Differential regulation of the fibroblast growth factor (FGF) family by α₂-macroglobulin: evidence for selective modulation of FGF-2–induced angiogenesis

Fibroblast growth factors (FGF) constitute a family of heparin-binding proteins that exert pleiotropic effects on cells from all embryonic lineages.1,2 Sequencing of FGF-1 and FGF-2 has led to the identification of at least 19 proteins that are members of this mammalian family.3,4 The FGFs are expressed during both embryogenesis and in mature organisms.5 They play important roles in angiogenesis, mitogenesis, embryonic pattern formation and development, cellular differentiation, and wound healing.1,6-10 Only FGF-1 and FGF-2 are expressed at high levels in adults. FGF-1 expression is predominantly confined to the central nervous system, but FGF-2 is ubiquitously expressed throughout all adult tissues.5FGF-2 is a potent mitogen for mesoderm-derived cells, such as endothelial cells.1,11 FGF-2 induces cell proliferation in endothelial cells derived from large vessels or capillaries with a median effective concentration of 1.5 to 2.6 pM.12 In model systems of angiogenesis, such as the rabbit cornea, the chick chorioallantoic membrane, and the hamster cheek pouch, FGF-2 exerts a potent angiogenic effect.6,13-15 FGF-1 and -2 are both involved in vasculogenesis, because epiblast cells can be induced to differentiate into endothelial cells by incubation with either of these FGFs.8 

FGF-2 up-regulates the urinary plasminogen activator receptor and stimulates the release of urinary plasminogen activator and collagenase in endothelial cells.16-18 Additionally, FGFs act as chemoattractants for endothelial cells.19 The ability of the FGFs to exert their effects depends on their interaction with both cell-surface receptor tyrosine kinases and extracellular and cell-surface–bound heparan sulfate proteoglycans.20 21However, the nature of the interactions of the FGFs with proteins and molecules that are predominantly fluid-phase has not been extensively investigated.

The plasma protein α₂-macroglobulin (α₂M) is a 718-kd homotetrameric glycoprotein that inactivates proteinases from all 4 mechanistic classes.22 Proteinase inhibition involves a major conformational change and compaction of the native α₂M structure.23 This conformational change results in exposure of receptor binding domains that mediate interaction of α₂M with its receptors, the low-density lipoprotein-related protein,24 and the α₂M signaling receptor.24-26 Ammonia or small amines such as methylamine can directly attack the thiolester bonds present in each α₂M subunit. This converts native α₂M into a conformational variant indistinguishable from activated α₂M-proteinase complexes with respect to receptor recognition27 and cytokine binding.28 Forms of α₂M activated by either small nucleophiles or proteinases are considered functionally equivalent and are designated α₂M*.

In vitro and in vivo, α₂M binds to numerous cytokines and growth factors.29,30 Binding to α₂M and α₂M* typically inhibits the ability of a given growth factor to regulate cellular functions in vitro, although such binding may enhance the activity of platelet-derived growth factor (PDGF).31,32 An important function of α₂M is to serve as a reservoir for cytokines and growth factors. Examples of this include tumor growth factor (TGF)-β and PDGF, where 80% to 90% of these growth factors are associated with α₂M in the circulation.33-36 α₂M binds FGF-2 in serum,16 and an equilibrium dissociation constant (K_d) is available for its binding to α₂M* but not α₂M.28 

In the present study, we sought to determine whether α₂M and α₂M* interact with other members of the FGF family and to derive equilibrium K_ds for these interactions. Further, given the high expression of FGF-2 in adult tissues and its important role in angiogenesis, we investigated whether α₂M or α₂M* regulate the ability of FGF-2 to induce proliferation and differentiation of endothelial cells, both important steps in angiogenesis.37 We report that binding to α₂M and α₂M* is restricted to certain members of the FGF family. We also demonstrate that although α₂M* is able to inhibit the activity of FGF-2 in proliferation assays, it does not block the ability of FGF-2 to induce endothelial tubule formation on Matrigel basement membranes or on collagen gels. These studies indicate that α₂M* and, to some extent, α₂M can regulate the activity of FGF-2 in the fluid phase. The basement membrane, however, appears slightly more important than α₂M* in modulating the activity of FGF-2. This may serve to confine FGF activity to the site of ongoing proteinase activity associated with angiogenesis, inflammation, and tissue repair because only basement membrane or extracellular matrix (ECM)-bound FGF-2 can evade α₂M* inhibition.

