An Alternatively Spliced Fibroblast Growth Factor (FGF)-5 mRNA Is Abundant in Brain and Translates into a Partial Agonist/Antagonist for FGF-5 Neurotrophic Activity*

We detected in the brain and then cloned two novel, short forms of human and mouse fibroblast growth factor (FGF)-5 mRNA, which were designated human FGF-5S (hFGF-5S) and mouse FGF-5S (mFGF-5S), respectively. Genomic analysis indicated that mFGF-5S and authentic mFGF-5 mRNAs were transcribed from a single gene; hFGF-5S and mFGF-5S mRNAs were generated by excluding the second exon of the respective FGF-5 genes, and the alternatively spliced mRNAs encoded for 123- and 121-amino acid proteins, respectively. Indeed, a neuron-like cell line expressing mFGF-5S mRNA secreted a protein of the expected size and with FGF-5 antigenicity. In PC12 cells, expression of hFGF-5 or exposure to hFGF-5 protein induced differentiation. Neither expression of hFGF-5S, alone, nor co-expression of hFGF-5S with hFGF-5 induced significant differentiation. At high concentrations, hFGF-5S protein partially antagonized FGF-5 activity, whereas by itself, hFGF-5S exerted very weak neurotrophic activity. hFGF-5S protein binds to FGF receptor (FGFR)-1 on PC12 transfectants and partially inhibits hFGF-5-induced tyrosine phosphorylation of FGFR-1 and an FGFR substrate, but it also induces phosphorylation by itself. These results suggest that FGF-5S is a naturally expressed partial agonist/antagonist of FGF-5 neurotrophic activity in the brain and that its effects are exerted in part at the level of the receptor.

RNA Preparation-Total RNAs from mouse tissues and cultured cells were prepared using Isogen solution (Nippon Gene, Tokyo) according to the manufacturer's protocol (27). Human cerebellar total RNA was purchased from CLONTECH (Palo Alto, CA).
DNA Cloning and Sequence Analysis-PCR products were separated by agarose gel electrophoresis and recovered with DEAE-cellulose membranes (NA-45, Schleicher and Schuell, Dassel, Germany). The fragments were blunt-ended with Escherichia coli DNA polymerase I (Takara Syuzo, Kyoto, Japan), cloned into an EcoRV-digested pBluescript SKϩ vector (Stratagene, La Jolla, CA), and introduced into E. coli DH5. The recombinant plasmids were propagated and sequenced using the DyeDeoxy terminator cycle sequencing kit and model 373A DNA sequencer (Applied Biosystems). All processes were carried out according to the standard procedures (29).
Genomic Southern Hybridization Analysis-Ten micrograms of genomic DNA isolated from ICR mouse liver were digested with restriction enzymes, BamHI, EcoRI, or HindIII. The DNAs were electrophoresed through 1% agarose gel, transferred to nylon membranes (Hybond Nϩ, Amersham) in a 10ϫ SSC capillary flow and UV-cross-linked. A set of PCR primers (5Ј-GGAATTCCATATGAGCCTGTCCTTGCTCTTCCT-C-3Ј and 5Ј-CCCGGATCCCTTTATCCGTAAATTTGGCTTAACACAC-3Ј) and a template (p5.1) were used for amplification of the entire open reading frame of FGF-5S cDNA. The amplified fragment was labeled with digoxygenin-dUTP using a digoxygenin DNA labeling kit (Boehringer Mannheim, Germany). Hybridization and CSPD chemiluminescence detection were performed essentially as described previously (30).
Northern Hybridization Analysis-Ten micrograms of total cellular RNA were used for detection of FGFR-1 mRNA (30). A FGFR-1 probe was synthesized by PCR in the presence of digoxygenin-dUTP (Boehringer Mannheim) using the FGFR-1 plasmid and a primer pair encompassing its extracellular domain (21,28).
RNase Protection Analysis-RNase protection analysis was carried out as described previously (31) with some modifications. Templates for FGF-5 sense and antisense riboprobes (p5.1) were digested with EcoRI and HindIII, respectively. They were then transcribed in vitro using T7 or T3 RNA polymerase (Life Technologies, Inc.) in the presence of digoxygenin-dUTP; DNase-treated; purified on a Sephadex G-50 column; and stored in 1 ml of 50% ethanol solution for later use. Ten micrograms of total RNA were isolated from cells and ethanol precipitated with 1 l of either sense or antisense riboprobes plus 40 g of E. coli tRNA. The RNA pellet was resuspended in 30 l of hybridization buffer consisting of 40 mM PIPES (pH 6.4), 1 mM EDTA (pH 8.0), 0.4 M NaCl, and 80% formamide. After heat denaturation at 85°C for 10 min, hybridization was carried out at 45°C for 16 h. Unannealed RNAs were digested by adding an RNase digestion mixture (300 mM NaCl, 10 mM Tris (pH 7.4), 5 mM EDTA and 2 g/ml RNase T1 (Sigma)). RNase A treatment was omitted, because it degrades the digoxygenin-RNA:RNA hybrid completely (32,33). After incubating at 30°C for 2 h, RNA was precipitated with ethanol. Single stranded RNAs were separated in a 6% polyacrylamide, 7 M urea, TBE gel followed by semidry electroblotting onto a nylon membrane (Hybond Nϩ) in 0.2ϫ TBE at 0.5 mA/cm 2 for 45 min. The RNAs were fixed by UV cross-linking and subjected to chemiluminescence detection using CSPD (30).
