Semirational design of a potent, artificial agonist of fibroblast growth factor receptors

© 1999 Nature America Inc. • http://biotech.nature.com

RESEARCH

Semirational design of a potent, artificial agonist of fibroblast growth factor receptors Marcus D. Ballinger, Venkatakrishna Shyamala, Louise D. Forrest, Maja Deuter-Reinhard, Laura V. Doyle, Jian-xin Wang, Lootsee Panganiban-Lustan, Jennifer R. Stratton, Gerald Apell, Jill A. Winter, Michael V. Doyle, Steven Rosenberg, and W. Michael Kavanaugh* Chiron Corporation, 4560 Horton St., Room 4.4144, Emeryville, CA 94608. *Corresponding author ([email protected]).

© 1999 Nature America Inc. • http://biotech.nature.com

Received 12 May 1999; accepted 13 September 1999

Fibroblast growth factors (FGFs) are being investigated in human clinical trials as treatments for angina, claudication, and stroke. We designed a molecule structurally unrelated to all FGFs, which potently mimicked basic FGF activity, by combining domains that (1) bind FGF receptors (2) bind heparin, and (3) mediate dimerization. A 26-residue peptide identified by phage display specifically bound FGF receptor (FGFR) 1c extracellular domain but had no homology with FGFs. When fused with the c-jun leucine zipper domain, which binds heparin and forms homodimers, the polypeptide specifically reproduced the mitogenic and morphogenic activities of basic FGF with similar potency (EC 50 = 240 pM). The polypeptide required interaction with heparin for activity, demonstrating the importance of heparin for FGFR activation even with designed ligands structurally unrelated to FGF. Our results demonstrate the feasibility of engineering potent artificial agonists for the receptor tyrosine kinases, and have important implications for the design of nonpeptidic ligands for FGF receptors. Furthermore, artificial FGFR agonists may be useful alternatives to FGF in the treatment of ischemic vascular disease. Keywords: fibroblast growth factor, protein design, receptors, ligands, cytokines, mimetics, angiogenesis

The fibroblast growth factors (FGFs) are a large family of pleiotropic growth factors that promote survival, proliferation, and differentiation of a variety of cell lines and tissues 1,2. Treatment with FGF induces formation of new blood vessels and improves blood flow and organ function in animal models of peripheral vascular disease3,4 and myocardial ischemia5–7. In addition, FGF treatment can reduce infarct size following experimental coronary and cerebral artery occlusion8,9. Clinical studies with FGF and other angiogenic molecules have suggested a need to develop therapeutic molecules with improved properties. For example, the plasma half-life of FGFs is short, implying that repeated administration of recombinant protein, or gene therapy, may be necessary for optimal therapeutic effects. Furthermore, the existence of multiple FGF receptor subtypes in various tissues suggests that molecules with greater receptor specificity may have fewer systemic toxicities. We wished to determine whether it is possible to use a semirational molecular design approach to engineer artificial ligands that reproduce the activities of FGF. Such molecules might ultimately provide useful alternatives to FGF in the treatment of ischemic vascular disease with improved pharmacokinetics, receptor specificity, and manufacturing costs. Fibroblast growth factor mediates its biological effects by binding to the FGF receptor (FGFR) tyrosine kinase. Peptidic and nonpeptidic cytokine mimetics capable of activating receptors that signal through JAK/STAT (Janus kinase/signal transducer and activator of transcription) pathways recently have been discovered10–12. Furthermore, a nonpeptidic agonist of the insulin receptor with micromolar potency was recently reported13. However, novel agonists of the receptor tyrosine kinases with potencies approaching those of the corresponding endogenous ligands have not yet been NATURE BIOTECHNOLOGY VOL 17 DECEMBER 1999

http://biotech.nature.com

identified. The rational design of agonists for the FGFRs is particularly challenging, because initiation of signaling requires formation of a complex between FGF, FGFR, and heparin by a mechanism that is poorly understood. However, evidence strongly suggests that activation of FGFRs requires: (1) high-affinity binding of FGF to the FGFR extracellular domain; (2) binding of FGF to heparan sulfate proteoglycans or heparin glycosaminoglycans on the cell surface or extracellular matrix; and (3) induction of FGFR dimerization or oligomerization14–20. We reasoned that by identifying separate functional domains that mediate each of these processes and by combining them into a single fusion protein, we could engineer a novel molecule that efficiently activates FGFRs. We report the successful design and construction of a small polypeptide ligand for FGFRs that is structurally unrelated to all known FGFs. The ligand potently and specifically reproduces the biological effects of basic FGF (bFGF; FGF-2) and, like bFGF, requires interaction with heparin for activity. Results Identification of a FGFR-binding peptide. Phage display was used to identify novel peptides that specifically bind to the extracellular domain (ECD) of FGFR 1c, which is one of the major FGF receptor isoforms that bind bFGF. A library of ∼107 different, random, 26residue peptides displayed on the NH 2 terminus of the pIII protein of M13 phage was alternately selected against immobilized purified FGFR ECD protein and against Sf9 cells overexpressing FGFRs on their surface. We identified a phage (designated C19) that was more than 1,000-fold enriched after selection, compared with control phage that expressed either wild-type pIII or control peptide–pIII fusion sequences (recovery of input phage: C19, 0.92%; control, 1199

