Human fibroblast growth factor-18 stimulates fibroblast cell ... - Nature

Oncogene (1999) 18, 2635 ± 2642 ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $12.00 http://www.stockton-press.co.uk/onc

Human ®broblast growth factor-18 stimulates ®broblast cell proliferation and is mapped to chromosome 14p11 Mickey C-T Hu*,1, You-ping Wang1 and Wan R Qiu1 1

Department of Cell Biology and Functional Genomics, Amgen Inc., 14-1-D, Thousand Oaks, California, CA 91320, USA

The ®broblast growth factors (FGFs) play crucial roles in controlling embryonic development, cell growth, morphogenesis, and tissue repair in animals. Furthermore, FGFs may have a role in angiogenesis and may be involved in tumor invasion and metastasis. Here, we present the cloning and sequence of human FGF-18, a novel member of the FGF family. Sequence comparison indicates that FGF-18 is conserved with the other FGFs and most homologous to FGF-8 among the FGF family members. We showed that human FGF-18 was expressed primarily in the heart, skeletal muscle, and pancreas, and at lower levels in the other tissues. FGF-18 was also expressed at low levels in certain cancer cell lines. FGF18 contains a typical signal peptide and was secreted when it was transfected into mammalian cells. Recombinant FGF-18 protein stimulated proliferation in the ®broblast cell line NIH3T3 in a dose-dependent manner, suggesting that FGF-18 is a functional growth factor. Finally, the FGF-18 gene was evolutionarily conserved, and localized to human chromosome 14p11. Keywords: FGF; growth factor; cell proliferation; chromosome mapping

Introduction The ®broblast growth factors (FGFs) comprise a family of at least 18 growth factors and oncogenes that share a signi®cant degree of sequence homology and often exhibit overlapping biological activities (for reviews, see Fernig and Gallagher, 1994; Mason, 1994). Individual FGFs play important roles in various physiological and pathological processes, including embryonic development, cell growth, tissue repair, morphogenesis, in¯ammation, angiogenesis, vascular tone (Zhou et al., 1998), and tumor growth and invasion (reviewed by Mason, 1994). FGF-1 (acidic FGF) and FGF-2 (basic FGF) were identi®ed as mitogens for a variety of cell types (Gospodarowicz et al., 1978, 1986), whereas FGF-3 (int-2), FGF-4 (hst/KFGF) and FGF-5 were ®rst identi®ed as oncogenes (reviewed by Basilico and Moscatelli, 1992; Burgess and Maciag, 1989). FGF-6 was isolated according to its sequence similarity to FGF-4 (Marics et al., 1989), and FGF-7 (KGF) was identi®ed as a mitogen for epithelial cells (Rubin et al., 1989). Many new FGF members were recently isolated based on their sequence homology to known FGFs.

*Correspondence: M Hu Received 20 July 1998; accepted 15 December 1998

Thus far, the best characterized members are FGF-1 and FGF-2 which promote the growth of mesodermal and neuroectodermal cells during both embryogenesis and adulthood (Gospodarowicz et al., 1978, 1986). While FGFs may aect the pattern of dierentiation of ectodermal precursor cells in early embryos (Kimelman et al., 1988; Slack et al., 1988), the function of FGFs is often to stimulate tissue repair (wound healing) in the adult (Clarke et al., 1993; Cuevas et al., 1988). This repair function may be mobilized in certain pathological conditions; for instance, diseases of the retina, muscular dystrophy, rheumatoid arthritis and Alzheimer's disease (reviewed by Gospodarowicz, 1991). Furthermore, it appears that the inappropriate or altered expression of FGFs and their receptors occurs in a variety of cancers, including many common carcinomas (for a review, see Fernig and Gallagher, 1994). FGFs may also have a role in tumor vascularization. In fact, one of the ®rst angiogenic factors isolated from tumors was FGF-2 (basic FGF) (reviewed by Folkman and Klagsbrun, 1987). The ability of FGFs to stimulate the secretion of collagenase and plasminogen activator may be involved in tumor invasion and metastasis as well as angiogenesis. For instance, neovascularization and tumorigenicity of ®brosarcomas in transgenic mice carrying the bovine papilloma virus genome are associated with elevated FGF-2 secretion (Kandel et al., 1991). Here, we presented the cloning and sequence of human FGF-18, a novel member of the FGF family. Sequence analysis revealed that FGF-18 is signi®cantly conserved with the other FGF family members. We showed that human FGF-18 was expressed primarily in the heart, skeletal muscle, and pancreas, and at lower levels in certain cancer cell lines. Recombinant FGF-18 protein stimulated proliferation in the ®broblast cell line NIH3T3 in vitro in a dose-dependent manner, suggesting that FGF-18 is a functional growth factor. The FGF-18 gene is evolutionarily conserved, and localized to human chromosome 14p11. The full de®nition of the structures of all FGFs and their cognate receptors will enable a greater understanding, at the molecular level, of the role that FGFs play in developmental and pathological processes.