¹²⁵I-labeled Bolton-Hunter reagent and [methyl-³H]thymidine were purchased from NEN Life Science Products (Boston, MA). RPMI 1640, Dulbecco phosphate-buffered saline (PBS), L-glutamine, penicillin/streptomycin, and fetal bovine serum (FBS) were purchased from Life Technologies (Gaithersburg, MD). Endothelial basal medium (EBM), human endothelial growth factor, bovine pituitary extract, and hydrocortisone were purchased as the endothelial growth medium kit from Clonetics (San Diego, CA); 5,5′-dithio-bis(2-nitrobenzoic acid), methylamine (CH₅NHCl), gelatin, and bovine serum albumin (BSA) were purchased from Sigma (St Louis, MO). Electrophoresis reagents were purchased from Bio-Rad Laboratories (Richmond, CA). All other reagents were of the highest quality commercially available.

Human native α₂M was purified from plasma according to a previously described protocol.38 The α₂M was at least 90% active against proteinases as determined by thioester titration using the 5,5′-dithio-bis(2-nitrobenzoic acid) assay.39 The protein concentration was based on A_280nm (1%/1 cm) = 8.93 and molecular mass = 718 kd. The α₂M-methylamine (α₂M*) was prepared as previously described40 and employed throughout these experiments. Recombinant, carrier-free TGF-β₁ and FGF-1, -2, -4, -5, -6, -7, -9, and -10 were purchased from R&D Systems (Minneapolis, MN). FGFs and α₂M* were labeled with¹²⁵I-labeled Bolton-Hunter reagent according to the manufacturer's recommended protocol. Both labeled and unlabeled FGFs were reconstituted with 0.5% BSA in PBS, pH 7.4.¹²⁵I-labeled FGFs were used within 2 weeks of labeling.

¹²⁵I-FGFs (15 ng) were incubated with α₂M or α₂M* for 2 hours at 37°C in the presence of 0.1% BSA in PBS, pH 7.4. Following incubation, the mixture was subjected to electrophoresis on nondenaturing pore-limit gels to separate ¹²⁵I-FGF bound to α₂M or α₂M* from unbound ¹²⁵I-FGF. Following electrophoresis, gels were stained with Coomassie brilliant blue to verify the location of α₂M or α₂M*. Gels were subsequently dried and exposed to a Phosphorimager 410A (Molecular Dynamics, Sunnyvale, CA) plate. The plate was developed, and the radioactivity associated with the bands corresponding to α₂M or α₂M* was quantified employing ImageQuant analysis software (version 3.3, Molecular Dynamics). Simultaneously, ¹²⁵I-FGFs were electrophoresed under denaturing conditions to determine total radioactivity added. Equilibrium K_ds were determined by direct fit of the Phosphorimager counts to a one-site binding model employing SigmaPlot (version 3.02, Jandel Scientific, San Rafael, CA). TheK_d determined for the binding of FGF-2 to α₂M and α₂M* was corrected for nonspecific binding.