© 1999 Nature America Inc. • http://biotech.nature.com

RESEARCH

C19 Sequence

6X His tag

S

S

S

S

nt ro l bF GF C1 9j un

c-jun leucine zipper

nt ro bF l GF C1 9j un

(GGGS)5 linker

AA

co

Dimerization

co

Receptor Binding

- 51

PhosphoMAPK

- 207

IP: FGFR Blot: APT

- 51

- 121

Total MAPK

- 81

C C

BB 0.5

1.4 1.2

0.4 0.3 0.2

© 1999 Nature America Inc. • http://biotech.nature.com

0

0.0005%). C19 did not bind to immobilized erythropoietin receptor ECD protein under similar conditions, demonstrating that the binding was specific (data not shown). The sequence of the C19 peptide (Fig. 1) was not homologous to FGFs or to any other protein. Construction of an FGFR agonist. To create an FGFR agonist, we combined the C19 peptide insert sequence with a protein domain that mediated dimerization and bound heparin. The leucine zipper region of the transcription factor c-jun was chosen because it is small (39 amino acids), is well characterized structurally, forms homodimers when expressed in Escherichia coli21, and binds heparin in vitro (data not shown). The C19 sequence was fused to residues 276–314 of human c-jun through a flexible linker sequence, and a polyhistidine tag was placed at the C terminus for convenient affinity purification (Fig. 1). The c-jun leucine zipper domain is expressed as a noncovalent dimer that can disassociate in vivo 22. Therefore, cysteines were placed on either end of this domain to direct covalent homodimerization23,24 and ensure that the fusion protein was maintained as a dimer under experimental conditions. As predicted, purified C19–c–jun fusion protein (C19jun) was principally expressed as a 22 kDa homodimer. It bound immobilized FGFR ECD protein in vitro with high affinity and specificity (Table 1; also see below) and quantitatively and specifically bound to heparin agarose beads through the leucine zipper domain (data not shown). C19jun potently reproduces the activities of bFGF. To determine whether C19jun could activate FGFRs in vivo, Swiss 3T3 cells were treated with bFGF or purified C19jun protein. Cell extracts were then immunoprecipitated with anti-FGFR antibodies, analyzed by SDS–PAGE, and immunoblotted with antiphosphotyrosine antibodies. Extracts from bFGF- and C19jun-stimulated 293 cells expressing FGFR 1c were also immunoblotted with phosphoMitogen-Activated Protein (MAP) kinase-specific antibodies. Both bFGF and C19jun induced FGFR autophosphorylation and MAP kinase phosphorylation (Fig. 2A). These results demonstrate that C19jun can activate FGFRs in cells and initiate intracellular signaling. To examine whether C19jun is a mitogen for cells expressing FGFR, Swiss 3T3 fibroblasts were stimulated with varying concentrations of C19jun in the presence of 15 U/ml heparin and assayed 24 h later for incorporation of BrdU, a measure of induction of DNA synthesis and entry into S phase. C19jun induced BrdU incorporation in these cells, with a potency and maximal response similar to that of bFGF (EC50 = 240 versus 140 pM) (Fig. 2B). Heparin alone had no effect, nor did a control protein expressed and purified in the same manner in which the C19 sequence was deleted (Fig. 5A; see Fig. 3 and below for additional controls). C19jun (2 nM) also stimulated proliferation of these cells in long-term assays to an extent equal to or greater than 1 nM bFGF (Fig. 2C), and promoted the proliferation of human endothelial cells in similar assays. Thus, 1200

bFGF C19-jun

0.1

D

0.01

0.1

1

OD (460-690nm)

Figure 1. Structure of C19jun. The sequence of C19 was followed by: five repeats of the sequence Gly-Gly-Gly-Ser (GGGS) as a flexible linker; the leucine zipper domain of c-jun (residues 276–314) flanked by cysteines to direct covalent homodimerization; a second short linker segment Gly-Gly-Ser-Gly-Gly; and a polyhistidine affinity tag at the C terminus. The sequence of C19 is: NH 2-AESGDDYCVLVFTDSAWTKICDWSHFRN-COOH

OD (370-492 nm)

Heparin Binding

C19-jun

1 0.8 0.6

bFGF

0.4 0.2

10

control

0 0

[Ligand], nM

1

2

3

Time (days)

D

control

bFGF (1 nM)

C19jun (2 nM)

Figure 2. C19jun is a potent FGFR agonist. (A) Swiss 3T3 cells were stimulated with vehicle (control), bFGF, or refolded C19jun protein for 15 min at 37oC and lysed. Immunoprecipitated (IP) FGFR was immunoblotted with antiphosphotyrosine (APT) antibodies (left panel). 293 cells expressing full-length FGFR 1c were stimulated with bFGF or refolded C19jun protein, and aliquots of lysate were immunoblotted with phospho-MAP kinase-specific antibodies (right panel, top), or anti-total MAP kinase antisera as a control (right panel, bottom). Molecular mass markers in kilodaltons are shown at right. (B) Colorimetric ELISA of BrdU incorporation in Swiss 3T3 fibroblasts stimulated with the indicated concentrations of bFGF or C19jun and heparin for 24 h. (C) Colorimetric assay of Swiss 3T3 cell proliferation following a single addition of 2 nM C19jun or 1 nM bFGF with heparin or heparin alone (control) on day 0. (D) Neurite outgrowth in PC12 cells stimulated for four days with the indicated concentrations of bFGF or C19jun and heparin, or heparin alone (control). (A and D) Representative experiments. Data in (B) and (C) are averages of triplicates ± standard deviations.