Production of Recombinant FGF-5 and FGF-5S Proteins in E. coli-Full-length, recombinant FGF-5 protein was obtained using a bacterial expression vector system, pET-3c (35). The FGF-5 cDNA fragment was amplified with the primer set, 5Ј-CGGAATTCCATATGGGTGAAAAG-CGTCTCGCCCCCAAA-3Ј (sense) and 5Ј-CGCCATATGTTTATCCAAA-GCGAAACTT-3Ј (antisense), using a pLTR122 template (1); an Nterminal hydrophobic signal sequence for the original FGF-5 protein was not included in this construct (36). The amplified fragment was cloned into pBluescript SKϩ, sequenced as described above, digested with the restriction enzyme NdeI (underlined), and inserted into the NdeI site of the pET-3c vector. The recombinant plasmid was maintained in E. coli (strain BL21(DE3)pLysS) (35), and the desired protein was expressed and extracted as described (36).
Human FGF-5S polypeptide was produced from an FGF-5S/malE fusion protein (37). The open reading frame sequence of the human FGF-5S cDNA was amplified by PCR using the primer set, 5Ј-TCA-GAATTCGGGGAGAAGCGTCTCGCC-3Ј (sense) and 5Ј-CGCGGATCC-CTTTATCTGTGAACTTGGCTTAACATATTGGCTTCGTGcGATCCA-T-3Ј (antisense). EcoRI and BamHI recognition sites were attached to the sense and antisense primers, respectively (underlined), and a silent mutation (G to "c") was introduced to eliminate a BamHI site within the open reading frame. After amplification with this primer set and with pLTR122 DNA as a template, the amplified fragment was cloned into pBluescript SKϩ and sequenced. The inserts were then cut out using EcoRI and BamHI and introduced into a pMAL-c2 plasmid vector (New England Biolabs, Beverly, MA) digested with the same two restriction enzymes. The chimeric plasmid, designated pMAL/humFGF5S, codes for a fusion protein, maltose-binding protein, MalE, plus human FGF-5S. As described above, an N-terminal hydrophobic signal sequence was not included. E. coli (strain BL21) (38) harboring the recombinant pMAL/humFGF5S plasmid was grown in L broth. The recombinant fusion protein, designated MBP/hFGF-5S, was then induced, purified on an amylose resin (New England Biolabs) affinity column, and cleaved using protease factor Xa (New England Biolabs) as described previously (39). The resultant FGF-5S polypeptide was further purified by high performance liquid chromatography (HPLC; Gilson, Middleton, WI) using a Mono-S column (Pharmacia Biotech, Uppsala, Sweden), eluted with a 0.25 M (pH 6.0) to 1 M (pH 7.5) sodium phosphate linear gradient, and stored at Ϫ80°C. The amino acid sequence of the peptide was confirmed using a protein sequencer (model 476A, Applied Biosystems).
Transfection of Expression Plasmids into PC12 Cells-Expression plasmids were transfected into PC12 cells using a lipofection procedure. Two micrograms of the plasmid were mixed with Lipofectamine reagent (Life Technologies); the mixture was incubated for 15 min at room temperature and then added to the cells. After 24 h, the mixture was removed and replaced with DMEM and the appropriate concentration of serum. For selection, 500 g/ml G418 (Geneticin, Sigma) was added to the culture, and selection was continued for 2 weeks.
Measurement of Acetylcholinesterase Activity-Acetylcholinesterase activity was measured using the colorimetric method (40,41). PC12 cells were plated at a density of 5 ϫ 10 6 cells/9-cm dish and cultured first in the DMEM plus 10% fetal bovine serum and 5% horse serum for 12 h and then in low serum medium (1% fetal bovine serum and 0.5% horse serum) for 5 days. Cell pellets were homogenized in 400 l of ice-cold buffer containing 10 mM Tris-HCl (pH 7.2), 1 M NaCl, 50 mM MgCl 2 and 1% Triton X-100. The crude extract was added to a cuvette containing 2.6 ml of 0.1 M sodium phosphate buffer (pH 8.0), 0.1 ml of 10 mM dithiobisnitrobenzoic acid (dissolved in 0.1 M sodium phosphate buffer (pH 7.0) containing 15 mg of sodium bicarbonate/10 ml), and 20 l of 75 mM acetylthiocholine iodide. Changes in absorbance were recorded at 412 nm over a 10-min period using a UV spectrophotometer. Protein concentrations were measured using a protein assay (Bio-Rad) according to the manufacturer's protocol. Enzyme activity was expressed as nmol of acetylthiocholine hydrolyzed per min per mg of protein.
PC12 Cell Differentiation Assay-PC12 cells were seeded in 12-well microtiter plates at approximately 1 ϫ 10 4 cells/well in DMEM supplemented with 10% fetal bovine serum and 5% horse serum. The cells were allowed to attach to the bottom of the wells for about 5 h, and then the medium was replaced with DMEM supplemented with 0.1% bovine serum albumin (fraction V, Sigma). Various combinations of growth factors were then added to each well, and in some experiments, 5 g/ml heparin was also added. After 24 h, the cells were examined using phase-contrast microscopy under low magnification. Randomly selected fields containing approximately 100 cells were photographed, and the numbers of undifferentiated and differentiated cells were counted. Morphological features used as experimental criteria for distinguishing differentiated from undifferentiated cells were cell body enlargement and neurite growth; cells having neurites greater in length than the average undifferentiated cell body diameter were considered differentiated, as were cells with cell body diameters at least 2 times larger than the average undifferentiated cell diameter.