Nonspecific binding of ¹²⁵I–FGF-2 to α₂M and α₂M* was determined as described previously.41 Briefly, 96-well plates were coated with 1 μg of either α₂M or α₂M* by incubation in 0.015 M Na₂CO₃, 0.035 M NaHCO₃, and 0.04% NaN₃ for 4 hours at 25°C. The coated wells were blocked overnight with 0.02 M HEPES and 0.15 M NaCl, pH 7.5, containing 0.1% (vol/vol) Tween 20 and stored at 4°C prior to performing binding studies with FGF-2. The total amount of α₂M and α₂M* in the wells was determined in concurrent experiments that determined the amount of¹²⁵I-α₂M and¹²⁵I-α₂M* that could be coated onto the wells. ¹²⁵I–FGF-2, 10 ng, was added to the wells in the presence or absence of a 1000-fold molar excess of unlabeled FGF-2, and the wells were then incubated at 37°C for 2 hours. The wells were then washed with 0.02 M HEPES and 0.15 M NaCl, pH 7.5, to remove any unbound ¹²⁵I–FGF-2. The radioactivity associated with each well was determined by γ-counting. Nonspecific binding of¹²⁵I–FGF-2 to α₂M and α₂M* was 15%, which is similar to previous determinations of nonspecific binding of cytokines and growth factors to α₂M.42 43 

Ninety-six–well plates were coated with 1 μg of either α₂M or α₂M* by incubation in 0.015 M Na₂CO₃, 0.035 M NaHCO₃, and 0.04% NaN₃ for 4 hours at 25°C. The coated wells were blocked overnight with 0.02 M HEPES and 0.15 M NaCl, pH 7.5, containing 0.1% (vol/vol) Tween 20 and stored at 4°C prior to performing competition binding studies with FGF-2 and TGF-β₁. The total amount of α₂M and α₂M* in the wells was determined in concurrent experiments that determined the amount of¹²⁵I-α₂M and¹²⁵I-α₂M* that could be coated onto the wells. ¹²⁵I–FGF-2, 10 ng, was added to the wells in the presence or absence of a 100-fold molar excess of unlabeled TGF-β₁, and the wells were then incubated at 37°C for 2 hours. The wells were then washed with 0.02 M HEPES and 0.15 M NaCl, pH 7.5, to remove any unbound ¹²⁵I–FGF-2. The radioactivity associated with each well was determined by γ-counting.

Fetal bovine heart endothelial (FBHE) cells were obtained from American Type Culture Collection (Manassas, VA) and cultured in gelatinized 75-cm² flasks in RPMI 1640 medium supplemented with 50 ng/mL FGF-2, 10% FBS, 2 mM L-glutamine, and 3% penicillin/streptomycin. Forty-eight hours prior to experimentation, FBHE cells were trypsinized from flasks and incubated in gelatinized 96-well plates at 2000 cells per well in FGF-2–deficient medium to reach quiescence. On the day of experimentation, the cell medium was replaced with one containing 0.5 ng/mL FGF-2 that had been incubated with varying concentrations of α₂M or α₂M* for 2 hours at 37°C, and the plates were then allowed to incubate for 48 hours at 37°C. Following incubation, each well was pulsed with [methyl-³H] thymidine (3.7 × 10⁴ Bq) and harvested with a Combi Cell Harvestor (Skatron, Norway) onto glass fiber filters for scintillation counting.

This assay was also performed with human umbilical vein endothelial cells (HUVECs) obtained from Clonetics (San Diego, CA) and employed at passage 4. HUVECs were cultured in gelatinized 75-cm²flasks in EBM supplemented with 500 ng/mL endothelial growth factor, 10% FBS, 1 ng/mL hydrocortisone, 24 ng/mL bovine pituitary extract, and gentamicin/amphotericin. Forty-eight hours prior to experimentation, HUVECs were trypsinized from flasks and incubated in gelatinized 96-well plates at 2000 cells per well in EBM supplemented with 2% FBS, 1 ng/mL hydrocortisone, and gentamicin/amphotericin to reach quiescence. On the day of experiment, this quiescent cell medium was replaced with one containing 0.5 ng/mL FGF-2 that had been incubated with varying concentrations of α₂M or α₂M* for 2 hours at 37°C. The plates were then allowed to incubate for 48 hours at 37°C, after which the wells were pulsed with [methyl-³H] thymidine (1 μCi) and harvested with a Combi Cell Harvestor onto glass fiber filters for scintillation counting.