C19jun reproduces the mitogenic effects of bFGF on multiple FGFresponsive cell types (data not shown). In addition to being a mitogen, FGF is also capable of inducing differentiation in some cells. PC12 cells respond to FGF treatment with the extension of neurite outgrowths and eventual growth arrest25,26. To determine whether C19jun can induce differentiation of these cells, PC12 cells were treated with 1 nM bFGF or 2 nM C19jun for four days in serum-containing media and examined for induction of neurite outgrowth. C19jun stimulated neurite outgrowths to a similar extent as bFGF in the presence of heparin, while heparin alone had no effect (Fig. 2D). Therefore, C19jun potently reproduced the morphogenic activity of bFGF. Furthermore, as PC12 cells can proliferate without differentiation in response to treatment with other growth factors, such as epidermal growth factor26, C19jun is not a nonspecific mitogen for cells. C19jun is a specific FGFR agonist. Swiss 3T3 cells were stimulated with C19jun in the presence of protein representing the soluble NATURE BIOTECHNOLOGY VOL 17 DECEMBER 1999

http://biotech.nature.com

© 1999 Nature America Inc. • http://biotech.nature.com

RESEARCH

2

0.1

1 10 [agonist], nM

bFGF

5 0

0.01

0.1

1

10

100

http://biotech.nature.com

120 100

A

B

80 60 40 20 0

Ctrl

C19jun

120 100

Ctrl

jun

C19jun C19pep

D

C

E

80 60 40 20

- Heparin

+ Heparin

C19jun+ C19-Ig

C19-Ig

0

C19jun

ECD of various receptors used as binding competitors. The mitogenic activity of C19jun was abolished by preincubation of the fusion protein with excess FGFR 1c ECD, but was unaffected by incubation with identical concentrations of the related plateletderived growth factor receptor ECD or with erythropoietin receptor ECD (Fig. 3A). Therefore, C19jun binds specifically to FGFRs. As an additional test of specificity, the activities of C19jun and bFGF were assayed in L6 myoblast cell lines stably expressing FGFR 1c and in control L6 cells that do not express FGFR27. Both bFGF and C19jun stimulated c-fos mRNA expression in the FGFR-expressing cells, but neither was active in control cells (Fig. 3B). Similar results were observed when phosphorylation of MAP kinase was analyzed in these cells (data not shown). These results demonstrate that C19jun activity is specific for the FGFR and requires the presence of the FGFR for activity. Analysis of C19jun binding to FGFRs. The experiments described above demonstrate that C19jun is a potent FGFR agonist with activities similar to those of bFGF. Data from several different experiments suggested that C19jun bound to an FGFR region that is similar, but not identical, to that bound by bFGF. First, a series of soluble FGFR–Fc fusion proteins that contained deletions in the ECD of FGFR were generated and tested for their ability to bind to C19jun and competitively inhibit its activity in mitogenesis assays as described above. A construct that lacked the first Ig-like domain and the acid box of FGFR1c (dIg1dAB:Fc) bound C19jun as effectively as a construct containing the entire ECD sequence (FGFR:Fc) (Fig. 3A). Therefore, C19jun interacts with the second and/or third Ig-like loops of FGFR 1c, as does bFGF28. Second, we wished to determine whether the specificity of C19jun for different FGFRs was similar to that of bFGF. Using the same approach as described in Figure 3A and above, we tested the binding of C19jun to the soluble form of three different FGFRs: 1b, 1c (also known as the bFGF receptor), and 2b (also known as the Keratinocyte Growth Factor [KGF] receptor). C19jun bound to FGFR 1c, but bound poorly or not at all to FGFR 1b and FGFR 2b (Table 2). This pattern of receptor binding was similar to that of bFGF assayed in the same experiment, but was different from acidic FGF, which bound well to all three receptors. Moreover, because

FGFR 1c and FGFR 1b differ only by alternative splicing in the second half of the third Ig-like domain, the data suggest that this region determines C19jun receptor specificity, as it does for endogenous FGFs27. Finally, we compared the ability of C19jun and bFGF to compete for binding of 125I-labeled bFGF to FGFR 1 in vitro. Both bFGF and C19jun competed for binding with an IC50 of approximately 3 nM under these conditions (Fig. 4). However, only approximately 50% of 125I-bFGF binding could be inhibited by C19jun even at high concentrations, whereas unlabeled bFGF was able to completely inhibit 125IbFGF binding. This pattern of incomplete inhibition is typical for a negatively cooperative, noncompetitive binding antagonist that interacts at a separate site29. Taken together, the data suggests that, although C19jun interacts with some of the same regions on FGFRs as does bFGF, C19jun and bFGF do not occupy precisely the same binding site.