Heparin Affinity Chromatography-The heparin binding properties of bacterially produced hFGF-5 and hFGF-5S were measured using an immobilized heparin column. Appropriate amounts of hFGF-5 and/or hFGF-5S were diluted with 10 mM Tris (pH 7.5) and applied to a HiTrap heparin column (bead volume 1 ml; Pharmacia) attached to an HPLC apparatus (Gilson). Unbound materials were washed with low ionic strength buffer (10 mM Tris, pH 7.5, 50 mM NaCl) until absorbance levels dropped to base line. The column was then eluted with a linear gradient of 0.05-2 M NaCl in 10 mM Tris (pH 7.5) buffer. Peaks were monitored by UV absorbance (280 nm); after the fractions were collected, they were confirmed by 15% SDS-PAGE and Western blotting analysis using anti-FGF-5 serum.
Competitive Receptor Binding-Recombinant hFGF-5 proteins were iodinated using IODO-BEAD iodination reagent (Pierce) (42), and the labeled proteins were purified on an immobilized heparin column (heparin-Sepharose CL-6B; Pharmacia); 8.1 ϫ 10 4 cpm/ng of purified hFGF-5 protein was obtained. PC12 cells were plated to semiconfluence on 24-well microtiter plates and washed twice with binding buffer (PBS supplemented with 0.1% bovine serum albumin and 5 g/ml heparin); and then 0.25 ml of the same buffer was added to each well and cooled on ice. Labeled FGF-5 (0.16 nM; 4 ng/0.25 ml) as well as various concentrations of nonlabeled recombinant hFGF-5 or hFGF-5S were then applied to the cultures. Receptor binding was performed at 4°C for 3 h. The cells were then washed one time with binding buffer and two times with 2 M NaCl, 10 mM Tris (pH 7.5) and lysed with 0.5 N NaOH. The retained membrane-bound hFGF-5 ligand was counted with a ␥-counter.
Detection of FGF-5 and FGF-5S Protein in Vivo-Anti-FGF-5 antibodies reactive to epitopes common to FGF-5/FGF-5S were purified from rabbit antiserum using MBP/FGF-5S protein coupled to Sepharose 4B (CNBr-activated, Pharmacia) as described by Huet (43). 102b67 cells were grown to confluence in 15-cm dishes. The medium was then changed to 25 ml of serum-free medium ASF104 (Ajinomoto, Tokyo, Japan), and 25 l of heparin-Sepharose beads (Pharmacia) were added to the cultures to adsorb secreted FGF-5 proteins. After 3 days of culture, the heparin-Sepharose beads were recovered by centrifugation and washed with 10 mM Tris (pH 7.4), 50 mM NaCl. The proteins adsorbed on the beads were boiled in sample buffer, resolved by electrophoresis on 15% SDS-polyacrylamide gels, and transferred to nitrocellulose membranes. The membranes were blotted with the affinitypurified antibody, and the signals were detected using horseradish peroxidase-conjugated anti-rabbit IgG and ECL.
Detection of Phosphorylated FRS2-PC12/FGFR-1 cells were serumstarved overnight in DMEM supplemented with 0.1% bovine serum albumin and 0.5 g/ml heparin. Growth factors were added to the cells for a period of 10 min, and the cells were then lysed (lysis buffer: 6 M urea, 10 mM sodium phosphate, pH 7.2, 1% SDS, and 10 mM 2-mercaptoethanol). The lysates were separated by 8% SDS-PAGE; Western transferred to polyvinylidene difluoride membranes; and then detected with anti-phosphotyrosine antibody, horseradish peroxidase-conjugated anti-mouse IgG, and ECL.
Detection of Receptor Tyrosine Phosphorylation of FGF Receptor-Cells stimulated with various samples for 5 min were lysed (lysis buffer: 10 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 10 mM Na 4 P 2 O 7 , 10 mM NaF, 2 mM Va 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, and 1% Nonidet P-40) and centrifuged. Anti-FGFR-1 antibody was added to the cleared supernatant. The immune complex was then precipitated by adding protein G-Sepharose (Pharmacia), separated by electrophoresis on 8% SDS-polyacrylamide gels, and transferred to polyvinylidene difluoride membranes. The membrane was probed with an anti-phosphotyrosine monoclonal antibody, and the signals were detected using horseradish peroxidase-conjugated anti-mouse IgG and ECL.
Cross-linking MBP/hFGF-5S Fusion Proteins to FGF Receptors-PC12/FGFR-1 cells were grown to confluence in 6-cm dishes, washed with PBS, and incubated with MBP/hFGF-5S fusion protein (0.2 or 2 M) in 1 ml of binding buffer (PBS containing 0.1% bovine serum albumin and 5 g/ml heparin) at 4°C for 4 h with gentle shaking. The same concentration of MBP/LacZ fusion protein was used as a negative control. The cells were washed with cold PBS, cross-linked with disuccinimidyl suberate (0.2 mM in PBS) at 4°C for 15 min, and then washed again with PBS. The cells were then lysed in 1 ml of lysis buffer (10 mM Tris, pH 7.5, 150 mM NaCl, and 1% Nonidet P-40) and cleared by centrifugation, and FGFR-1 was precipitated by the sequential addition of anti-FGFR-1 antibody and protein G-Sepharose (Pharmacia). The precipitated proteins were separated by 6% SDS-PAGE and transferred to a polyvinylidene difluoride membrane, and the signals were detected using anti-MBP serum (New England Biolabs), AP-conjugated antirabbit IgG, and CDP-Star.