Results Molecular cloning and structure of human FGF-18 cDNA A 495 base pairs (bp) partial mouse cDNA sequence with signi®cant homology to FGF-8 was identi®ed from an Amgen EST database derived from a mouse

Human FGF-18 MC-T Hu et al

2636

fetal lung cDNA library. Using this mouse cDNA as a probe, we have also isolated two 1.6 kb cDNA clones from a human heart cDNA library. The nucleotide sequence of 1546 bp contains a single open reading frame of 621 bp encoding a polypeptide of 207 amino acids (Figure 1). After the initiation codon ATG, there is a stretch of 27 hydrophobic amino acids with the characteristics of a signal peptide (von Heijne, 1986), and there are two potential N-linked glycosylation sites having the consensus sequence N-X-S/T, suggesting that it is a candidate secreted glycoprotein. The initiation codon is preceded by three in-frame termination codons and presumably represents the true translation start site. A homology search of the available databases did not reveal any amino acid sequence identical to that of this clone. However, the deduced amino acid sequence of this cDNA is substantially homologous to those of FGF proteins, suggesting that it is a novel member of the FGF family. Therefore, we have designated this growth factor as FGF-18. Among all other FGF family

members, FGF-18 is most homologous to FGF-8 (Tanaka et al., 1995) and FGF-17 (Hoshikawa et al., 1998). Human FGF-18 shares 60% identity with human FGF-8 and 58% identity with human FGF17. The amino acid sequence of human FGF-18 was aligned with those of known FGF family members, using the CLUSTAL W program (Thompson et al., 1994) (Figure 2). Two cysteine (Cys) residues are conserved among the 18 known FGF family members (highlighted by asterisks), and the Cys127 of FGF-18 is conserved throughout the entire FGF family. Expression of FGF-18 in various human tissues and cancer cell lines To examine the tissue distribution of FGF-18, we investigated the level of FGF-18 mRNA in several human adult tissues by Northern blot analysis. A major FGF-18 transcript (approximately 4.5 kb) was identi®ed primarily in the heart, skeletal muscle and pancreas (Figure 3a). Lower-level expression of this

Figure 1 Nucleotide and predicted amino acid sequences of human FGF-18 cDNA. The deduced amino acid sequence of FGF-18 is indicated below the ®rst nucleotide of each codon and the termination codon is marked with an asterisk. The predicted signal peptide is underlined, the potential N-glycosylation sites are boxed, and the 5' end in-frame stop codons are highlighted in black boxes


