Proliferation of neointimal smooth muscle cells after arterial injury. Dependence on interactions between fibroblast growth factor receptor-2 and fibroblast growth factor-9.

The growth factor signaling mechanisms responsible for neointimal smooth muscle cell (SMC) proliferation and accumulation, a characteristic feature of many vascular pathologies that can lead to restenosis after angioplasty, remain to be identified. Here, we examined the contribution of fibroblast growth factor receptors (FGFRs) 2 and 3 as well as novel fibroblast growth factors (FGFs) to such proliferation. Balloon catheter injury to the rat carotid artery stimulated the expression of two distinctly spliced FGFR-2 isoforms, differing only by the presence or absence of the acidic box, and two distinctly spliced FGFR-3 isoforms containing the acidic box and differing only by the presence of either the IIIb or IIIc exon. Post-injury arterial administration of recombinant adenoviruses expressing dominant negative mutant forms of these FGFRs were used to assess the roles of the endogenous FGFR isoforms in neointimal SMC proliferation. Dominant negative FGFR-2 containing the acidic box inhibited such proliferation by 40%, whereas the dominant negative FGFR-3 forms had little effect. Expression of FGF-9, known to be capable of binding to all four neointimal FGFR-2/-3 isoforms, was abundant within the neointima. FGF-9 markedly stimulated both the proliferation of neointimal SMCs and the activation of extracellular signal-related kinases 1/2, effects which were abrogated by the administration of antisense FGF-9 oligonucleotides to injured arteries and the expression of the dominant negative FGFR-2 adenovirus in cultured neointimal SMCs. These studies demonstrate that, although multiple FGFRs are induced in neointimal SMCs following arterial injury, specific interactions between distinctly spliced FGFR-2 isoforms and FGF-9 contribute to the proliferation of these SMCs.

Atherosclerosis, vasculopathies, post-transplant arteriosclerosis, and restenosis are characterized by expansion of the arterial intima as a result of the infiltration of mononuclear leukocytes, the accumulation of the extracellular matrix, and the proliferation of vascular smooth muscle cells (SMCs) 1 (1)(2)(3).
To date, the agents responsible for the proliferation of SMCs within the intima have eluded identification, and strategies aimed at inhibiting intimal SMC proliferation after arterial injury by either preventing platelet-derived growth factor expression or neutralizing fibroblast growth factor-2 (FGF-2) have been unsuccessful (4,5). This has led to postulates stating that phenotypically unique SMCs that can exhibit autonomous growth in the intima and secrete unknown heparin binding growth factors may be responsible for expansion of the SMC population in the intima (6,7). Findings that fibroblast growth factor receptors (FGFRs), (8) but not FGF-1 or FGF-2, are critical for vascular development implicate other members of the FGF superfamily in vessel development (9,10). Such FGFs are most likely also responsible for SMC proliferation within the intima of healing arteries following arterial injury during angioplasty and during the development of arteriosclerosis.
Immunohistochemical studies have identified the presence of multiple FGFR types in healing and atherosclerotic arteries, including FGFR-1, FGFR-2, and FGFR-3 (11)(12)(13). Although FGFR-1 has been shown to be involved in the medial SMC proliferation within healing arteries, it does not play a major role in neointimal SMC proliferation (11). Furthermore, the contribution of FGFR-2 and FGFR-3 to neointimal SMC proliferation has not been elucidated. The structural features of FGFRs include an extracellular ligand binding domain, a single transmembrane domain, and an intracellular tyrosine kinase domain. The extracellular region of FGFRs prototypically consists of three immunoglobulin-like domains, a short motif of acidic amino acids (the "acidic box") located just distal to Ig-like domain I, and a heparin binding domain also interposed between the first and second Ig-like domains; however, alternative RNA splicing can result in several isoforms of an FGFR from the one respective gene that differ in their extracellular sequence and possess unique ligand binding properties (14). One such splicing event results in a deletion of the exon encoding the amino-terminal Ig-like domain (domain I), resulting in a shorter, two-Ig-like domain isoform of the receptor (15); such shorter FGFRs ("␤-forms") frequently have higher affinities for some FGFs than do the corresponding longer three-Ig-like domain FGFRs ("␣-forms") and, in some instances, have been associated with tumor progression (16). Another splicing event that can dramatically alter the FGF binding ability of FGFR-1, -2, and -3 involves the alternate usage of either the "IIIb" or "IIIc" exon to encode the carboxyl-terminal portion of Ig-like domain III (14). Expression of these FGFR isoforms appears to be regulated in a tissue-specific manner, with IIIb exon-containing isoforms apparently restricted in expression to epithelial cell lineages and IIIc exon-containing isoforms restricted in expression to mesenchymal cell lineages (17). As a consequence of such splicing events, FGF ligands can have either a diverse or an extremely restricted FGFR binding profile. For example, FGF-1 binds to all FGFR isoforms with high affinity, FGF-2 exhibits a high affinity for binding to FGFR-1IIIb, FGFR-1IIIc, and FGFR-3IIIc, and FGF-9 exhibits a high affinity for binding to FGFR-2IIIc, FGFR-3IIIb, and FGFR-3IIIc (17); conversely, FGF-7 and FGF-10 are restricted in binding, with high affinity only to FGFR-2IIIb (14,17,18). Finally, it has become increasingly apparent that the cellular potency of FGFR signaling is also influenced by post-translational receptor modification, in particular the enhanced mitogenic capacity of FGFR-2␤ isoforms arising from glycosaminoglycan modification at the acidic box motif (19).