RESULTS
PCR Amplification and Cloning of mFGF-5 and mFGF-5S cDNAs-A pair of PCR primers were designed to amplify most of the protein coding region of mFGF-5 mRNA; the sense and antisense primers annealed with the first and third exons, respectively (Fig. 1a). When RNAs from the brains of embryonic, neonatal, and adult mice were reverse-transcribed and amplified with these primers, two distinct DNA bands of slightly different size were reproducibly observed by gel electrophoresis (approximately 650 and 550 bp; Fig. 1a).
To determine the nucleotide sequences of the isolated DNAs, DNA fragments were recovered from the agarose gels and cloned into a plasmid vector. Sequencing of the clone designated p5.1 revealed that the larger PCR product was 652 bp long and was identical to previously reported mouse FGF-5 mRNA (13). The smaller PCR product (p5short.1) was 548 bp in size, and the nucleotide sequence was identical to the 652-bp product except for a 104-bp deletion in the middle. Three independently obtained clones containing the smaller fragment yielded the same results and confirmed that the sequence was not a cloning artifact. Comparison with FGF-5 genomic organization (13) showed that the 548-bp cDNA was amplified from an alternatively spliced mRNA that lacks the second exon of the FGF-5 gene (Fig. 1, a and b). In this mRNA, the first and third exons were directly joined, which generated an immediate termination codon and a new frame in the third exon. Consequently, this alternative splicing introduces four new amino acids, Gln-Ile-Tyr-Gly, as a new carboxyl terminus following the Ser 117 of the authentic 264-amino acid mFGF-5 protein. This novel form of mFGF-5 mRNA translates into a 121-amino acid polypeptide, which we designated mFGF-5S.
RNase Protection Analysis-We further confirmed the existence of the mFGF-5S mRNA in vivo using RNase protection analysis (Fig. 1c). In this experiment, neuron-like 102b67 cells were used, because, as determined by RT-PCR, they express more FGF-5 and FGF-5S mRNAs than the mouse brain (data not shown). No protected band was seen with the sense riboprobe. On the other hand, the antisense riboprobe generated three protected bands after RNase T1 digestion. The largest protected fragment (652 bases) corresponded in size to the full-length mRNA for FGF-5. The other observed bands (222 and 326 bases) corresponded to the probe hybridized with mFGF-5S mRNA at exons I and III. A schematic drawing depicting generation of these three bands is shown in Fig. 1c.
The above results indicate that FGF-5S mRNA, an alternatively spliced form of FGF-5 mRNA, is indeed expressed in vivo and is not an artifact of RT-PCR. They also show that the FGF-5 gene is transcribed in a single direction, in contrast to the FGF-2 gene, which is bidirectionally transcribed in mammals (44) and in Xenopus (45).
Detection of FGF-5S Protein Expression in Vivo-To demonstrate the presence of mFGF-5S mRNA translation products, in vivo, we carried out a Western blotting experiment. hFGF-5 contains a secretion signal peptide at its N terminus, and its secretion from NIH 3T3 transfectants overexpressing FGF-5 has been shown previously (1,25). mFGF-5 peptide sequences and the predicted mFGF-5S proteins share the same N terminus. Therefore, we tested for the presence of secreted mFGF-5S protein in the culture medium of 102b67 cells; because hFGF-5S binds to heparin-Sepharose beads (described below), beads were included in the culture medium and then later examined for FGF-5 antigenicity. To avoid any cross-reactivity derived from serum components, serum-free medium was used.
In addition, to eliminate other nonspecific signals, anti-FGF-5 antibody was affinity-purified from antiserum using an epitope common to both hFGF-5S and hFGF-5. These efforts enormously improved the signal-to-noise ratio of the Western blot analysis.