control

Figure 3. Specificity of C19jun for FGFR. (A) Mitogenesis assay of 10 nM C19jun preincubated with 100 nM of the indicated receptor extracellular domain-.Fc fusion proteins (denoted by:Ig). unstim, unstimulated; FGFR, full-length FGFR 1 ECD; PDGFR, plateletderived growth factor receptor; EPOR, erythropoietin receptor; dIg1dAB, FGFR1 ECD with deletions of the first Ig-like domain and the acid box. Data are averages of triplicates ± standard deviations. (B) bDNA assay of endogenous c-fos mRNA levels induced in FGFRexpressing (closed symbols) or control L6 myoblast cell lines, which do not express FGFR (open symbols) by the indicated concentrations of bFGF (squares) or C19jun (circles). Shown is a representative experiment, expressed as counts per second (CPS).

Figure 4. Competition of C19jun and bFGF for 125I-bFGF binding to FGFR1. 125I-bFGF was incubated with FGFR1 immobilized on microtiter plates in the presence of the indicated concentrations of C19jun or unlabeled bFGF, washed extensively, and the total amount of radioactivity bound was determined as described in Experimental Protocol.

Mitogenesis (% max FGF Response)

© 1999 Nature America Inc. • http://biotech.nature.com

C19-jun

NATURE BIOTECHNOLOGY VOL 17 DECEMBER 1999

10

[Competitor], nM

Mitogenesis (% max FGF Response)

+ PDGFR:Fc

+ EPOR:Fc

+ dIg1dAB:Fc

+ FGFR:Fc

0

control

unstim

0

4

15

C19jun∆Hep

0.1

L6 + bFGF L6-FGFR + bFGF L6 + C19-jun L6-FGFR + C19-jun

C19jun

0.2

6

control

0.3

8

C19jun

20

C19jun

0.4

10

control

0.5

25

C19jun

0.6

control

c-fos mRNA (CPS x 103)

OD (370-492 nm)

0.7

125I-bFGF bound (cpm x 103)

B 12

A

Figure 5. Activities of C19jun variants. Mitogenesis assays of Swiss 3T3 fibroblasts stimulated with 10 nM refolded C19jun protein, with vehicle (control), and with: (A) 10 nM control protein in which the C19 sequence was deleted (jun); (B) 1 µM synthetic C19 peptide (C19pep), without the leucine zipper region; (C) control or C19jun in the absence or presence of 15 U/ml heparin; (D) 17 nM C19jun∆hep protein; and (E) 1 µM C19-Ig alone or in combination with 10 nM C19jun. Assays were performed and data presented as in Figure 2B. 1201

© 1999 Nature America Inc. • http://biotech.nature.com

RESEARCH Table 1. Approximate affinities of C19 variants for FGFR1c ECD1. Kd (nM) C19 synthetic peptide Jun leucine zipper only C19jun C19 junghep C19-Ig

400 >10,000 10 10 90

© 1999 Nature America Inc. • http://biotech.nature.com

1Apparent affinities for immobilized dimeric FGFR1c:Fc protein were determined by BIAcore real-time kinetic analysis in the absence of heparin as described in Experimental Protocol.

Structure-function analysis of C19jun. We next investigated the structural requirements for C19jun activity. Deletion of the C19 sequence from the leucine zipper fusion protein abrogated mitogenic activity (Fig. 5A). A synthetic peptide containing the C19 sequence with an intramolecular disulfide bond was able to bind to FGFR 1c (Table 1). However, this peptide, which does not include the c-jun leucine zipper domain, did not induce mitogenesis even at high concentrations (Fig. 5B). Mutation of either of the two cysteines in the C19 peptide sequence to alanines eliminated the ability of C19 to bind to the FGFR (data not shown). These experiments demonstrate that both the C19 peptide and the leucine zipper domain are required for C19jun mitogenic activity. Furthermore, the cysteines in the C19 sequence are essential for FGFR binding, probably by constraining C19 peptide structure through an intramolecular disulfide bond. To determine the importance of heparin to C19jun activity, Swiss 3T3 fibroblasts were stimulated with C19jun in the presence and absence of heparin. The addition of heparin was required for optimal mitogenic activity of the C19jun fusion protein (Fig. 5C). Heparin was also required for the morphogenic activity of C19jun on PC12 cells (data not shown). To determine whether the heparinbinding site on C19jun was required for activity, two lysines and two arginines predicted to comprise a major portion of the heparinbinding site in the c-jun leucine zipper domain were changed to glutamines. The mutated C19jun protein (C19jun∆hep) was expressed as a homodimer and bound FGFR ECD in the absence of heparin with the same apparent affinity as C19jun (data not shown, and Table 1). However, unlike C19jun, C19jun∆hep protein had very low affinity for heparin and was inactive in Swiss 3T3 mitogenesis assays (data not shown and Fig. 5D). Similar results were obtained when the c-jun leucine zipper was replaced with the Fc portion of IgG1 (C19-Ig), which also mediates spontaneous dimerization but does not bind heparin. C19-Ig was expressed as a homodimer and bound FGFR with high apparent affinity (Table 1) but bound heparin poorly and did not stimulate mitogenesis at concentrations up to 1 µM (data not shown and Fig. 5E). Preincubation of cultures with 1 µM C19-Ig completely inhibited C19jun-induced mitogenic activity, demonstrating that C19-Ig was competent for receptor binding but not for receptor activation (Fig. 5E). Taken together, these experiments demonstrate that the interaction of heparin with C19jun is essential for stimulation of mitogenesis. Discussion Previously, the only molecules shown to bind FGFRs have been FGFs, peptide fragments of FGF, or peptide sequences with homology to FGFs16,30,31. The C19 sequence is not homologous to any member of the FGF family, and does not bind to precisely the same region of the FGFR as does bFGF (Fig. 3A, Fig. 4, and Table 2). Therefore, the design of FGFR agonists by the methods described in this report is not limited to molecules that are structurally similar to FGFs or that precisely interact with the FGF binding site on the receptor. Peptides derived from FGF sequences previously have been reported to bind to FGFR at approximately micromolar concentra1202