Here, we demonstrate that arterial injury induces the predominant expression by neointimal SMCs of two distinctly spliced FGFR-2IIIc ␤-isoforms, differing from each other only by the presence/absence of the acidic box motif, and two distinctly spliced FGFR-3 ␣-isoforms, both containing the acidic box motif and differing from each other only by the alternate IIIb/IIIc composition of Ig-like domain III. Using adenoviruses expressing dominant negative mutants of specific receptor types truncated at their transmembrane domain and devoid of tyrosine kinase activity, we demonstrate that signaling through FGFR-2IIIc is a major contributor to neointimal SMC proliferation. We further demonstrate that a specific FGF ligand, FGF-9, known to have high affinity for the neointimal FGFR-2/-3 isoforms, shows an elevated neointimal expression after arterial injury and contributes to neointimal SMC proliferation through an interaction with FGFR-2IIIc. These findings, for the first time, demonstrate a major regulatory role for an FGFR2-IIIc/FGF-9 interaction in the stimulation of neointimal SMC proliferation in the injured arterial wall.

EXPERIMENTAL PROCEDURES
Carotid Artery Injury-Balloon catheter injury to the rat carotid artery was carried out as described previously (4,20). Briefly, after anesthetizing the rats with a mixture of pentobarbitone (30 mg/kg), methohexitone (40 mg/kg), and atropine sulfate (3 mg/kg) administered by intraperitoneal injection, a 2F Fogarty arterial embolectomy catheter (Baxter) was passed through an arteriotomy in the left common femoral artery and into the left common carotid artery to its bifurcation. The balloon was inflated with 25 l of saline and withdrawn with a rotating action to the aortic arch and then reintroduced and withdrawn a further two times. The femoral artery was then ligated, the incision was closed, and the animals were allowed to recover in a humidified, warmed chamber for 1-2 h. Subsequently, the animals either underwent a second surgical procedure (see below) or were sacrificed 2 to 14 days later with an overdose of pentobarbitone (100 mg/kg intraperitoneal), and the carotid arteries were collected for mRNA analyses, Western blotting, or immunohistochemistry.
Delivery of Adenoviruses/Antisense Oligonucleotides to Injured Arteries and Determination of SMC Proliferation-After anesthetizing the rats (see above), a midline neck incision was made, and blunt end dissection to the carotid bifurcation was performed. A fine catheter was then inserted through an arteriotomy in the external carotid artery and placed into the common carotid artery. Temporary ligatures were then placed around the catheter in the common carotid artery before flushing the vessel with a solution of adenovirus or oligonucleotide and then placing a second ligature on the vessel toward the aortic arch. Solutions of either adenovirus (5 ϫ 10 9 plaque-forming units in 100 l saline) or oligonucleotides (20 M in 100 l of LipofectAMINE/Opti-MEM) were then infused into the ligated segment of the common carotid artery for 30 min at a pressure of 80 -100 mm Hg. The ligatures and catheter were then removed, the external carotid ligated, the incision closed, and the animals allowed to recover (see above). To measure cell proliferation in the healing vessels, 50 mg/kg bromodeoxyuridine (BrdUrd) was administered by intraperitoneal injection 24 and 8 h before sacrifice.
Tissue Collection and Immunohistochemistry-Carotid arteries were removed from the sacrificed animals and cleaned of non-vascular tissue. Then, arteries intended for mRNA analyses or Western blotting underwent careful removal of their neointima under a dissecting microscope, and the neointimal tissue was then rapidly frozen in liquid nitrogen. Carotid arteries for immunohistochemical analyses were mounted in OCT compound (Tissue Tek; Miles Scientific), frozen in pentane over liquid nitrogen, and stored at Ϫ70°C. Immunohistochemical detection of FGFR and FGF peptide expression and BrdUrd incorporation was carried out using 6 m of frozen cross-sections and the avidin-biotinperoxidase complex kit (Vector Laboratories) with 3,3Ј-diaminobenzidine tetrachloride as the chromogenic substrate, essentially as described previously (21). Primary antibodies to FGFR-2 (1:500, rabbit anti-human polyclonal; catalog number sc-122), FGFR-3 (1:500, rabbit anti-human polyclonal; catalog number sc-123), and FGF-9 (1:500, goat anti-human polyclonal; catalog number sc-1368) were from Santa Cruz Biotechnology Inc., the anti-FGF-1 antibody (1:500, rabbit antibovine polyclonal; catalog number G5081) was from Promega, the anti-FGF-2 antibody (1:500, mouse anti-human monoclonal; catalog number DE6) was from du Pont du Nemours and Company, and the anti-BrdUrd antibody (1:100, mouse monoclonal; catalog number 03-3900) was from Zymed Laboratories Inc.. The appropriate peroxidase-labeled secondary antibodies (Vector Laboratories) were used at a concentration of 1:200. After completing the immunohistochemical procedures, the sections were lightly counterstained with hematoxylin.