Three immunoreactive proteins with three different mobilities were reproducibly detected in the conditioned medium of 102b67 cells, which naturally express both forms of FGF-5 mRNA (Fig. 1d, lane 1; 40, 38, and 18.5 kDa). The molecular masses of mFGF-5 and mFGF-5S after secretion were 27 and 11 kDa, respectively, as calculated from their amino acid compositions. An earlier study showed that, based on the calculated molecular weight, the actual migration of simple FGF-5 protein on SDS-PAGE is at variance with the expected migration, perhaps due to its unusual amino acid composition (36). Indeed, bacterially expressed hFGF-5 and hFGF-5S (see below), which lack the signal sequences, exhibited slower mobilities (31 and 15 kDa, respectively) than would be deduced from their molecular masses (27 and 11 kDa, respectively; Fig. 1d,  lanes 2 and 3). Moreover, hFGF-5 protein runs more slowly on  (30), but the final stringent wash was with 0.1 ϫ SSC, 0.1% SDS at 65°C (right). f, distribution of FGF-5 and FGF-5S mRNAs in mouse tissues. Tissues were obtained from BALB/c mice and dissected, and total RNAs were extracted. The mRNAs transcribed from the FGF-5 gene were amplified using the primers 5Ј-CCTTCGGGGCGC-CGGACCGGCA-3Ј and 5Ј-CTTGGCTTTCCCTCTCTTGTTC-3Ј. Amplified cDNA fragments were separated on an agarose gel, transferred to a nylon membrane, and hybridized with an FGF-5 cDNA probe; specific bands were visualized by chemiluminescence detection. Tissues used were brain (lane 1), skeletal muscle (lane 2), heart (lane 3), lung (lane 4), gut (lane 5), and liver (lane 6). Lane 7 is a negative control without a template. As a control, primers and a probe for GAPDH were used (left). Signal intensities on the film (optical density) were scanned and presented as a graph (right). SDS-PAGE than its core protein due to glycosylation (25). The predicted N-glycosylation site on FGF-5 is retained in FGF-5S; thus, we interpret our findings as indicating that the 18.5-kDa protein is a translation product of FGF-5S mRNA with glycosyl modification(s). Because immunoblot signals were unaffected by the inclusion of protease inhibitors in the culture (data not shown), we think it unlikely that the 18.5-kDa band represents a proteolytic degradation product of full-length FGF-5. The 40and 38-kDa proteins are likely to contain the hFGF-5 core protein with additional heterogenous modifications.
Genomic Southern Hybridization Analysis-In order to determine whether the mFGF-5S mRNA was generated from the known mFGF-5 gene, we performed a genomic Southern hybridization experiment. Mouse liver genomic DNA was digested with one of three restriction enzymes (BamHI, EcoRI, or HindIII), elecrophoresed, and transferred to membranes. As a probe, we used a cDNA fragment encompassing the entire first exon of the mouse FGF-5 gene that was contained in both mFGF-5 and mFGF-5S mRNAs. In each case, the cDNA probe hybridized to a single restriction fragment of mouse genomic DNA, clearly demonstrating that the two forms of mRNA were generated from a single gene on the genome (Fig. 1e).
Distribution of FGF-5 and FGF-5S mRNA in the Tissues of Adult Mouse-FGF-5 gene expression in adult mouse tissue was determined by RT-PCR analysis. Total RNAs from brain, skeletal muscle, heart, lung, gut, and liver were reverse transcribed and subjected to RT-PCR using the primer set, 5Ј-CCTTCGGGGCGCCGGACCGGCA-3Ј and 5Ј-CTTGGCTTTC-CCTCTCTTGTTC-3Ј, which should produce DNA fragments of 351 bp (mFGF-5) and 247 bp (mFGF-5S). The PCR products were separated by agarose gel electrophoresis followed by Southern blot hybridization with a mFGF-5 cDNA probe; a GAPDH primer set was used as a control in parallel with each sample. FGF-5 gene expression was predominantly seen in the brain (Fig. 1f), where the FGF-5S cDNA signal was also very strong; FGF-5S levels in the brain were only slightly lower than those of FGF-5. The high levels of expression in the brain of these 2 messages were reproducibly observed throughout repeated trials, which suggests that a relatively high copy number of FGF-5S mRNA must exist in the adult mouse brain. FGF-5 gene expression was also found in all other tissues studied, but the levels of expression were much lower than in the brain (Fig. 1f). FGF-5S expression was detected in skeletal muscle, lung, and occasionally in gut at very low levels but not in liver or heart. FGF-5:FGF-5S mRNA expression ratios were diverse among the tissues and are exemplified by the contrasting levels of expression in the brain and liver (Fig. 1f). Thus, both FGF-5S and FGF-5 mRNAs are most abundantly expressed in the brain, and the transcription and splicing of the FGF-5 gene appears to be differentially regulated in a tissuespecific manner.
Detection of FGF-5S mRNA in Human Brain-We next considered whether FGF-5S is expressed in the human brain as well as in mice. RT-PCR was performed using adult human cerebellar RNA and a set of new oligonucleotide primers for human FGF-5 (Fig. 2, a and b). The reaction amplified two DNA fragments with lengths of about 430 and 330 bp, respectively; on membranes, the fragments hybridized with an hFGF-5 probe (Fig. 2a, lane 3). The larger fragment was identical in size to one predicted from authentic hFGF-5 mRNA (13). The shorter fragment (331 bp) was cloned into a pBluescript SKϩ vector and sequenced. Its sequence revealed the fragment to be an alternatively spliced form of hFGF-5 mRNA generated, as was observed in mice, by direct linkage of the first and third exons (Fig. 2, b and c).
Effect of Endogenously Expressed hFGF-5 and hFGF-5S in PC12 Cells-Because we found that a neuron-like cell expressed both FGF-5 and FGF-5S, it seemed likely that, in the brain, an individual cell type might produce both splice variants. To test the effects of endogenously expressed hFGF-5 and hFGF-5S, PC12 cells were transfected with one or both of the mammalian expression vectors encoding for these proteins, and the stable transfectants were selected. Expression of the mRNAs was examined by RNase protection analysis (Fig. 3a), which revealed that protected RNA fragments of 611 bases for FGF-5 and/or 355 bases for FGF-5S were present in the cells. In culture, the PC12 transfectants expressing FGF-5 extended neurites reflecting the neurotrophic activity of the factor (Fig.  3c). In contrast, the transfectants expressing FGF-5S or transfectants expressing both FGF-5 and FGF-5S showed very few morphological changes (Fig. 3c).