tions and to be poorly mitogenic or inactive by themselves16,30,31. Although the apparent affinity of the C19 synthetic peptide for FGFR in vitro is similar to those peptides, C19jun binds to FGFR with 40-fold higher affinity (Table 1) and is a fully effective FGFR agonist at subnanomolar concentrations on multiple FGF-responsive cell types (Fig. 2). The higher affinity of the C19jun dimer compared with the monomeric C19 peptide in the absence of heparin is presumably due to an avidity effect. This suggests that both peptides in the C19jun molecule interact with FGFRs simultaneously and might promote receptor dimerization. However, dimeric forms of the C19 sequence are not sufficient for activation of the FGFR in the absence of interaction with heparin (Fig. 5). Our results demonstrate that designed agonists for FGFR that are not structurally related to FGFs have the same requirement for interaction with heparin as do the endogenous FGFs. Several models proposed to explain the requirement of heparin in FGFR activation14,19,20,32 have been reconciled by the recent report of the crystal structure of basic FGF bound to the ECD of FGFR33. In this structure, two FGF:FGFR complexes form a twofold symmetric dimer. The two receptors are in contact with each other, and each molecule of FGF contacts both receptors. The two FGF molecules are on opposite sides of the dimer and do not interact with each other. In the absence of heparin, these interactions are presumably insufficient to stabilize the receptor dimer for activation. The authors suggest that heparin stabilizes the dimer by binding in a groove (the heparin binding “canyon”) spanning both FGF molecules and both receptors in the complex. This structure and the data presented here suggest a model for the interaction of C19jun with receptors that is analogous to that for FGF. The C19 peptides in C19jun are tethered to the jun leucine zipper by long and flexible linker segments. The receptor dimer therefore could be spanned by a single C19jun molecule contacting both receptors, as is the case with FGF. As described above, the data in Table 1 directly support this idea, and the data in Figure 4 suggest that the contact site(s) on the receptors are not precisely the same as for FGF. Because both bFGF and C19jun require interaction with heparin for activity (Fig. 5), the leucine zipper of C19jun must also be positioned in proximity to the heparin-binding “canyon,” where heparin can bind to both C19jun and to receptors and thereby stabilize the dimer for activation. Artificial FGFR agonists could be useful alternatives to FGFs in the treatment of ischemic vascular disease. For example, polypeptides that are specific for one of the many FGF receptor subtypes might be designed, thereby limiting undesired effects during systemic administration. Furthermore, recent results in our laboratory suggest that simpler molecules with equivalent activities can be designed using the strategies described in this report (data not shown). Our results have important implications for attempts to identify nonpeptidyl agonists of FGFRs. The requirement for our polypeptide agonist to bind specifically to both the receptor and to heparin suggests that orally available, low-molecular-weight comTable 2. Relative affinities of acidic FGF, basic FGF, and C19jun for FGFR isoforms1. Estimated IC50 of soluble FGFR (nM) Factor

FGFR1b:Fc

FGFR1c:Fc

FGFR2b:Fc

Acidic FGF Basic FGF C19jun

6 No inhibition No inhibition

45 100 100

0.7 >500 >500

1Mitogenesis assays of acidic FGF (1 nM), basic FGF (1 nM), and refolded C19jun protein (50 nM) were performed in the presence of soluble forms of the indicated FGFRs used as competitive binding antagonists, as described in Experimental Protocol and in Figure 3A. IC50 values were estimated from fourparameter curve fits. “>500” indicates that minor binding and inhibition were detected at the highest concentration of receptor tested (500 nM).

NATURE BIOTECHNOLOGY VOL 17 DECEMBER 1999

http://biotech.nature.com

© 1999 Nature America Inc. • http://biotech.nature.com

RESEARCH pounds may be difficult to develop, despite recent optimism that such an approach is feasible for other growth factor receptors10,13.