Oligonucleotide Primers for RT-PCR and Antisense FGF-9 Oligonucleotides-FGFR-2 and FGFR-3 primers were designed to provide definitive structural information on distinct regions of the receptors encoded by each amplified cDNA (see Fig. 2A). Specifically, to detect the presence/absence of Ig-like domain I and the acidic box motif in the extracellular region of each receptor, primer sets designated as "Region A" were utilized in which the sense and antisense primers were chosen such that they were sufficiently distal from the 5Ј and 3Ј splicing boundaries of both consecutive exons that code for Ig-like domain I and the acidic box; for detection of alternatively spliced IIIb or IIIc exons in the carboxyl terminus of Ig-like domain III and in the presence of the transmembrane-spanning region, primer sets designated as "Region B-TM" in the case of FGFR-2 and "Region A-IIIb", "Region A-IIIc", or Region B-TM in the case of FGFR-3 were utilized. The precise location of each of the primer pairs within each cDNA and the predicted size of their amplification products were as follows. For FGFR-2 cDNA (Gen-Bank TM accession numbers Z35138, L19104, and D83495), in Region A the sense primer comprised nucleotides 19 -42 and the antisense primer comprised nucleotides 585-614 (596 bp); in Region B-TM the sense primer comprised nucleotides 884 -913 and the antisense primer comprised nucleotides 1426 -1445 (562 bp). To detect the presence in this cDNA fragment of the IIIb or IIIc exon, the restriction endonuclease HpaI, a site for which is present in the IIIc exon but not the IIIb exon, was used to digest the amplified product. For FGFR-3 cDNA (GenBank TM accession numbers M81342 and L26492), in Region A the sense primer comprised nucleotides 6 -33 and the antisense primer comprised nucleotides 433-452 (447 bp). To detect either the IIIb exon or IIIc exon in the carboxyl-terminal half of Ig-like domain III, the sense primer of Region A was used with a IIIb exon-specific antisense primer (Region A-IIIb, nucleotides 914 -939; 934 bp) or a IIIc exon-specific antisense primer (Region A-IIIc, nucleotides 956 -981; 976 bp). For the transmembrane-spanning domain (Region B-TM), the sense primer comprised nucleotides 913-942 and the antisense primer comprised nucleotides 1676 -1696 (784 bp). To detect FGF-9 cDNA (GenBank TM accession number D14839), the sense primer comprised nucleotides 222-247 and the antisense primer comprised nucleotides 508 -537 (316 bp). Phosphorothioate-modified FGF-9 antisense and sense oligonucleotides were made, corresponding to nucleotides 170 -184 of the rat cDNA.
In Situ RT-PCR-To detect FGF-9 mRNA in the neointima, 6-m cross-sections of carotid arteries fixed onto gelatinized slides with formalin were subjected to in situ RT-PCR as described previously (23). Briefly, FGF-9 mRNA was detected by a "one-step" RT-PCR method (35 cycles) utilizing EZ rTth polymerase (PerkinElmer Life Sciences) according to the manufacturer's protocol and digoxigenin-11-dUTP. Messenger RNA expression/localization was then determined with an antidigoxigenin antibody coupled to alkaline phosphatase; color detection was carried out using the substrate nitro blue tetrazolium. Controls included tissue sections not exposed to PCR primers and reaction mixtures not subjected to reverse transcription.
Recombinant Dominant Negative FGFR-2 and FGFR-3 Adenoviruses-The pAd-Easy1 virus and shuttle plasmid pAd-TrackCMV were obtained from Dr. R. Hannan, Peter MacCallum Cancer Institute, Melbourne, Australia (24). Adenoviruses expressing dominant negative (DN) FGFR-2␤IIIcAB (devoid of Ig-like domain I and containing the acidic box), FGFR-3␣IIIbAB, and FGFR-3␣IIIcAB (both containing all three Ig-like domains and the acidic box and differing from each other only in the alternate presence of either the IIIb or IIIc exon in the carboxyl terminus of Ig-like domain III) were prepared by amplifying their cDNAs from rat SMC RNA and truncating these cDNAs at the transmembrane domain with the inclusion of a FLAG epitope tag immediately distal to the last amino acid of the transmembrane domain. These cDNAs were cloned into pAd-TrackCMV using the following PCR primers incorporating KpnI and XbaI restriction sites: FGFR-2␤IIIcAB, 5Ј-taccggtaccATGGTCAGCTGGGGGCGCTTCATC-3Ј (sense; nucleotides 1-24 of cDNA; KpnI site is lowercase and underlined and the initiation codon is bold and italicized) and 5Ј-ccggtctagaCTACTTGTCAT-CGTCGTCCTTGTAGTCGCTGCTGAAGTCTGGCTTCTTGGT-3Ј (antisense; nucleotides 1207-1236 of cDNA; XbaI site is lowercase and underlined, termination codon is bold and italicized, and the FLAG epitope is in small capitals); and FGFR-3␣IIIbAB or FGFR-3␣IIIcAB, 5Ј-cacggtacc-ATGGTAGTCCCGGCCTGCGTGCTA-3Ј (sense; nucleotides 1-24 of cDNA; KpnI site is lowercase and underlined and the initiation codon is bold and italicized) and 5Ј-ccggtctagaCTACTTGTCATCGTCGTCCTTGTAGTCG-CCCAGGCCCTTCTTTGGGGGACT-3Ј (antisense; nucleotides 1180 -1209 of cDNA; the XbaI site is lowercase and underlined, the termination codon is bold and italicized, and the FLAG epitope is in small capitals). The integrity of each construct was confirmed by automated DNA sequencing, and the differentiation between either the FGFR-3␣IIIbAB or the FGFR-3␣IIIcAB constructs was confirmed by the use of restriction enzymes specific for either the IIIb or IIIc exon. Recombinant adenoviruses were generated by bacterial homologous recombination between pAd-TrackCMV containing DN-FGFR cDNAs and pAdEasy-1. Large scale amplification and purification using human embryonic kidney 293 cells was then carried out as described previously (25).