Production of acetylcholinesterase was also analyzed as a marker of differentiation in PC12 cells (Fig. 3b). Acetylcholinesterase levels in PC12 transfectants expressing FGF-5 were about 1.8-fold higher than levels found in mock-transfected cells. In transfectants co-expressing FGF-5 and FGF-5S, however, acetylcholinesterase levels were similar to those observed in mock transfectants (Fig. 3b), and expression of FGF-5S alone also had a very weak effect on acetylcholinesterase activity. These results indicate that endogenous expression of FGF-5, but not FGF-5S, induces differentiation in PC12 cells. Moreover, when co-expressed, FGF-5S antagonizes FGF-5-induced neurotrophic activity.
Production of Recombinant hFGF-5S and hFGF-5 Proteins in E. coli-We expressed the human FGF-5S protein as part of a fusion protein with a maltose-binding protein, MalE; this allowed for its one-step purification from a lysate of the expressing bacteria. The FGF-5S polypeptide was purified by Mono-S ion exchange chromatography following cleavage of the fusion protein with a specific endopeptidase. The purified FGF-5S polypeptide appeared on SDS-polyacrylamide gels as a single 15-kDa band (Fig. 1d, lane 3); determination of the amino acid sequence verified that the protein had the expected amino terminus. To prepare the hFGF-5 protein, it was efficiently expressed using the T7 prokaryotic expression system and purified. The purified proteins were then used for biological assays. Consistent with an earlier observation (36), hFGF-5 and hFGF-5S proteins exhibited slower mobilities on SDSpolyacrylamide gels (Fig. 1d, lanes 2 and 3, respectively) than would be expected from their calculated molecular masses (27.2 and 10.7 kDa, respectively).
Heparin Binding Abilities of hFGF-5S and hFGF-5 Proteins-The heparin binding capabilities of recombinant hFGF-5S and hFGF-5 proteins were analyzed using immobilized heparin column chromatography. Bound hFGF-5 protein was eluted at approximately 1.4 M NaCl (Fig. 4, middle panel), which is in good agreement with previous findings using eukaryotically expressed hFGF-5 (1). The hFGF-5S polypeptide was also bound by the heparin column but eluted with 0.4 M NaCl (Fig. 4, left panel). Separating a mixture of these two forms of FGF-5 protein did not change their independent elution profiles (Fig. 4, right panel).
Effect of Exogenously Applied hFGF-5S Proteins on PC12 Cell Differentiation-The neurotrophic activities of bacterially produced hFGF-5 or hFGF-5S proteins were tested in PC12 cells (Fig. 5). Selected concentrations of one or the other protein were added to low serum test medium with or without heparin. Neurotrophic activity was then assessed by the resultant morphological changes in the cells. hFGF-5 protein clearly induced PC12 cells to assume a more neuron-like shape, and the addition of heparin enhanced this neurotrophic effect (Fig. 5a,  circles). hFGF-5S protein was a much weaker agonist, but a small amount of neurotrophic activity was detected at high concentrations, suggesting that FGF-5S may be a partial agonist (Fig. 5a, squares).
We next tested the effect of hFGF-5S on the capacity of hFGF-5 and other neurotrophic factors to induce differentiation in PC12 cells (Fig. 6). Various amounts of hFGF-5S polypeptide were added to PC12 cultures in the presence of heparin together with the indicated concentration of hFGF-5 FIG. 3. Effects of ectopic expression of hFGF-5 and hFGF-5S on PC12 differentiation. a, stable PC12 transfectants with their respective expression vectors were lysed, and the RNA was extracted. hFGF-5 (upper arrow) and hFGF-5S mRNAs (lower arrow) were detected by RNase protection as described under "Experimental Procedures." Positions of size markers are shown on the left with the number of bases. GAPDH mRNA was used as an internal control. The expression vectors used were pMexNeo (1), pMex/hFGF-5 (2), pMex/hFGF-5S (3), and pMex/hFGF-5 and pMex/hFGF-5S (4). b, the transfectants were lysed, and acetylcholinesterase activity was measured as described under "Experimental Procedures." c, morphology of the PC12 transfectants.
protein. We observed that increasing amounts of hFGF-5S partially suppressed PC12 cell differentiation elicited by 0.08 nM or 0.4 nM hFGF-5 ( Fig. 6a) but that excess amounts of FGF-5S were necessary for strong suppression.
To determine whether the suppressive effect of FGF-5S was specific for FGF-5 activity, we first examined its effect on NGF-induced cell differentiation. At a concentration of 0.4 nM, NGF exhibited the same degree of neurotrophic activity as 0.4 nM hFGF-5, but no suppressive effect of hFGF-5S was seen (Fig. 6b). Indeed, even at the lowest concentration tested (0.04 nM), NGF-induced neurotrophic activity was unaffected by hFGF-5S. The specificity of FGF-5S's antagonism of cell differentiation was further examined by simultaneously adding FGF-5S and FGF-1 or FGF-6 to PC12 cell cultures. FGF-1 efficiently activates all types of FGF receptors. In contrast, FGF-6 activates only a limited number of FGF receptors (FGFR-1 type IIIc, FGFR-2 type IIIc and FGFR-4), and its spectrum of activity is similar to that of FGF-5 (FGFR-1 type IIIc, FGFR-2 type IIIc) (26). As shown in Fig. 6c, suppression of FGF-1 neurotrophic activity by hFGF-5S was not detected even when hFGF-5S was present in high concentrations (up to 250 nM). On the other hand, the neurotrophic activity of FGF-6 appeared to be somewhat suppressed by hFGF-5S, although the effect was very weak. Thus, hFGF-5S does not act as an antagonist of either NGF-or FGF-1-induced PC12 cell differentiation. Indeed, the data suggest that only a limited number of FGF family members with specific binding characteristics are antagonized by hFGF-5S.