transcript was detected in the rest of tissues except spleen where the transcript was barely detected. Another minor FGF-18 transcript (approximately 1.8 kb) was also detected in the heart. Since our Northern blot analysis was performed under high stringency and no other evident FGF members were picked up by the probe, this 2.0 kb transcript might be due to alternative RNA splicing. The 4.5 kb FGF-18 transcript was expressed at low level in several human cancer cell lines, while this transcript was not readily detected in the promyelocytic leukemia HL-60, colorectal adenocarcinoma SW480 and melanoma G361 cell lines. The 1.8 kb FGF-18 transcript was signi®cantly detected in the colorectal adenocarcinoma cell line SW480 that might be due to alternative splicing as described above. Furthermore, we examined the expression of FGF-18 mRNA in the mouse embryos. A Northern blot analysis of dierent days of embryos indicated that the expression of FGF-18 was not signi®cant until the 11-day's embryo and reached a plateau in the 15-day's embryo (data not shown). Analysis of recombinant human FGF-18 proteins produced in mammalian cells To determine whether human FGF-18 could be secreted in mammalian cells, localization of the Flagtagged FGF-18 (FGF-18F) protein in cultured 293EBNA cells transfected with the FGF-18F expression vector was examined by Western blotting with antiFlag M2 monoclonal antibody. The reactive protein was detected primarily in the culture medium with a molecular mass approximately 33 kD (Figure 3b, lane 2). Moreover, the amino acid sequence analysis of the puri®ed FGF-18F protein indicated that FGF-18F contained the N-terminal sequence of EENVDFRIHV as predicted (data not shown). These data indicated that this protein was secreted from the cells with proper cleavage of the N-terminal signal peptide. Recombinant FGF-18 stimulates proliferation of ®broblast cells Since it has been well documentated that FGF-1 can strongly activate DNA synthesis in a ®broblast cell line NIH3T3 (Gospodarowicz, 1974), we investigated whether FGF-18 protein could stimulate DNA synthesis in NIH3T3 cells in culture. Similar to FGF1 (positive control), recombinant FGF-18 exerted a dose-dependent increase of DNA synthesis in NIH3T3 cells (Figure 4), suggesting that it stimulates proliferation of ®broblast cells. In comparison to the negative control (PBS), one-half maximal stimulation of FGF18 occurred at *2 ng/ml. Furthermore, similar results were obtained with the MTS cell proliferation assay (Promega, Inc.), where FGF-18 showed a dosedependent augmentation on NIH3T3 cell growth (data not shown). However, the speci®c activity of FGF-18 on NIH3T3 cells was of a considerably lesser magnitude than that induced by FGF-1. The FGF-18 gene is a single copy in the human genome and evolutionarily conserved To examine whether the human genome contains a single FGF-18 gene, we digested human genomic DNA

with several restriction endonucleases as indicated and detected the FGF-18 gene by Southern blot analysis using FGF-18 cDNA as a probe. The genomic Southern blot showed that the patterns of restriction fragments hybridized to the FGF-18 cDNA probe were unique, and only one band was strongly positive in most patterns (Figure 5a). These results imply that the FGF-18 gene may be a single copy gene in the human genome. We further examined if the FGF-18 gene is conserved among dierent species. Thus, a zoo-blot analysis was performed, using FGF-18 cDNA as a probe. Most of the species tested, including yeast, showed speci®c bands hybridized to the probe (Figure 5b), suggesting that this gene is conserved evolutionarily. The FGF-18 gene is localized on human chromosome 14p11 Using the ¯uorescence in situ hybridization (FISH) technique (Heng et al., 1992; Heng and Tsui, 1993), the biotinylated FGF-18 cDNA probe was used to map the human chromosome. A speci®c region of one chromosome showed the FISH positive with the FGF-18 probe (Figure 6a). Under the condition used, the hybridization eciency was approximately 65% for this probe (among 100 checked mitotic ®gures, 65 of them revealed positive signals on one pair of the chromosomes). Since the DAPI banding was used to identify the speci®c chromosome (Figure 6b), the assignment between signal from the probe and the short arm of chromosome 14 was established. The detailed position was further determined based on the summary from ten photographs. There was no additional locus picked by FISH detection under the condition used; therefore, the gene of FGF-18 was mapped to human chromosome 14, region p11 (Figure 6c). Discussion We have recently identi®ed and cloned a novel FGF family member, human FGF-18, whose sequence is most similar to FGF-8 (AIGF) among the FGF family members. Unlike FGF-1, FGF-2, and FGF-9 (Miyamoto et al., 1993), which lack a secretory signal sequence, the ®rst 26 amino acids of human FGF-18 protein consist of hydrophobic residues that are predicted to be signal peptides for secretion. Indeed, the FGF-18 protein was secreted into the tissue culture media when it was expressed in 293-EBNA cells (see Figure 2b). The puri®ed recombinant human FGF-18 protein was biologically active in vitro. Like FGF-1, FGF-18 activated DNA synthesis and cell proliferation in the NIH3T3 ®broblast cell line. However, its speci®c activity on NIH3T3 cells was of a lesser magnitude than that of FGF-1. This result suggests that FGF-18 may be less potent than FGF-1 in ®broblasts. Alternatively, ®broblasts may not be the primary physiological target cell type for FGF-18. The mechanism by which FGFs stimulate cell proliferation is thought to be mediated by a dualreceptor system, i.e. FGF receptors (FGFR) and heparan sulfate proteoglycans (for reviews, see Fernig and Gallagher, 1994; Klagsbrun and Baird, 1991). The FGFR family consists of four distinct tyrosine kinase