© 1999 Nature America Inc. • http://biotech.nature.com

Experimental protocol Phage display. A library of random 26-residue peptides, preceded by Ala-Glu and followed by six prolines, was displayed on the N terminus of pIII of M13 as described34. The random peptide was made using the codon definition of NNB, where N = ACGT and B = TCG. We incubated 1Ð2 X 1011 phage for 1 h at room temperature with Lentil Lectin Sepharose 4B beads (Sigma, St. Louis, MO) containing bound monomeric FGFR ECD protein 35 blocked in phosphate-buffered saline (PBS), 10% fetal calf serum, or with 5 X 10 6 Sf9 cells overexpressing full-length, transmembrane FGFR blocked in Grace media, and 2% nonfat milk. The beads or cells were then washed with PBS and eluted with 6 M urea, pH 2.2, and phage were neutralized with 2 M Tris base and amplified. After at least four rounds of selection, single plaques were sequenced and characterized34. Construction, expression, and purification of C19jun. C19jun was cloned into the pET23 vector (Fig. 1), expressed in BL21 DE3 pLysS Escherichia coli and purified by nickel-nitrilotriacetic acid metal-affinity chromatography as described by the kit manufacturer (Qiagen, Valencia, CA). Exponentially growing cultures were induced with 1 mM isopropylthiogalactoside for 4 h at 37oC and lysed in 6 M guanidine HCl, 5 mM imidazole, 0.1 M NaH2PO4, and 0.01 M Tris-HCl, pH 8.0. The protein was purified under denaturing conditions. Purified protein was refolded by dilution with five volumes of 0.5 mM reduced glutathione, 0.5 mM oxidized glutathione, 1 mM EDTA, 0.01M TrisHCl, pH 8.5, and 0.1M NaH2PO4; incubation for 24 h at 4oC; and subsequent extensive dialysis against PBS. Purified protein (>95%) was characterized by amino acid composition analysis, N-terminal sequencing, mass spectroscopy, and SDS–PAGE. The purified protein was approximately 80% dimeric. For some experiments (Fig. 2, 3B, and 4), the His 6 tag was replaced with a Glu-Glu antibody epitope and protein purified by immunoaffinity and anion exchange chromatography under native conditions without refolding, as described36. This material was approximately 10-fold more active than the refolded protein. Other constructs. Protein representing the linker segments and c-jun leucine zipper without the C19 sequence (Fig. 1) was expressed, purified, and characterized as for C19jun. C19 peptide with an intramolecular disulfide bond was synthesized, purified by HPLC, and characterized by mass spectroscopy. C19jun∆hep protein was prepared by changing human c-jun residues Arg276, Arg279, Lys285, and Lys288 in the leucine zipper domain to glutamines. Expression, purification, and characterization was as for C19jun. C19-Ig fusion protein was prepared by cloning the C19 sequence and GlyGly-Gly-Ser linker segments (Fig. 1) in frame with the Fc portion of human IgG1, expressing the protein in baculovirus and purifying protein with protein A affinity chromatography. The extracellular ligand-binding domain of human FGFR 1c (residues 1–377), FGFR 1c-dIg1dAB (FGFR 1c ECD with residues 31–147 deleted), FGFR 1b (residues 1–288), FGFR 2b (residues 1–376), and human erythropoietin receptor (residues 1–250) were fused to the Fc region of human IgG1, expressed in baculovirus and purified by protein A affinity chromatography. Activity of the fusion proteins was confirmed by binding of 125I-labeled ligands. Platelet-derived growth factor receptor Fc fusion protein was from R&D Systems (Minneapolis, MN). The approximate apparent affinities (Kd) of C19 synthetic peptide, jun leucine zipper, C19jun, C19 jun∆hep, and C19-Ig fusion proteins for immobilized dimeric FGFR 1c:Fc protein in the absence of heparin were determined by real-time kinetic analysis using the BIAcore biosensor (Pharmacia, Piscataway, NJ) and published methods37. Immunoprecipitations and immunoblotting. Swiss 3T3 fibroblasts or 293 cells expressing full-length FGFR 1c were incubated in quiescing media (DMEM, 1 µg/ml insulin, 5 µg/ml transferrin, 0.5 mg/ml bovine serum albumin) for 24–48 h and then stimulated with 1.0 nM bFGF or 50 nM refolded C19jun protein in the presence of 15 U/ml heparin for 15 min at 37 oC. The cells were lysed, and aliquots of lysate containing equal amounts of total protein were either analyzed directly by immunoblotting with phospho-MAP kinase-specific antibodies and anti-total MAP kinase antisera (New England Biolabs, Beverly, MA) or immunoprecipitated with anti-FGFR antisera (gift of L.T. Williams, Chiron Corporation, Emeryville, CA) and immunoblotted with antiphosphotyrosine antibodies (Upstate Biotechnology, Lake Placid, NY). c-fos mRNA expression. c-fos mRNA levels were assayed by bDNA assay in L6 myoblast cell lines stably expressing FGFR 1c and in control L6 cells that NATURE BIOTECHNOLOGY VOL 17 DECEMBER 1999