Neointimal SMC Culture and Proliferation Assay-Neointimal SMCs were isolated by elastase/collagenase digestion of 2-3 neointimas obtained from healing carotid arteries and cultured as described previously (21). These SMCs exhibited the "cobblestone" morphology described previously for such SMCs (26). Proliferation stimulated by FGF-9 was determined on semiconfluent neointimal SMC cultures that had been serum-deprived for 48 h prior to the addition of the growth factor (50 ng/ml) and then for a further 18 h with the subsequent measurement of cellular [ 3 H]thymidine incorporation, essentially as described previously (27). For such experiments, SMCs at passages 1-3 were used.
Western Blot Analyses-Protein was extracted from the neointima of injured arteries and subjected to Western blot analysis using either the anti-FGF-9 antibody (see above), an anti-FLAG epitope antibody (mouse monoclonal, Sigma catalog number F-3165), or an antibody to green fluorescent protein (GFP; mouse anti-jellyfish monoclonal, Roche Applied Science catalog number 1814460) and the enhanced chemiluminescence Western blotting system (Amersham Biosciences) essentially as described previously (4). To measure DN-FGFR adenovirus expression in neointimal SMCs infected 48 h earlier, the cells were washed in ice-cold phosphate buffered saline and lysed using a solution consisting of 1% Nonidet P-40, 10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 2 mM sodium orthovanadate, 100 mM sodium fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, and 100 M phenylmethylsulfonyl fluoride. The solution was clarified by centrifugation at 10,000 ϫ g for 10 min, and the protein was quantified using the Coomassie Plus protein assay kit (Pierce) using bovine serum albumin as standard. SDS-electrophoresis was performed using 10% acrylamide gels followed by transfer to polyvinylidene difluoride membranes (Millipore), and confirmation of mature DN-FGFR peptide expression was carried out by using the anti-FLAG epitope antibody. The ability of FGF-9 to stimulate the phosphorylation of p42/44 mitogen-activated protein ki-nases (ERK 1/2) in the neointimal SMCs, either with or without prior DN-FGFR-2 infection, was assessed with Western blot analysis using a mouse anti-human monoclonal phospho-p42/44 mitogen-activated protein kinase (Thr-202/Tyr-204) antibody (Cell Signaling Technology catalog number 9106); in these SMCs total ERK levels were assessed using a rabbit anti-human polyclonal antibody to ERK 1/2 (Santa Cruz Biotechnology Inc. catalog number sc-93), and levels of DN-FGFR-2␤II-IcAB were assessed using the anti-FLAG antibody.
Statistical Analyses-Results are expressed as mean Ϯ S.E. The inhibitory effect of the DN-FGFR adenoviruses and the antisense-FGF-9 oligonucleotide on neointimal SMC proliferation in injured arteries was assessed using unpaired t tests. Significance was established by a value of p Ͻ 0.05.

FGFR Expression in Proliferating SMCs of the Neointima-
Because FGFRs bind FGF family members with varying affinities, and alternative hnRNA splicing leads to receptor isoforms that possess unique FGF binding properties (14,17), we initially investigated the expression of two FGFRs, FGFR-2 and FGFR-3, in the rapidly proliferating SMCs within the growing neointima of healing rat carotid arteries injured with an inflated balloon catheter. Ten days after injury, high concentrations of FGFR-2 and FGFR-3 were expressed in the neointima and the media of these arteries (Fig. 1, A and B). Four days later, FGFR-2 levels were greatly reduced (Fig. 1C), but FGFR-3 continued to be expressed (Fig. 1D). FGFR-2 and FGFR-3 were also expressed in the media of these vessels as early as 2 days after injury (not shown), but not in uninjured carotid arteries (Fig. 1, E and F). To gain an insight into the potential role of these receptors in neointimal SMC proliferation and the FGF ligands with which they might interact, we determined the expression of distinctly spliced FGFR-2 and FGFR-3 isoforms by RT-PCR analysis of their mRNAs isolated from the neointima, followed by nucleotide sequencing of specific amplified fragments. Analysis of FGFR-2 mRNA was performed using oligonucleotide primers (depicted in Fig. 2A) encompassing the extracellular region that includes Ig-like domains I and II (Region A, nucleotides 19 -614 of the rat cDNA) and the region that includes Ig-like domain III and the transmembrane-spanning domain (Region B-TM, nucleotides 884 -1445 of the rat cDNA). Amplification using Region A primers indicated that only trace amounts of full-length FGFR-2 containing all three Ig-like domains was expressed, as shown by the low abundance of the expected PCR product (596 bp; Fig.   FIG. 1. FGFR-2 and FGFR-3

FGFR-2:FGF-9 Interactions Regulate Intimal SMC Proliferation
2B, lane 1); an additional two minor FGFR-2 cDNA species running below the full-length fragment were determined by diagnostic restriction endonuclease digestion (data not shown) to arise from aberrant splicing events within the Ig-like domain I exon. The two predominant amplification products were in the size range 250 -350 bp and were characterized further by nucleotide sequencing, which confirmed that the larger fragment was 332 bp in length and was the consequence of splicing events that resulted in the deletion of the 264-bp exon encoding Ig-like domain I; the smaller fragment was 254 bp in length and had arisen from the deletion of both the Ig-like domain I exon and the subsequent 78-bp exon encoding the acidic box motif. Amplification using Region B-TM primers yielded the expected product (562 bp), which was completely cleaved with HpaI (Fig. 2B, lane 3), a restriction endonuclease specific for the IIIc exon, therefore indicating the presence of only the IIIc exon in this cDNA fragment. Thus, SMCs within the growing neointima of healing arteries express predominantly two FGFR-2 isoforms that lack the Ig-like domain I ("FGFR-2␤" forms), both of which contain exon IIIc and differ only by the presence or absence of the acidic box motif (Fig. 2C). Such isoforms, designated "FGFR-2␤IIIcAB" and "FGFR-2␤IIIc," respectively, have been shown in a number of cell types to interact strongly with FGF-1, FGF-4, and FGF-9, but not with FGFs such as FGF-7 (14,17,18).