Binding of hFGF-5S to Receptors on PC12 Cells-The capacity of hFGF-5S protein to bind to receptors on the surface of PC12 cells was tested using competitive receptor binding assays. hFGF-5 protein was labeled with 125 I, and its binding to the high affinity FGF receptor on PC12 cells was assayed at 4°C in the presence of various concentrations of unlabeled FGF-5S or FGF-5 protein. As shown in Fig. 7, FGF-5S and FGF-5 proteins competitively antagonized 125 I-FGF-5 binding with equal potency, indicating that hFGF-5S and hFGF-5 proteins bind with similar affinities to the same high affinity FGF receptor(s) on PC12 cells.
Activation of FGF Receptor 1 Type IIIc by hFGF-5S-To study hFGF-5S activation of FGFR-1, we initially used wildtype PC12 cells and examined FGF-5S-induced tyrosine phosphorylation of the receptors and the downstream intracellular signaling. The very low levels of FGF receptor expression in PC12 cells, however, caused the analysis to be imprecise (data not shown). We, therefore, established stable transfectants of PC12 cells that overexpress FGFR-1 type IIIc on their surfaces. FGFR-1 type IIIc was selected because FGFR-1 plays the most important role in mediating FGF-induced neurotrophic activity in PC12 cells (46) and because FGF-5 activates FGFR-1 type IIIc most efficiently (26). Once the overexpression of FGFR-1 mRNA (Fig. 8a) and protein (not shown) in the transfectants (PC12/FGFR-1 cells) was confirmed, the PC12/FGFR-1 cells were treated with hFGF-5S or hFGF-5. Cell lysates were then prepared, FGFR-1 was immunoprecipitated, and tyrosine phosphorylation of FGFR-1 was assayed. It was observed that FGF-5S is a weak agonist for FGFR-1 (Fig. 8, b and c); at 50 nM, hFGF-5S induced tyrosine phosphorylation of FGFR-1, but the potency was much less than hFGF-5 at 2 nM (Fig. 8b). When cells were treated with FGF-5S at very high concentration (250 nM), the receptor phosphorylation became apparent (Fig. 8c). This correlates with the apparent induction of neurotrophic differentiation of PC12 cells by 250 nM hFGF-5S (Fig. 6c, filled  squares). Cross-linking Study-To further confirm that FGF-5S binds to FGFR-1, a cross-linking experiment was performed. The small number of tyrosine residues in the hFGF-5S primary structure did not allow for 125 I labeling with sufficiently high specific activity, and direct visualization of the cross-linked material using anti-FGF-5 antibodies did not provide sufficient sensitivity. Consequently, the MBP/hFGF-5S fusion protein was utilized to identify the FGF-5S binding protein on the cell surface. The MBP/hFGF-5S fusion protein partially retained the neurotrophic activity of FGF-5S (data not shown), and cross-linking MBP/hFGF-5S to the cell surface generated a complex of ϳ200-kDa (Fig. 8d). By subtracting the molecular mass of MBP/hFGF-5S (53 kDa), the molecular mass of the FGF-5S-binding protein on the cell surface was calculated to be ϳ150 kDa, which is similar to molecular mass of glycosylated FGFR-1 (47). In a control reaction where MBP/LacZ fusion protein (53 kDa) was used instead of MBP/hFGF-5S, no band corresponding to its complex was seen on the gel (Fig. 8d). Taken together with the results of the aforementioned binding assays (Fig. 7), this cross-linking experiment makes it highly likely that FGF-5S binds directly to FGFR-1.
hFGF-5 and hFGF-5S Activity on FGFR-1 Signaling-Because hFGF-5S appears to bind to FGFR-1, we investigated the extent to which hFGF-5S inhibits hFGF-5-mediated intracellular signaling in the receptor-dependent pathway. We found that excess hFGF-5S suppressed hFGF-5-induced tyrosine phosphorylation of FGFR-1 on PC12/FGFR-1 cells (Fig. 8e). Moreover, when the cells were stimulated with hFGF-5 in the presence of various concentrations of hFGF-5S, tyrosine phosphorylation of FRS2, a 90-kDa intracellular protein that is specifically and rapidly tyrosine-phosphorylated in response to FGF receptor activation (48), was inhibited in a concentrationdependent fashion (Fig. 8f). This result provides clear evidence that that FGF-5S suppresses intracellular signaling through FGFR-1. DISCUSSION We have identified two novel splice variants of the FGF-5 gene transcript; designated hFGF-5S and mFGF-5S, they were generated from single human and mouse genes, respectively. As demonstrated by RT-PCR and RNase protection analysis, FGF-5S mRNA was expressed abundantly in human and mouse brain as well as in neuron-like 102b67 cells, but it was poorly expressed in other tissues. In addition, we have confirmed that 102b67 cells secrete both FGF-5 and FGF-5S proteins. Assayed with respect to binding activity and neurotrophic activity in PC12 cells using recombinant proteins, we found that FGF-5S protein functions as a partial agonist/antagonist for FGF receptors. FGF-5S was found to bind to FGFR-1 type IIIc and antagonize FGF-5-mediated activation of the receptor, although at high concentrations FGF-5S elicits very weak signaling. A similar form of rat FGF-5 mRNA was detected solely by RT-PCR while this manuscript was in preparation (49), but its function has remained unstudied.