http://biotech.nature.com

do not express FGFR27 (gifts of L.T. Williams, Chiron Corp., Emeryville, CA) essentially as described38, except that the rat c-fos sequence was used to design the capture oligonucleotides. Mitogenic and proliferation assays. To assay cells for DNA synthesis, Swiss 3T3 fibroblasts (1 × 104 cells/well in 96-well microtiter plates) were incubated at 37oC in quiescing media for 24 h, stimulated with the indicated molecules for 18 h in the presence of 15 U/ml heparin, labeled with BrdU for 4–6 h, and assayed for BrdU incorporation by ELISA as described by the kit manufacturer (Boehringer Mannheim, Chicago, IL). To measure cell proliferation, Swiss 3T3 fibroblasts were seeded at 3 × 103 cells/well in 96-well microtiter plates, the indicated substances were added in the presence of 15 U/ml heparin, and the cultures were incubated for three to five days at 37oC. Proliferation was assessed by measurement of the extent of cleavage of the tetrazolium salt WST-1 by viable cells in a colorimetric assay as described by the kit manufacturer (Boehringer Mannheim). In some experiments, C19jun was preincubated with 0.01–500 nM of soluble forms of growth factor receptors for 15–30 min at 4oC before assay. Neurite outgrowth by PC12 cells. PC12 cells were seeded in 24-well plates coated with poly-L-lysine and mouse laminin (Collaborative Biomedical, Bedford, MA) at 1 × 104 cells/well in duplicate in RPMI media containing 5% fetal calf serum and 10% horse serum and allowed to attach overnight at 37oC (day 0). The indicated substances and 15 U/ml heparin were added on day 0 and day 2 without changing the media, and the cultures were examined for neurite extensions on day 4. 125I-bFGF binding assay. Microtiter wells were coated with 100 µl of 50 µg/ml streptavidin, blocked with PBS, 1.0% gelatin, and coated with 50 µl of 20 nM biotinylated, monomeric FGFR 1 ECD protein for 2 h at room temperature. The plate was then washed with PBS and 0.05% Tween 20, and 50 µl of 15 pM 125I-bFGF (NEN Life Science Products, Boston, MA; 119 µC/µg) was added per well in the presence of the indicated concentrations of unlabeled bFGF or C19jun in binding buffer (PBS, 0.1% gelatin, 0.1% Triton X100, 10 µM heparin) for 2.5 h. The wells were then washed five times and the wells counted in a gamma counter.

Acknowledgments The authors thank Jaime Escobedo, Anke Klippel, and Lewis T. Williams for helpful discussions and provision of reagents; Hermel Manalo and Hamid Khoja for expert technical assistance; and Leah Conroy, Gwynn Pardee, and Stephania Widger for baculovirus-expressed proteins. 1. Mason, I.J. The ins and outs of fibroblast growth factors. Cell 78, 547–552 (1994). 2. Wilkie, A.O., Morriss-Kay, G.M., Jones, E.Y. & Heath, J.K. Functions of fibroblast growth factors and their receptors. Curr. Biol. 5, 500–507 (1995). 3. Asahara, T. et al. Synergistic effect of vascular endothelial growth factor and basic fibroblast growth factor on angiogenesis in vivo. Circulation 1,II365–II371 (1995). 4. Baffour, R. et al. Enhanced angiogenesis and growth of collaterals by in vivo administration of recombinant basic fibroblast growth factor in a rabbit model of acute lower limb ischemia: dose-response effect of basic fibroblast growth factor. J. Vasc. Surg. 16, 181–191 (1992). 5. Harada, K. et al. Basic fibroblast growth factor improves myocardial function in chronically ischemic porcine hearts. J. Clin. Invest. 94, 623–630 (1994). 6. Landau, C., Jacobs, A.K. & Haudenschild, C.C. Intrapericardial basic fibroblast growth factor induces myocardial angiogenesis in a rabbit model of chronic ischemia. Am. Heart J. 129, 924–931 (1995). 7. Lazarous, D.F. et al. Effects of chronic systemic administration of basic fibroblast growth factor on collateral development in the canine heart. Circulation 91, 145–153 (1995). 8. Yanagisawa-Miwa, A. et al. Salvage of infarcted myocardium by angiogenic action of basic fibroblast growth factor. Science 257, 1401–1403 (1992). 9. Koketsu, N. et al. Pretreatment with intraventricular basic fibroblast growth factor decreases infarct size following focal cerebral ischemia in rats. Ann. Neurol. 35, 451–457 (1994). 10. Tian, S.S. et al. A small, nonpeptidyl mimic of granulocyte-colony-stimulating factor. Science 281, 257–259 (1998). 11. Cwirla, S.E. et al. Peptide agonist of the thrombopoietin receptor as potent as the natural cytokine. Science 276, 1696–1699 (1997). 12. Wrighton, N.C. et al. Small peptides as potent mimetics of the protein hormone erythropoietin. Science 273, 458–464 (1996). 13. Zhang, B. et al. Discovery of a small molecule insulin mimetic with antidiabetic activity in mice. Science 284, 974–977 (1999). 14. Waksman, G. & Herr, A.B. New insights into heparin-induced FGF oligomerization. Nat. Struct. Biol. 5, 527–530 (1998). 15. Bellot, F. et al. Ligand-induced transphosphorylation between different FGF receptors. EMBO J. 10, 2849–2854 (1991). 16. Baird, A., Schubert, D., Ling, N. & Guillemin, R. Receptor- and heparin-binding domains of basic fibroblast growth factor. Proc. Natl. Acad. Sci. USA 85, 2324–2328 (1988). 17. Gallagher, J.T. & Turnbull, J.E. Heparan sulphate in the binding and activation of