Given these findings, it was therefore of interest to also delineate the FGFR-3 isoforms expressed by the neointimal SMCs by using the PCR primers depicted in Fig. 2A. Amplification with Region A primers, which encompass the extracellular region that includes Ig-like domains I and II (nucleotides 6 -452 of rat cDNA), yielded the predicted sized product of 447 bp (Fig. 2D,  lane 1), indicating the presence of the exon encoding for Ig-like domain I. Additional amplification of the extracellular region by using the sense primer of Region A in combination with an antisense primer specific for either the IIIb exon (Region A-IIIb, nucleotides 6 -939 of rat cDNA) or the IIIc exon (Region A-IIIc, nucleotides 6 -981 of rat cDNA) yielded the predicted sized fragments of 934 bp (Fig. 2D, lane 2) and 976 bp (Fig. 2D, lane 3), respectively, indicating the simultaneous expression of FGFR-3 isoforms that contain either the IIIb or IIIc exon in the carboxyl terminus of the third Ig-like domain. Confirmation that such isoforms were transmembrane-spanning was shown by the amplification with primers encompassing a 784-bp region (Region B-TM, nucleotides 913-1696 of rat cDNA) that surrounds the transmembrane domain, yielding the predicted sized product (Fig. 2D, lane 4). Thus, SMCs in the growing neointima, in addition to the two FGFR-2 isoforms, express two FGFR-3 isoforms that contain all three Ig-like domains and the acidic box "FGFR-3␣" forms) and that differ from each other only by the alternate presence of either the IIIb or the IIIc exon and are designated as either FGFR-3␣IIIbAB or FGFR-3␣IIIbAB, respectively (Fig. 2E). Previously, it has been demonstrated in transfection studies that expression of either FGFR-3 isoform into L6 or BaF3 cells markedly increases their mitogenic responses to FGFs, including FGF-1 and FGF-9 (17,28). Therefore, the FGFR-2/FGFR-3 isoforms expressed in the SMCs of the growing neointima in the injured carotid artery possess the ability to contribute to neointimal SMC proliferation, depending on the co-expression profile of the FGF ligands with which they are known to interact.
Dominant Negative Mutant FGFRs and the Inhibition of Neointimal SMC Proliferation-To determine whether one or more of these FGFRs contributed to neointimal SMC proliferation, we next examined the extent to which dominant negative mutant forms of the receptors could inhibit such proliferation. Previously, overexpression of dominant negative mutant

FGFR-2:FGF-9 Interactions Regulate Intimal SMC Proliferation
FGFRs that contain the extracellular domains but are devoid of the intracellular tyrosine kinase domain has been used to differentiate the roles of endogenous FGFR-1, FGFR-2, and FGFR-3 in lens fiber differentiation and the development of diabetes (29,30). Initially, we infected cultured neointimal SMCs with recombinant adenoviruses expressing cDNAs encoding DN-mutant forms of FGFR-2␤IIIcAB, FGFR-3␣IIIbAB, and FGFR-3␣IIIcAB to confirm their expression as transmembrane-anchored proteins of the appropriate size. Two days after infection with the DN-FGFR-2␤IIIcAB adenovirus (Fig. 3A,  lane 2), SMCs expressed a minor peptide band of 37 kDa, which is the predicted size of the receptor protein in the absence of any significant glycosaminoglycan modification, and an intense broad band of ϳ55-80 kDa, which is consistent with this receptor isoform undergoing extensive post-translational modification by glycosaminoglycans (19). In contrast, infection of the SMCs with either the DN-FGFR-3␣IIIbAB adenovirus (Fig. 3A,  lane 4) or the DN-FGFR-3␣IIIcAB adenovirus (Fig. 3A, lane 6) resulted in the predominant expression of the predicted 44-kDa protein with very little post-translational modification evident, which is consistent with findings that the presence of Ig-like domain I in these FGFR-3 isoforms largely abrogates any glycosaminoglycan modification occurring near the acidic box (19). To assess the ability of the DN-FGFR adenoviruses to inhibit neointimal SMC proliferation, carotid arteries injured 7 days earlier were exposed to each adenovirus (100 l of a 5 ϫ 10 9 -plaque-forming unit solution), and 72 h later the effects on DNA synthesis were assessed by immunohistochemical detection and quantification of BrdUrd incorporation (percentage of BrdUrd-positive neointimal SMCs). In such experiments, the DN-FGFR-2␤IIIcAB adenovirus was able to inhibit neointimal SMC proliferation by 40% relative to the level of proliferation present after infection with the control, a GFP-expressing adenovirus (p Ͻ 0.01; Fig. 3, B and C). Thus, as in NIH 3T3 cells (19), the endogenously expressed FGFR-2␤IIIcAB isoform appears to be a potent stimulator of neointimal SMC proliferation. In contrast, infection of the vessels with an adenovirus expressing DN-FGFR-3␣IIIcAB did not affect neointimal SMC proliferation (Fig. 3B) and, although infection with an adenovirus expressing DN-FGFR-3␣IIIbAB reduced neointimal SMC proliferation by ϳ10%, this effect was not statistically significant (p Ͼ 0.05; Fig. 3B). Using Western blotting, we also assessed the level of DN-FGFR proteins and GFP (the expression of which is driven independently by a cytomegalovirus promoter in all four adenoviruses) (25) and confirmed that equivalent adenoviral expression occurred within arteries in each of the four treatment groups (Fig. 3B, insets).

