction was then gently rotated for 90 min at 4°C. Bound receptor and FGF were recovered by centrifugation (10 s at 12,000 rpm (4000 ϫ g), 4°C in a microcentrifuge), and washed two times with 500 l of ice-cold phosphate-buffered saline. 125 I-FGF binding was determined by counting the washed tubes directly in a ␥ counter (Beckman). Binding to cell surface FGF receptors was performed as for the soluble FGF receptors except that anti-alkaline phosphatase-Sepharose and FGFR-alkaline phosphatase were replaced with 3 ϫ 10 5 FGFR-expressing BaF3 cells. 20,000 -60,000 cpm of iodinated FGF were incubated with cells in the presence of 2 g/ml heparin for 2 h at 4°C. Cells were washed and counted as described above.
Mitogenic Assays-Full-length cDNAs encoding the three immunoglobulin-like domain form of FGFRs 1b, 1c, 2b, and 2c were cloned into expression vector MIRB. MIRB contains the Moloney murine leukemia virus long-terminal repeat, unique EcoRI, BamHI, and SpeI sites, followed by the IRES-NEO gene (previously described in Ref. 8) in the Bluescript KS plasmid.
The plasmid pSVFGFR1b was provided by Werner et al. (7). A 2.9-kb BamHI-SpeI fragment was cloned into the corresponding sites of MIRB. FGFR1c (14) was cloned as a 3.2-kb EcoRI fragment into MIRB by converting a 3Ј-Asp-718 site into an EcoRI site and then excising with EcoRI. FGFR2b (26) was cloned as a 2.9-kb BamHI fragment into MIRB by converting a 5Ј-Asp-718 site into a BamHI site and then excising with BamHI. FGFR2c (27) was cloned into MIRB as a 3.6-kb SpeI fragment by converting unique NarI and XbaI sites into SpeI sites.
The FGFR3 cDNAs were reengineered to enhance signaling in BaF3 cells by constructing chimeric cDNAs encoding the extracellular region of FGFR3 fused to the cDNA encoding the tyrosine kinase domain of FGFR1. FGFR31c cells express an FGFR that has the extracellular region from FGFR3c (28) and transmembrane domain and tyrosine kinase domain from FGFR1; FGFR31b cells express an FGFR that has the extracellular region and transmembrane domain derived from FGFR3b (8) and tyrosine kinase domain derived from FGFR1. The details of these chimeric receptors will be described elsewhere.
The mouse FGF-9 cDNA was excised with HindIII and EcoRI, blunted with the Klenow fragment of DNA polymerase, ligated to BamHI linkers, re-excised with BamHI, and then cloned into the BamHI site of the MIRB expression vector. MIRB-FGF-9 was then transfected into mammalian cells.
BaF3 cells expressing specific FGFRs were washed and resuspended in Dulbecco's modified Eagle's medium, 10% neonatal bovine serum, L-glutamine. 22,500 cells were plated per well in a 96-well assay plate in media containing 2 g/ml heparin, except where indicated. Test reagents were added to each well for a total volume of 200 l per well. The cells were then incubated at 37°C for 2 days. To each well, 1 Ci of [ 3 H]thymidine was then added in a volume of 50 l. Cells were harvested after 4 -5 h by filtration through glass fiber paper. Incorporated [ 3 H]thymidine was counted on a Wallac ␤ plate scintillation counter.

RESULTS
Cloning and Expression of Mouse FGF-9-Mouse FGF-9 was cloned by reverse transcription-PCR from mouse brain RNA (see "Experimental Procedures"). Sequencing of the amplified 851-base pair cDNA identified the entire FGF-9 open reading frame (209 amino acids) located between bases 146 and 773 in Fig. 1. When compared to the sequence of human FGF-9, murine FGF-9 exhibited 92% identity in its nucleotide sequence and 99% identity in its amino acid sequence. The predicted molecular mass of the FGF-9 protein is 23,532 Da. The FGF-9 cDNA was expressed in E. coli and purified to near homogeneity by chromatography on heparin-Sepharose and fast protein liquid chromatography using a mono-Q column (see "Experimental Procedures"). The final purified FGF-9 protein was analyzed on an 18% SDS-polyacrylamide gel and stained with Coomassie Blue. The purified protein was resolved as a strong band running at approximately 24 kDa (Fig. 2), which is consistent with its predicted molecular mass.
Receptor Binding Specificity of FGF-9-To determine the binding specificity of recombinant FGF-9, we tested the ability of 125 I-labeled FGF-9 to bind to FGFR1c, -2b, -3b, and -3c expressed on BaF3 cells. Significant binding was observed on cells expressing FGFR3c. No significant binding was detected on cells expressing FGFR1c, -2b, or -3b (data not shown). Because the sensitivity of this assay is poor, we further characterized the binding properties of FGF-9 using the extracellular domains of soluble FGFRs (1c, 2b, 3b, and 3c) fused to alkaline phosphatase (8) (Fig. 3A). Labeled FGF-9 binds best to FGFR3c (17.1-fold over background) and does not bind to FGFR1c. A small amount of binding was observed with FGFR2b (2.7-fold over background) and FGFR3b (3.5-fold over background). In all cases, FGF-9 binding could be competed with non-radioactive FGF-9.