2637


2638

receptors and many isoforms of these receptors which are generated by alternative RNA splicing. Therefore, a number of the FGFR isoforms can bind several FGFs. At present, it is not clear whether FGF-18 binds to any known member of the FGFR family or whether it requires a novel receptor to transduce its proliferative signal into cells. Identi®cation of the speci®c receptor for FGF-18 and its expression pattern in embryonic and adult tissues should help elucidate its physiological role. We showed that the human FGF-18 gene was evolutionarily conserved, and localized to human chromosome 14p11. Many human FGF genes have been mapped and most of them are not on the same chromosome (Verdier et al., 1997). Thus far, the FGF18 gene was the only FGF member localized to human

chromosome 14, indicating that this gene is probably not clustered with the other FGF family member's genes on the same chromosome. Although there are at least 18 members of the FGF gene family, these genes may not be duplicated from certain FGF genes during evolution. At present, it is unknown whether mutation or defect of the FGF-18 gene is involved in any diseases. Mutation analysis in patients with monogenic disorders that map to human chromosome 14p11 will evaluate this gene's involvement in these diseases. Moreover, targeted disruption (knockout) of the FGF-18 gene in mice may provide evidence for the relationship between its function and diseases. In summary, we have isolated, characterized, and studied the function of a novel member of the FGF family, human FGF-18, and determined its chromoso-


2639

Figure 2 Alignment of amino acid sequences of 18 FGF members using the single letter code. The sequences were aligned with the CLUSTAL W program (Thompson et al., 1994), and identical amino acids are boxed in black. The consensus (CONS) sequence is indicated at bottom and two conserved cysteines are marked by asterisks

Figure 3 mRNA expression patterns of human FGF-18 and recombinant FGF-18 protein expressed in 293-EBNA cells. (a) Northern blots of poly(A)+ RNAs from various human adult tissues and cancer cell lines were hybridized with the human FGF-18 cDNA probe. Human cancer cell lines are promyelocytic leukemia HL-60, epitheloid carcinoma HeLa S3, chronic myelogenous leukemia K-562, lymphoblastic leukemia MOLT-4, Burkitt's lymphoma Raji, colorectal adenocarcinoma SW480, lung carcinoma A549 and melanoma G361. As a control, the same blots were reprobed with human b-actin cDNA to check the integrity of the RNA (lower panel). (b) Western blot analysis of the Flag-tagged human FGF-18 (FGF-18/Flag) protein produced in 293-EBNA cells. The culture media of the untransfected (control, lane 1) and the FGF-18/Flag-transfected 293-EBNA cells (lane 2) were examined by Western blotting with anti-Flag M2 monoclonal antibody


2640

mal localization. FGFs are potent mitogens for a wide variety of cells of mesenchymal and neuroectodermal origin. Moreover, FGFs play a role in the differentiation of a variety of cells and are involved in

morphogenesis, development, and angiogenesis. Thus, identi®cation and characterization of a novel FGF may contribute to a better understanding of the role that FGFs have in normal development and pathological processes.

Materials and methods Isolation of human FGF-18 cDNA

Figure 4 Proliferation of NIH3T3 cells in response to recombinant human FGF-18 and FGF-1, using the BrdU labeling procedures. Cell cultures were prepared as described in Materials and methods. Increasing concentrations of FGF-18 (solid circle), or FGF-1 (solid square) in ng/ml was added. Phosphate-buered saline (PBS) was used as a negative control (open diamond). Proliferation was quanti®ed by measuring the absorbance at 490 nm using a scanning multi-well spectrophotometer (ELISA reader). Error bars indicate the mean and standard deviation for triplicate assay values