FIG. 3. Effect of inhibition of FGFR signaling on neointimal SMC proliferation.
A, to confirm both the correct processing of the DN-FGFR constructs and the extent to which they undergo post-translational modification, cultured neointimal SMCs were infected with either DN-FGFR-2␤IIIcAB (R-2IIIc), DN-FGFR-3␣IIIbAB (R-3IIIb), or DN-FGFR-3␣IIIcAB (R-3IIIc), and protein isolated 48 h later was analyzed by Western blotting using an anti-FLAG antibody. The absence (Ϫ) or presence (ϩ) of the DN-FGFR adenovirus is noted; arrows indicate predicted sized protein species; bracket on the left denotes the occurrence of post-translationally modified 55-80 kDa species of FGFR-2␤IIIcAB. B, 7 days after the initial balloon catheter injury, local delivery to the carotid artery of 5 ϫ 10 9 plaque-forming units (in 100 l) of each of the three DN-FGFR adenoviruses and a control, a GFPexpressing adenovirus (GFP), was carried out (as described under "Experimental Procedures") with immunohistochemical detection and quantification of BrdUrd incorporation measured 72 h subsequently. Data are presented as the percentage of neointimal SMCs that have incorporated BrdUrd into their nuclei, which was significantly lower in the DN-FGFR-2␤IIIcAB treatment (asterisk) relative to the GFP control and represent the mean Ϯ S.E. from five animals in each treatment group. At this time, the levels of DN-FGFR protein and GFP within such arteries were also determined by Western blotting using either anti-FLAG or anti-GFP antibody (see insets), confirming equivalent expression of each adenoviral construct within the four treatment groups. C, representative cross-sections of injured arteries indicating the lower number of proliferating neointimal SMCs (indicated by red arrowheads) after DN-FGFR-2␤IIIcAB adenoviral treatment (panel a) relative to the control, GFP adenoviral treatment (panel b). ni, neointima; m, media; upward arrows indicate internal elastic laminae. FIG. 4. FGF-1 and FGF-9 expression in injured rat carotid artery. A and B, immunohistochemical detection of FGF-1 and FGF-9 peptides, respectively, in carotid artery 10 days after the balloon catheter injury. C, immunohistochemical analysis demonstrating barely detectable FGF-9 expression in uninjured carotid arteries. D, neointimal expression of FGF-9 peptides was confirmed by using both RT-PCR (section a), with FGF-9-specific primers resulting in the amplification of the predicted sized 316-bp fragment (S represents X174 digested with HaeIII as size standards), and Western blot analysis (section b) detecting the 26-kDa FGF-9 peptide. ni, neointima; m, media; a, adventitia; arrows indicate the internal and external elastic laminae.

Expression of Fibroblast Growth Factors during Neointimal
SMC Proliferation-Because FGFR-2␤IIIcAB and FGFR-3␣II-IbAB are known to exhibit a high affinity for FGF-1 and FGF-9 with a resultant stimulation of cellular proliferation (17, 28), we next examined the potential for these two FGF ligands to contribute to the neointimal SMC proliferation in healing arteries. First, we examined by immunohistochemistry the expression of FGF-1, which does not possess a hydrophobic signal sequence and is not normally secreted by cells (31). Ten days after the injury, when up to half of the SMCs in the neointima are proliferating, FGF-1 is barely detectable in this region of the healing artery, contrasting with its high concentration in the relatively quiescent medial region (Fig. 4A). In contrast, FGF-9 is highly expressed in the neointima at this time after injury, being diffusely dispersed throughout this region of the vessel (Fig. 4B) but barely detectable in uninjured arteries (Fig.  4C). Neointimal FGF-9 expression was confirmed both by RT-PCR (Fig. 4D, section a) and Western blot analyses (Fig. 4D,  section b) of RNA and protein extracted from neointimal tissue obtained 10 days after injury. To determine the relationship between neointimal SMC proliferation and FGF-9 expression in the healing arteries, we carried out in situ RT-PCR for FGF-9 mRNA in arterial cross-sections collected 14 days after the injury. This analysis showed FGF-9 mRNA expression to be clearly apparent in SMCs adjacent to the lumen of the vessel ( Fig. 5A; specificity of amplification/staining is shown by comparison to an equivalent cross-section not subjected to reverse transcription (Fig. 5B)), where BrdUrd incorporation indicated that SMCs were still proliferating at elevated levels (Fig. 5C). In addition, FGF-9 expression was seen in ϳ30 -40% of SMCs deep within the neointima and also adjacent to the vessel's media (Fig. 5A), regions in which BrdUrd incorporation indicated a lower number of proliferating SMCs (Fig. 5C).