The observed binding of FGF-9 to soluble FGFR3b was surprising in that no other FGF ligand, besides FGF-1, can bind to or activate this receptor (8). 2 Therefore, to further characterize this binding interaction, we tested the ability of FGF-1, FGF-2, and FGF-9 to compete with the binding of labeled FGF-9 to FGFR3b. Both FGF-1 and FGF-9 can efficiently compete for binding to FGFR3b; however, FGF-2 was only a weak competitor (Fig. 3B). This is consistent with the inability of FGF-2 to bind to or activate FGFR3b (8).
FGF-1 binds with high affinity to all known FGFRs (7-10, 26, 28 -30). Competition binding experiments demonstrate that FGF-9 cannot compete with labeled FGF-1 for binding to FGFR1c yet does demonstrate an increasing ability to compete with FGF-1 for binding to FGFR2b, FGFR3b, and FGFR3c (Fig.  3C), respectively. These data are consistent with the direct binding experiments in which FGF-9 preferentially binds to FGFR3c and less well to FGFR2b and FGFR3b. The ability of FGF-9 to compete with labeled FGF-1 for receptor binding is poor compared to that of unlabeled FGF-1. However, this observation is consistent with previous results in which FGF-7 could only partially compete with FGF-1 for binding to its primary receptor, FGFR2b (8).
Receptor Activation by Recombinant FGF-9-To further test the biological activity of FGF-9, we assayed its ability to activate the "b" and "c" splice forms of FGFRs 1-3. FGFRs 1b, 1c, 2b, and 2c were expressed in BaF3 cells as full-length receptors. FGFRs 3b and 3c were expressed in BaF3 cells as fusion proteins, which contain the entire extracellular domain of FGFR3 genetically fused to the intracellular domain of FGFR1. Each of these cell lines responds well to FGF-1 (Fig. 4), consistent with previous binding and activity data, which indicate that FGF-1 can interact with and activate all FGFRs. Two additional control ligands were also used, FGF-7 and FGF-8. FGF-7 can only activate FGFR2b (Fig. 4C), while FGF-8 specifically activates FGFR2c and FGFR3c (Fig. 4, D and F) (21). Similar to FGF-8, recombinant FGF-9 most efficiently activates FGFR2c and FGFR3c; however, unlike with FGF-8, FGF-9 can also partially activate FGFR1c and FGFR3b (Fig. 4). FGF-9 shows no activity toward cells expressing FGFR1b and FGFR2b. The strong activity toward FGFR3c is consistent with the cell surface and soluble receptor binding data. The weak activity toward FGFR3b is consistent with binding to soluble receptor. However, we do not detect binding to cell surface or soluble FGFR1c even though cells expressing this receptor can 2 D. M. Ornitz, unpublished observations. FIG. 1. cDNA sequence of FGF-9. The nucleotide and amino acid sequence of human and murine FGF-9 (hFGF-9 and mFGF-9, respectively) is shown. Oligonucleotide primers used to amplify FGF-9 are underlined. Coding sequence begins at human base pair 151 and ends at base pair 774 and is shown above the nucleotide sequence. Differences between the human and murine nucleotide sequence are depicted in bold type in the murine sequence. The two amino acid differences are depicted in bold type with the human amino acid listed before the murine amino acid. be weakly activated by FGF-9. Conversely, we do detect weak binding to soluble (but not cell surface) FGFR2b even though cells expressing this receptor are not activated by FGF-9. The physiological significance of these activities toward FGFR1c and FGFR2b remain to be defined.
To assess the relative mitogenic activity on individual FGFR splice variants, we normalized the mitogenic data in Fig. 4 to that of FGF-1. The relative mitogenic activity for each ligand ([ 3 H]thymidine incorporation) at concentrations of 312 and 1250 nM was calculated. These values were then averaged and plotted in Fig. 5. This analysis clearly demonstrates that the best receptors for FGF-9 are FGFR2c and FGFR3c. These FGFRs can be activated by FGF-9 with 89 and 96% of the activity of FGF-1, respectively. FGFR3b-expressing cells respond to FGF-9 with 42% of the activity of FGF-1, and FGFR1cexpressing cells respond to FGF-9 with 21% of the activity of FGF-1.