A novel mouse expressed sequence tag (EST) cDNA fragment of 495 base pairs (bp) with signi®cant homology to human FGF-8 was identi®ed in the Amgen EST data base. This EST cDNA was used as a probe to screen a human heart lTripEx cDNA library (Clontech Laboratories, Inc.). For hybridization, replicate ®lters were prehybridized for 1 h at 688C in Express hybridization buer (Clontech Laboratories) and hybridized 12 h at 688C in the same solution with the 32P-dCTP-labeled probe. After hybridization, the ®lters were washed several times at high stringency, at 658C in 0.26SSC (16SSC: 150 mM NaCl and 15 mM sodium citrate), 0.1% SDS and subjected to autoradiography. Twenty positive clones were picked and puri®ed after screening 46106 phages. The cDNA inserts of these positive phage clones were subsequently converted in vivo into pTripEx plasmid vector, according to the manufacturer's instructions (Clontech Laboratories). After analysis of the

Figure 5 Genomic Southern and zoo blot analyses of the human FGF-18 gene. Ten micrograms of human genomic DNA was digested with the indicated restriction enzymes and hybridized with the FGF-18 cDNA probe. l/HindIII DNA size markers were included in lane 1, and molecular sizes in kilobases (kb) are indicated. (b) A genomic Southern blot of EcoRI-digested DNA from various eukaryotic species (a zoo-blot) was hybridized with the FGF-18 cDNA probe. l/HindIII DNA size markers were included as previously described


inserts, the longest cDNA clone was sequenced on both strands, using a PCR procedure employing ¯uorescent dideoxynucleotides and a model 373A automated sequencer (Applied Biosystems). Sequence comparisons were aligned with the Best®t program of the GCG sequence analysis software package (Wisconsin Package Version 9.0) and the CLUSTAL W program (Thompson et al., 1994). Northern blot analysis Poly(A)+ RNAs from various human tissues or mouse embryos were obtained from Clontech Laboratories. Each sample (2 mg) was denatured and electrophoresed on a 1.2%

agarose gel containing formaldehyde and then transferred to a Hybond-N membrane (Amersham) in 206SSC as described (Sambrook et al., 1989). Human FGF-18 or bactin cDNA was labeled with 32P-dCTP to a speci®c activity of approximate 108 d.p.m./mg. Membranes were hybridized with either the human FGF-18 or b-actin cDNA probe, then washed at high stringency, at 658C in 0.26SSC, 0.1% SDS, and subjected to autoradiography. Probes were removed in 0.5% SDS at 95 ± 1008C. Southern blot analysis Ten micrograms of human genomic DNA (Clontech Laboratories) were digested with the following restriction enzymes (Boehringer Mannheim): BamHI, EcoRI, HindIII, PstI, SalI, XbaI, XhoI, NotI. The digested DNA samples and l/HindIII DNA size markers were electrophoresed, transferred to a Hybond-N+ membrane (Amersham), and hybridized with human FGF-18 cDNA probe as described above. Similarly, a genomic Southern blot of EcoRI-digested DNA from various eukaryotic species (a zoo-blot from Clontech Laboratories) was hybridized with the human FGF18 cDNA probe. Probes were removed as previously described. Production and puri®cation of recombinant human FGF-18 Full-length human FGF-18 cDNA (amino acids 1 ± 207) was cloned into the mammalian expression vector pCEP4 (Invitrogen, Inc., Carlsbad, CA, USA) by polymerase chain reaction (PCR) using the two following forward and reverse primers: 5'-TATAAGCTTGGTACCGCCACCATGTATTCAGCGCCCTCCGCC-3' and 5'-TATTGCGGCCGCTTATCATTTATCATCATCATCTTTATAATCGCCGGGGTGAGTGGGGCGGATC-3'. The second primer included a Flag tag at the 3' end, and the resulting plasmid was designated pCEP/FGF-18/Flag. 293-EBNA cells were grown in Dulbecco's modi®ed Eagle's medium (DMEM) (Gibco ± BRL, Gaithersburg, MD, USA) supplemented with 10% fetal bovine serum (FBS) (Gibco ± BRL). Cells to be transfected were plated the day before transfection at a density of 26106 cells per 100 mm dish. Cells were transfected with expression plasmid (10 mg per dish), using the calcium phosphate precipitation protocol (Specialty Media, Inc.). The positive cell clones were isolated by hygromycin B (Boehringer Mannheim, Corp., Indianapolis, IN, USA) selection and expanded in serum-free medium. The conditioned medium was loaded onto an S-Sepharose column (Pharmacia, Inc., Piscataway, NJ, USA) equilibrated with 10 mM sodium phosphate, pH 7.0. The bound proteins were eluted with a linear NaCl gradient from 0.5 ± 3 M in the same (phosphate) buer. The fractions containing recombinant FGF-18, as determined by SDS ± PAGE, were eluted at 2 M NaCl. These fractions were pooled and dialyzed against the phosphatebuered saline using the same dialysis membrane, sterilely ®ltered, and stored at 4 8C. NIH3T3 cell proliferation assay in vitro