FGF-9 and Neointimal SMC Proliferation-We then evaluated more directly the potential of FGF-9 to act as a proliferative growth factor for these neointimal SMCs. Initially, we examined its ability to stimulate mitogenesis in quiescent cultured SMCs prepared from the neointimas of carotid arteries injured 10 days earlier. Exposing the SMCs to FGF-9 resulted in an 8-fold elevation in [ 3 H]thymidine incorporation levels (Fig. 5D); no increase in [ 3 H]thymidine incorporation levels was observed after exposing these SMCs to FGF-7, which is consistent with our finding that the high affinity FGFR for this ligand, FGFR-2IIIb, was not appreciably expressed by neointi- FIG. 5. Functional relationship between FGF-9 expression, neointimal SMC proliferation, and FGFR-2␤II-IcAB signaling. A, in situ RT-PCR localization of neointimal FGF-9 mRNA expression demonstrating high concentrations associated with lumenal SMCs (some of which are indicated by a red arrowhead), although expression is still associated with the SMCs deeper within the neointima (ni) and in the media (m) (in both instances some are indicated by a yellow arrowhead). Upward arrows indicate the internal elastic lamina. B, an equivalent cross-section, as shown in panel A, which, as a control for amplification/staining specificity, was not subjected to reverse transcription. C, BrdUrd-labeled neointimal SMCs at this time (darkly stained and indicated by a red arrowhead), primarily localizing close to the lumen of the vessel (section counterstained with hematoxylin; upward arrows indicate the internal elastic lamina) a, adventitia. D, effect of exogenous FGF-9 (50 ng/ml) on [ 3 H]thymidine incorporation in SMCs cultured from neointimas obtained 10 days after the balloon catheter injury. E, Western blot analysis demonstrating the kinetics of activation of ERK 1/2 (p-ERK row) in these SMCs exposed to FGF-9 over a 16 h time period (ERK row shows equivalent protein loading in each lane). C, no FGF-9 exposure. F, p-ERK row, left, ability of the DN-FGFR-2␤IIIcAB adenovirus, 48 h after the infection of these SMCs, to block ERK 1/2 activation in response to a 10-min exposure to 50 ng/ml FGF-9 (F-9) but not 50 ng/ml platelet-derived growth factor-BB (BB); P-ERK row, right, identical treatment in the absence of the adenovirus. The ERK row shows equivalent protein loading in each lane, whereas the FLAG row, left, shows equivalent expression of the DN-FGFR-2␤IIIcAB adenovirus in the infected SMCs. C, no FGF-9 or plateletderived growth factor-BB exposure.
FGFR-2:FGF-9 Interactions Regulate Intimal SMC Proliferation mal SMCs (see Fig. 2, B and C). Activation of the ERK 1/2 pathway has been shown to play a major role in neointimal SMC proliferation following arterial injury (32), so we next examined the ability of FGF-9 to activate these signaling proteins. In such experiments, FGF-9 induced a rapid, marked elevation in ERK 1/2 phosphorylation, which remained somewhat increased for at least 16 h (Fig. 5E, p-ERK row), suggesting an involvement of FGFR-2␤IIIcAB (19); additional probing for levels of total ERK (Fig. 5E, ERK row) confirmed equivalent protein loading of all samples. To confirm the involvement of this receptor, the SMCs were exposed to an adenovirus expressing the DN-FGFR-2␤IIIcAB for 48 h before the addition of FGF-9 and an assessment of ERK 1/2 activation. In such experiments, FGF-9 was unable to elicit ERK 1/2 activation after 10 min of exposure to the SMCs (Fig. 5F, p-ERK row, left) compared with the normal activation in the SMCs not preexposed to the DN-FGFR-2␤IIIcAB adenovirus (Fig. 5F, p-ERK row, right). Under both sets of experimental conditions the SMCs remained equally responsive to platelet-derived growth factor-BB, which elevated ERK 1/2 phosphorylation to a similar extent in the presence (Fig. 5F, left section of row labeled p-ERK) or absence (Fig. 5F, p-ERK row, right) of DN-FGFR-2␤IIIcAB; additional probing for levels of total ERK (ERK row) and DN-FGFR-2␤IIIcAB (FLAG row) confirmed, respectively, equivalent protein loading of all samples and equivalent expression of the DN-FGFR-2␤IIIcAB adenovirus within the infected SMCs. Taken together, these findings provide further support for the interaction between FGF-9 and endogenous FGFR-2␤IIIcAB to play a major role in the proliferation of neointimal SMCs subsequent to arterial injury.
Because FGF-9 was mitogenic for neointimal SMCs in vitro, we next determined its contribution to peak neointimal SMC proliferation in the injured arteries. Eight days subsequent to the initial balloon catheter injury, we delivered locally a 20 M solution of a phosphorothioate-modified antisense FGF-9 oligonucleotide, which was designed toward the 5Ј-end of the mRNA (bases 170 -184 of the rat cDNA) to most effectively disrupt the translation of the FGF-9 protein, and assessed neointimal SMC proliferation 48 h later. In such experiments, the antisense oligonucleotide inhibited neointimal SMC proliferation by ϳ52% (Fig. 6A), whereas identical administration of the corresponding phosphorothioate-modified sense oligonucleotide did not inhibit neointimal SMC proliferation (Fig. 6A). Immuno-histochemical assessment of FGF-9 protein levels in arteries administered either the sense oligonucleotide (Fig. 6B) or the antisense oligonucleotide (Fig. 6C) demonstrated a marked reduction in the amount of FGF-9 protein within the neointima of the antisense oligonucleotide-treated arteries, providing further support for a major role of this FGF in neointimal SMC proliferation.

DISCUSSION
The identity of the growth factor/growth factor receptor signaling systems responsible for the proliferation of SMCs within the arterial intima/neointima has remained largely unknown. An involvement of FGF:FGFR signaling in intimal/neointimal SMC proliferation has been suggested, based on studies demonstrating the presence of FGFR-1 in healing arteries (11) and the expression of FGFR-1, FGFR-2, and FGFR-3 in atherosclerotic arteries (12,13). However, although FGFR-1 has been shown to be involved in the early medial SMC proliferation stimulated by arterial injury, it appears not to be involved in neointimal SMC proliferation (11). To date, no study has determined the role that FGFR-2 and FGFR-3 play in this process or identified the FGF ligands responsible for neointimal SMC proliferation.