Receptor Activation Specificity of Native FGF-9-To assess the activity of native FGF-9 and to compare it to that of recombinant FGF-9, we transfected NIH3T3 cells with an FGF-9 cDNA expression vector (see "Experimental Procedures"). Transfected cells were either allowed to grow to confluency in a focus-forming assay or selected for expression of the trans-fected plasmid with the drug G418. Both selection methods result in morphological transformation of NIH3T3 cells. Foci or selected colonies grew in a disorganized manner, with the majority of cells assuming a spindle-shaped morphology (data not shown). Media conditioned by these cells was then assayed for mitogenic activity on BaF3 cells expressing FGFR1c, FGFR3b, or FGFR3c (Fig. 6). Similar to recombinant FGF-9, FGF-9-expressing NIH3T3 cell-conditioned media activate FGFR3c Ͼ FGFR3b Ͼ FGFR1c. Control NIH3T3 cell-conditioned media do not activate cells expressing these FGFRs. We conclude that the biological activity of both the native and recombinant FGF-9 is similar.
FGF-9 Requires Heparin for Optimal Mitogenic Activity-All known members of the FGF family bind heparin with relatively high affinity (K d ϳ10 Ϫ9 M) (1), and several of the FGFs have been shown to require heparin for optimal biological activity (14,18,28). In biological assays, heparin is likely to serve at least two functions: 1) heparin can increase the stability and thus the half-life of FGF molecules and 2) heparin can stabilize and thus increase the half-life of ligand-receptor interaction. Like other members of the FGF family, FGF-9 is also a heparin binding protein.
To determine the effect of heparin on the biological activity of FGF-9, we compared the heparin dependence of FGF-9 mitogenic activity to that of FGF-1 on BaF3 cells expressing FGFR3c (Fig. 7). Both FGF-1 and FGF-9 require heparin for maximal biological activity on FGFR3c-expressing BaF3 cells. Half-maximal activation of FGF-1 is seen at 670 ng/ml heparin, and half-maximal activation of FGF-9 is seen at 185 ng/ml heparin. Compared to FGF-1, FGF-2 (19) and FGF-9 require lower heparin concentrations for optimal activity. This difference in heparin dependence may reflect a higher affinity for heparin by these two FGFs. DISCUSSION FGFs compose a family of growth factors that play key roles in a variety of developmental events. Some FGFs are expressed in adult tissues and may be important for maintaining normal tissue homeostasis. FGFs are also involved in mediating a physiological response to injury (31). In adult mice, FGF-9 is expressed in both brain and kidney (2). During development, FGF-9 is expressed at low levels in mid-gestation mouse embryos. 3 Thus, it is likely that FGF-9 plays a role in both developmental events and in normal adult physiology. Important elements controlling the activity of FGFs include tissue-and temporal-specific gene expression and specificity of ligand-receptor interactions. In this study, we have determined the receptor specificity of FGF-9, thus identifying its potential physiologically relevant receptors.
Our data indicate that the preferred receptors for FGF-9 are FGFR2c, FGFR3c, and FGFR4. 4 Additionally, the binding and mitogenic data presented here demonstrate that FGF-9 can also bind to and activate FGFR3b. Although the activity of FGF-9 toward FGFR3b is only 42% of that of FGF-1, it is nevertheless significant because no other FGF ligand shows any activity toward this receptor (8). 4 FGFR2 is expressed in glial cells (32), low grade astrocytomas (33), and oligodendrocytes present in fiber tracts of the central nervous system (34). FGFR2 transcripts have been identified in the germinal epithelium of the developing central nervous system and, later in development, in a diffuse pattern consistent with expression in glial cells (35). FGFR2 is also prominently expressed in epithelial tissues in limb bud, kidney, stomach, and lung (35,36). FGFR3 mRNA has been localized to the germinal epithelium of the developing central nervous system, to glial cells (later in development), to the sensory epithelium of the cochlea, to proliferating cartilage of developing bone, and to the lens of the eye (37). Although the cellular localization of FGFR3b and FGFR3c is not known, RNase protection studies indicate that both splice forms are expressed in kidney (8). In the case of FGFR2, "b" splice forms are restricted to epithelial tissues, and "c" splice forms are expressed in mesenchymal tissues (10,38).
The identification of FGFR2 and FGFR3 expression in glial cells and the identification of a functional ligand, FGF-9, expressed in the mouse and rat brain suggests that FGF-9, FGFR2, and FGFR3 may participate in either autocrine or paracrine loops in the central nervous system. The discovery of FGF-9 secretion by a glioma cell line (2,4) also supports the functional pairing of FGF-9 with these receptors and may ac- count for the formation of the tumor that originally gave rise to this cell line. Additional support for FGF-9 forming physiologically relevant ligand-receptor pairs with FGFR2 and FGFR3 comes from gene expression in kidney where FGF-9 (2), FGFR2c, FGFR3, and FGFR4 are all expressed (30,37,39). FGF-9 may therefore play a role in both neural and renal development, tissue homeostasis, and response to injury. Spatial localization of FGF-9, FGFR2, and FGFR3 at the cellular level should help to define the physiological relationship between these signaling molecules.