Figure 6 Chromosomal localization of human FGF-18 gene by FISH. (a) The FISH signals (arrows) on a representative metaphase spread. (b) The respective 4',6'-diamidino-2-phenylindole (DAPI) banding patterns of the chromosomes. (c) Diagrammatic representation of map assignments for several metaphase spreads. Dots indicate the number of spreads that showed the double FISH signals at the indicated chromosome band

NIH3T3 cells were cultured in DMEM supplemented with 10% FBS. Cells were collected by trypsinization and a cell suspension (7500 cells per 100 ml) was dispensed into each well of 96-well tissue culture microtiter plates (Falcon, Becton Dickinson Labware, Lincoln Park, NJ, USA). After 24 h, the medium was changed to DMEM-0.1% FBS (starvation medium), and culturing was continued for 24 h. Growth factors (human FGF-18 or FGF-1) were added to each well at this time, and 5-bromo-2'-deoxy-uridine (BrdU) (Boehringer Mannheim, Corp.) was added to each well after 16 h. Recombinant human FGF-1 was included in these assays as a positive control, while phosphate-buered saline (PBS) was

2641


2642

used as a negative control. After 5 h incubation, the cells were ®xed with 70% ethanol containing 0.5 M HCl for 30 min at 7208C. Subsequently, DNA synthesis was assayed quantitatively by measuring BrdU incorporation into cellular DNA with an anti-BrdU antibody according to a standard cell ELISA protocol. The cells were treated with nuclease solution for 30 min at 378C, washed, and incubated with the anti-BrdU antibody labeled with peroxidase (Boehringer Mannheim, Corp.) for 30 min at 378C. After washing, the samples were incubated with peroxidase substrate at room temperature for 15 min. The absorbance (490 nm) in the samples was determined by using a scanning multi-well spectrophotometer (ELISA reader, Molecular Devices, Corp., Sunnyvale, CA, USA) with the excitation at 405 nm. Similarly, the tetrazolium MTS cell proliferation assay (Promega, Inc.) was performed according to the manufacturer's instruction. Cell culture and microscope slides preparation Lymphocytes isolated from human blood were cultured in aminimal essential medium (MEM) supplemented with 10% fetal bovine serum and phytohemagglutinin at 378C for 68 ± 72 h. The lymphocyte cultures were treated with BrdU (0.18 mg/ml, Sigma) to synchronize the cell population. The synchronized cells were washed three times with serum-free medium to release the block and recultured at 378C for 6 h in a-MEM with thymidine (2.5 mg/ml, Sigma). The cells were harvested and the cell slides were prepared by using standard procedures including hypotonic treatment, ®xation, and airdrying.

Fluorescence in situ hybridization (FISH) The procedure for FISH detection was performed as previously described (Heng et al., 1992; Heng and Tsui, 1993). Brie¯y, the cell slides were baked at 558C for 1 h. After RNase treatment, the slides were denatured in 70% formamide in 26SSC for 2 min at 708C followed by dehydration with ethanol. DNA probes were labeled with biotinylated dATP at 158C for 1 h, using the BRL BioNick labeling kit (Gibco ± BRL). Probes were denatured at 758C for 5 min in a hybridization buer containing 50% formamide and 10% dextran sulfate, and loaded onto the denatured chromosomal slides. After 16 ± 20 h hybridization, the slides were washed and incubated with ¯uorescein isothiocyanate (FITC)-conjugated avidin (Vector Laboratories), and the signal was ampli®ed as described (Heng et al., 1992). FISH signals and the 4',6'-diamidino-2-phenylindole (DAPI) banding patterns were recorded separately by taking photographs, and the assignment of the FISH mapping data with chromosomal bands was achieved by superimposing FISH signals with the DAPI banded chromosomes (Heng and Tsui, 1993).