1203

© 1999 Nature America Inc. • http://biotech.nature.com

© 1999 Nature America Inc. • http://biotech.nature.com

RESEARCH basic fibroblast growth factor. Glycobiology 2, 523–528 (1992). 18. Moy, F.J. et al. Properly oriented heparin-decasaccharide-induced dimers are the biologically active form of basic fibroblast growth factor. Biochemistry 36, 4782–4791 (1997). 19. DiGabriele, A.D. et al. Structure of a heparin-linked biologically active dimer of fibroblast growth factor. Nature 393, 812–817 (1998). 20. Spivak-Kroizman, T. et al. Heparin-induced oligomerization of FGF molecules is responsible for FGF receptor dimerization, activation, and cell proliferation. Cell 79, 1015–1024 (1994). 21. Riley, L.G. et al. Cloning, expression, and spectroscopic studies of the Jun leucine zipper domain. Eur. J. Biochem. 219, 877–886 (1994). 22. Patel, L.R., Curran, T. & Kerppola, T.K. Energy transfer analysis of Fos-Jun dimerization and DNA binding. Proc. Natl. Acad. Sci. USA 91, 7360–7364 (1994). 23. Crameri, R. & Suter, M. Display of biologically active proteins on the surface of filamentous phages: a cDNA cloning system for the selection of functional gene products linked to the genetic information responsible for their production. Gene 160, 139 (1995). 24. de Kruif, J. & Logtenberg, T. Leucine zipper dimerized bivalent and bispecific scFv antibodies from a semi-synthetic antibody phage display library. J. Biol. Chem. 271, 7630–7634 (1996). 25. Claude, P., Parada, I.M., Gordon, K.A., D’Amore, P.A. & Wagner, J.A. Acidic fibroblast growth factor stimulates adrenal chromaffin cells to proliferate and to extend neurites, but is not a long-term survival factor. Neuron 1, 783–790 (1988). 26. Marshall, C.J. Specificity of receptor tyrosine kinase signaling: transient versus sustained extracellular signal-regulated kinase activation. Cell 80, 179–185 (1995). 27. Werner, S. et al. Differential splicing in the extracellular region of fibroblast growth factor receptor 1 generates receptor variants with different ligand- binding specificities. Mol. Cell. Biol. 12, 82–88 (1992). 28. Wang, F., Kan, M., Xu, J., Yan, G. & McKeehan, W.L. Ligand-specific structural

1204

29. 30.

31.

32.

33. 34. 35.

36.

37.

38.

domains in the fibroblast growth factor receptor. J. Biol. Chem. 270, 10222–10230 (1995). Hulme, E.C. & Birdsall, N.J.M. in Receptor-ligand interactions: a practical approach (ed. Hulme, E.C.) 101–104 (Oxford University Press, New York; 1993). Ray, J., Baird, A. & Gage, F.H. A 10-amino acid sequence of fibroblast growth factor 2 is sufficient for its mitogenic activity on neural progenitor cells. Proc. Natl. Acad. Sci. USA 94, 7047–7052 (1997). Yayon, A. et al. Isolation of peptides that inhibit binding of basic fibroblast growth factor to its receptor from a random phage-epitope library. Proc. Natl. Acad. Sci. USA 90, 10643–10647 (1993). Pantoliano, M.W. et al. Multivalent ligand-receptor binding interactions in the fibroblast growth factor system produce a cooperative growth factor and heparin mechanism for receptor dimerization. Biochemistry 33, 10229–10248 (1994). Plotnikov, A.N., Schlessinger, J., Hubbard, S.R. & Mohammadi, M. Structural basis for FGF receptor dimerization and activation. Cell 98, 641–650 (1999). Doyle, M.V. et al. in Combinatorial libraries: synthesis, screening and application potential (ed. Cortese, R.) 171 (Walter de Gruyter, New York; 1996). Kiefer, M.C. et al. Molecular cloning of a human basic fibroblast growth factor receptor cDNA and expression of a biologically active extracellular domain in a baculovirus system. Growth Factors 5, 115–127 (1991). Grussenmeyer, T., Scheidtmann, K.H., Hutchinson, M.A., Eckhart, W. & Walter, G. Complexes of polyoma virus medium T antigen and cellular proteins. Proc. Natl. Acad. Sci. USA 82, 7952–7954 (1985). Laminet, A.A., Apell, G., Conroy, L. & Kavanaugh, W.M. Affinity, specificity, and kinetics of the interaction of the SHC phosphotyrosine binding domain with asparagine-X-X-phosphotyrosine motifs of growth factor receptors. J. Biol. Chem. 271, 264–269 (1996). Shyamala, V. et al. High-throughput screening for ligand-induced c-fos mRNA expression by branched DNA assay in Chinese hamster ovary cells. Anal. Biochem. 266, 140–147 (1999).

NATURE BIOTECHNOLOGY VOL 17 JULY 1999

http://biotech.nature.com