FGF-5 acts as a survival factor for spinal motoneurons (9, 10) and as a trophic factor for septal cholinergic and raphe serotonergic neurons (11), and it is expressed in adult brain (13,17,20). In that context, our present finding that the brain is the dominant tissue expressing FGF-5 mRNA suggests that it is very likely that FGF-5 serves as a physiological regulator in the neuronal system similar to FGF-1 and FGF-2 (2, 50). We tested this possibility using PC12 pheochromocytoma cells as a model, and found that the endogenous expression of FGF-5 or the addition of recombinant FGF-5 protein to cultures each exerted clear neurotrophic activity in PC12 cells. In contrast, the endogenous expression of FGF-5S mRNA or the addition of FGF-5S polypeptide exhibited very weak neurotrophic activity in the same experimental systems. Interestingly, co-expression of FGF-5S with FGF-5 in PC12 cells or the simultaneous addition of these proteins to PC12 cell cultures suppressed FGF-5mediated neurotrophic activity. Because FGF-5S did not inhibit cell differentiation induced by NGF or FGF-1, the receptor antagonist activity of FGF-5S is considered to be specific for a limited number of FGF family members and to regulate their activities in neural systems.
Our findings suggest that the partial agonist/antagonist effect of FGF-5S is exerted, at least in part, at the level of the cell surface receptor; in PC12 cells, FGF-5S was bound to FGFR-1 with an affinity similar to that of FGF-5, and it competitively inhibited high affinity FGF-5 binding. At high concentrations, FGF-5S suppressed FGF-5-induced activation of FGFR-1. It is interesting, then, that FGF-5S did not activate FGFR-1 efficiently. We interpret this to mean that the structure of FGF-5S is insufficient to evoke receptor dimerization and the subsequent intracellular signaling, but it is sufficient to compete with FGF-5 for its binding site on the receptor.
Sequence analysis revealed that both human and mouse FGF-5S mRNAs were generated by splicing out the second exons of the respective FGF-5 genes. Encoding for 123-and 121-amino acid polypeptides in human and mouse, respectively, the alternative splicing resulted in frameshifts and the introduction of immediate stop codons in the third exons. Mouse genomic DNA analysis revealed that FGF-5S and FGF-5 mRNAs originated from a single gene. Thus, it is clearly demonstrated that FGF-5 activity can be regulated at the level of its splicing and not solely at the level of transcription. This is in contrast to interleukin-1 ␣/␤, which are distantly related to FGFs (6). The interleukin-1 family has an antagonist (interleukin-1 receptor antagonist), encoded by an independent gene that has presumably been generated by gene duplication of an ancestral gene during evolution (51).
Messenger RNAs of both FGF-5S and FGF-5 were detected in the brain at all developmental stages and in adult cerebrum and cerebellum (20,21). Further suggesting that the FGF-5S plays an important role in brain, we found that both the absolute level of FGF-5S mRNA expression and its ratio to expression of FGF-5 mRNA were much higher in brain than in other organs. Simultaneous expression of FGF-5 and FGF-5S mRNAs proteins in 102b67 cells indicates that the alternative splice for FGF-5S and the subsequent translation of both messages can take place in a single cell type that expresses FGF-5. Actually, none of the tissues or cell lines examined so far express FGF-5S mRNA exclusively. It is therefore likely that a single cell type in the brain expresses both FGF-5 and FGF-5S, but this remains to be shown.
Alternative splicing such as we observed for FGF-5S has  1-4), and activated FRS2 was detected using anti-phosphotyrosine antibodies. The optical density of each signal is shown by the graph. been reported for FGF-1 (aFGF); as with the FGF-5 gene, the FGF-1 gene has three protein coding exons, and elimination of the second exon by alternative splicing generates a new termination codon yielding a much shorter form of the protein (6.7 kDa; aFGFЈ) (22). In BALB/c 3T3 cells, aFGFЈ and full-length FGF-1 bind to high affinity FGF receptors with similar affinities, but unlike FGF-1, aFGFЈ is barely able to elicit mitogenic responses (DNA synthesis and c-fos gene activation), and endogenous expression of aFGFЈ antagonizes the proliferative activity of the full-length FGF-1 (22). These findings are analogous to our results with FGF-5S and FGF-5. We did not detect aFGFЈ expression in brain at any developmental stage (see Ref. 20, Fig. 1) although the PCR primers used to amplify FGF-1 cDNA should have detected both FGF-1 and aFGFЈ mRNAs. Because the expression of aFGFЈ has only been described in fibroblast and medulloblastoma cell lines (22), expression of aFGFЈ may be limited to certain cell types/tissues as is the case for FGF-5S mRNA (Fig. 1f). Thus, aFGFЈ and FGF-5S are similar with respect to the mechanisms of their generation, their limited tissue distribution, and their antagonist activity. But because of their different distributions in tissues, they appear to have different functions.