Acknowledgements We thank B Sutton for DNA sequencing; Dr H Heng for assisting in FISH; J Speakman for the MTS cell proliferation assay; H Lu for partial amino acid sequence determination; E Han for ELISA data analysis; D Olivares for technical illustration; and Drs L Souza, R Bosselman and W Boyle for their support.

References Basilico C and Moscatelli D. (1992). Adv. Cancer Res., 59, 115 ± 165. Burgess WH and Maciag T. (1989). Annu. Rev. Biochem., 58, 575 ± 606. Clarke MSF, Khakee R and McNeil PL. (1993). J. Cell Sci., 106, 121 ± 133. Cuevas P, Burgos J and Baird A. (1988). Biochem. Biophys. Res. Commun., 156, 611 ± 618. Fernig DG and Gallagher JT. (1994). Prog. Growth Factor Res., 5, 353 ± 377. Folkman J and Klagsbrun M. (1987). Science, 235, 442 ± 447. Gospodarowicz D. (1974). Nature, 249, 123 ± 127. Gospodarowicz D. (1991). Cell Biol. Rev., 25, 307 ± 314. Gospodarowicz D, Greenburg G and Bialecki H. (1978). In Vitro, 14, 85. Gospodarowicz D, Neufeld G and Schweigerer L. (1986). Cell Dier., 19, 1 ± 17. Heng HHQ, Squire J and Tsui L-C. (1992). Proc. Natl. Acad. Sci. USA, 89, 9509 ± 9513. Heng HHQ and Tsui L-C. (1993). Chromosoma, 102, 325 ± 332. Hoshikawa M, Ohbayashi N, Yonamine A, Konishi M, Ozaki K, Fukui S and Itoh N. (1998). Biochem. Biophys. Res. Commun., 244, 187 ± 191. Kandel J, Bossy-Wetzel E, Radvanyi F, Klagsbrun M, Folkman J and Hanahan D. (1991). Cell, 66, 1095 ± 1104. Kimelman D, Abraham JA, Haaparanta T, Palisi TM and Kirschner MW. (1988). Science, 242, 1053 ± 1056.

Klagsbrun M and Baird A. (1991). Cell, 67, 229 ± 231. Marics I, Adelaide J, Raybaud F, Mattei M, Coulier F, Planche J, Lapeyriere O and Birnbaum D. (1989). Oncogene, 4, 335 ± 340. Mason IJ. (1994). Cell, 78, 547 ± 552. Miyamoto M, Naruo K-I, Seko C, Matsumoto S, Kondo T and Kurokawa T. (1993). Mol. Cell. Biol., 13, 4251 ± 4259. Rubin JS, Osada H, Finch PW, Taylor WG, Rudiko S and Aaronson SA. (1989). Proc. Natl. Acad. Sci. USA, 86, 802 ± 806. Sambrook J, Fritsch EF and Maniatis T. (1989). Molecular Cloning: A Laboratory Manual. 2nd ed. Cold Spring Harbor, New York: Cold Spring Harbor Laboratory. Slack JMW, Isaacs HV and Darlington BG. (1988). Development, 103, 581 ± 590. Tanaka A, Miyamoto K, Matsuo H, Matsumoto K and Yoshida H. (1995). FEBS Lett, 363, 226 ± 230. Thompson JD, Higgins DG and Gibson TJ. (1994). Nucleic Acids Res., 22, 4673 ± 4680. Verdier A-S, Mattei M-G, Lovec H, Hartung H, Goldfarb M, Birnbaum D and Coulier F. (1997). Genomics, 40, 151 ± 154. von Heijne G. (1986). Nucleic Acids Res., 14, 4683 ± 4690. Zhou M, Sutli RL, Paul RJ, Lorenz JN, Hoying JB, Haudenschild CC, Yin M, Con JD, Kong L, Kranias EG, Luo W, Boivin GP, Duy JJ, Pawlowski SA and Doetschman T. (1998). Nature Med., 4, 201 ± 207.