Here, we critically tested the contribution of FGFR-2, FGFR-3, and FGF-9 to neointimal SMC proliferation. We found that balloon catheter injury to the rat carotid artery stimulated the expression of two distinctly spliced FGFR-2␤IIIc isoforms, differing only by the presence/absence of the acidic box, as well as two distinctly spliced FGFR-3␣ isoforms containing the acidic box and differing from each other by the presence of either the IIIb or the IIIc exon. Using post-injury arterial administration of recombinant adenoviruses expressing dominant negative mutant forms of these FGFRs to assess the roles of the endogenous FGFR isoforms in neointimal SMC proliferation, we found that DN-FGFR-2␤IIIcAB significantly inhibited such proliferation, whereas the DN-FGFR-3␣ forms had little effect. We also found abundant expression in the neointima of FGF-9, an FGF ligand capable of binding to the FGFR-2/FGFR-3 isoforms. FGF-9 markedly stimulated in vitro the proliferation of cultured neointimal SMCs and the activation of ERK 1/2, the latter effects of which were abrogated by expression of the DN-FGFR-2␤IIIcAB adenovirus; finally, in vivo administration of antisense FGF-9 oligonucleotides to in- FGFR-2:FGF-9 Interactions Regulate Intimal SMC Proliferation jured arteries significantly inhibited neointimal SMC proliferation, concomitant with a reduction in FGF-9 protein levels.
In a number of cell systems, FGF-9 appears to act as an autocrine/paracrine growth factor and has been shown to stimulate endometrial stromal cell proliferation (33) and locomotion of human teratocarcinoma cells (34). It is actively secreted, and this feature distinguishes it from a number of other FGFs that are localized intracellularly and require cell damage for release (31). In vivo, FGF-9 exhibits cell type-specific effects. It has been implicated in sex determination, stimulation of mesenchymal cell proliferation, mesonephric cell migration, and Sertoli cell differentiation in embryonic testis (35,36). It supports survival of cholinergic neurons (37), and its overexpression disturbs skeletal development and linear bone growth (38). Our studies indicate that FGF-9 is a potent mitogenic factor that stimulates neointimal SMC proliferation in vitro and in injured arteries via its interaction with FGFR-2␤IIIcAB and the subsequent activation of the ERK 1/2 pathway. Our findings that FGF-9 induces persistent activation of ERK 1/2, maintained for at least 16 h after its exposure to neointimal SMCs, is consistent with earlier studies indicating that extensive glycosaminoglycan modification of FGFR-2␤IIIcAB at a serine residue located 5Ј to the acidic box increases the ability of the receptor to sustain both autophosphorylation and subsequent phosphorylation of its intracellular substrates (19). This results in a potentiation of the stimulatory effects of physiological concentrations of FGF ligands on DNA synthesis in NIH 3T3 cells via a sustained activation of ERK 1/2 (19). Consistent with this hypothesis is our demonstration that adenoviral expression of DN-FGFR-2␤IIIcAB in neointimal SMCs yields both a protein of expected molecular mass (37 kDa) and species of proteins of higher molecular masses (55-80 kDa), most likely representing the glycosaminoglycan-modified receptor. Furthermore, the absence of Ig-like domain I in this receptor isoform in conjunction with the glycosaminoglycan modification is likely to confer upon the ectodomain of FGFR-2␤IIIcAB an enhanced resistance to degradation by proteases such as matrix metalloproteinase-2 (39), which is known to be expressed in injured arteries (40). These properties may also act to protect interacting FGF ligands from protease degradation (41,42) and, in a "proteolytically active" environment, are consistent with an FGFR-2␤IIIcAB/FGF-9 interaction acting to stimulate neointimal SMC proliferation.
Despite findings that indicate that FGF-9 interacts with FGFR-3␣IIIbAB and FGFR-3␣IIIcAB (17), our studies with the dominant negative FGFR-3 forms indicate that only FGFR-3␣IIIbAB contributes to neointimal SMC proliferation and in a much less potent manner than does FGFR-2␤IIIcAB. In colonic epithelial cells FGFR-3␣IIIbAB has been reported to promote proliferation (43), and this isoform is known to be the major FGFR-3 isoform expressed by epithelial cells (44), whereas non-epithelial cell populations express predominantly FGFR-3␣IIIcAB isoforms (44). Our findings that these FGFR-3␣ isoforms do not significantly contribute to neointimal SMC proliferation appear to be consistent with the inability of these isoforms to be extensively post-translationally modified (in contrast to FGFR-2␤IIIcAB), most likely via the abrogation of this modification by Ig-like domain I (19). Thus, the function of such FGFR-3 isoforms in neointimal development remains to be fully elucidated. Although FGFR-3 signaling has been reported previously to induce growth arrest in chondroprogenitor cells (45), a similar action in neointimal SMCs appears unlikely given that the DN-FGFR-3␣ forms did not augment the proliferation of these cells.
In conclusion, we have for the first time defined a novel FGFR/FGF signaling pathway in blood vessels that is activated by injury and is responsible for a significant component of neointimal SMC proliferation in such injured arteries. The pathway involves injury-induced activation of the expression of a secreted FGF, FGF-9, and its high affinity interaction with the extensively glycosaminoglycan-modified FGFR-2␤IIIcAB isoform. This interaction leads to a sustained activation of the ERK 1/2 signaling pathway and the rapid proliferation of neointimal SMCs. Thus, targeting of this pathway may represent a useful approach aimed at inhibiting proliferative vascular diseases and related pathological processes involving neointimal SMC proliferation and accumulation.