Growth factor–reduced Matrigel was purchased from Collaborative Biomedical Products (Bedford, MA) and employed according to the manufacturer's recommended protocol to coat 24-well plates. FBHE cells that had been cultured in FGF-2–deficient medium for 48 hours were trypsinized from flasks and plated onto the Matrigel layers at 40 000 cells per well in the presence of varying concentrations of FGF-2. Where the effect of α₂M* on the ability of FGF-2 to induce endothelial tubule formation was tested, varying concentrations of α₂M* were incubated with FGF-2 for 2 hours at 37°C prior to addition to the cells. The FBHE cells were then incubated for 96 hours at 37°C and photographed.

Endothelial tubule formation was also studied employing collagen gels. Rat tail collagen type I was purchased from Collaborative Biomedical Products and employed according to the manufacturer's protocol to coat 48-well plates. HUVECs at passage 4 that had been cultured in EBM supplemented with 2% FBS, 1 ng/mL hydrocortisone, and gentamicin/amphotericin for 48 hours were trypsinized from flasks and plated onto the collagen gels in EBM supplemented with 1 ng/mL hydrocortisone and gentamicin/amphotericin at 30 000 cells per well in the presence of varying concentrations of FGF-2. Where the effect of α₂M* on the ability of FGF-2 to induce endothelial tubule formation was tested, varying concentrations of α₂M* were incubated with FGF-2 for 2 hours at 37°C prior to addition to the cells. The HUVECs were then incubated for 24 hours at 37°C and photographed.

Matrigel layers were prepared as above in 24-well plates.¹²⁵I–FGF-2 (15 ng) was added to the layers in cell medium and incubated for 48 hours at 37°C. The Matrigel layers were then washed 3 times with cell medium and solubilized with Matrisperse (Collaborative Biomedical Products) according to the manufacturer's instructions. The solubilized Matrigel was subjected to nonreducing 15% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), stained, and dried. The gels were exposed to a Phosphorimager plate, and the radioactivity associated with the¹²⁵I–FGF-2 was quantified employing ImageQuant. The α₂M* was tested for its ability to inhibit binding of I¹²⁵–FGF-2 to the Matrigel layer by incubating 0.3 mg/mL α₂M* with ¹²⁵I–FGF-2 for 2 hours at 37°C prior to addition to the Matrigel. Studies were performed to determine the amount of ¹²⁵I–FGF-2 that was associated with α₂M* on the Matrigel layer. The solubilized Matrigel, following incubation as above with ¹²⁵I–FGF-2 and α₂M*, was electrophoresed on nondenaturing pore-limit gels to allow separation of ¹²⁵I–FGF-2 bound to α₂M* from free ¹²⁵I–FGF-2, and the radioactivity associated with the α₂M* band was quantified as above. In all cases, 15 ng ¹²⁵I–FGF-2 alone was subjected to nonreducing SDS-PAGE and exposed simultaneously with the nondenaturing gels for which it served as 100% of total¹²⁵I–FGF-2 added. Binding of¹²⁵I-α₂M* to Matrigel was performed by incubating 4 μg ¹²⁵I-α₂M* with Matrigel layers in cell medium for 48 hours at 37°C. The supernatant was aspirated, and the Matrigel layers were washed 3 times with cell medium and solubilized with Matrisperse. The solubilized Matrigel and supernatant were electrophoresed on nondenaturing pore-limit gels, and the radioactivity associated with the α₂M* band was quantified as above and compared with 4 μg¹²⁵I-α₂M* alone.

The ability of FGF-2 to bind to collagen gels was also studied. Collagen gels were prepared as above in 48-well plates.¹²⁵I–FGF-2 (100 ng) was added to the gels in cell medium and incubated for 24 hours at 37°C. The collagen gels were then washed 3 times with cell medium and solubilized according to the manufacturer's instructions. The solubilized collagen was subjected to nonreducing 5% SDS-PAGE, stained, and dried. The gels were exposed to a Phosphorimager plate, and the radioactivity associated with the¹²⁵I–FGF-2 was quantified employing ImageQuant. The α₂M* was tested for its ability to inhibit binding of I¹²⁵–FGF-2 to the collagen gel by incubating 0.5 mg/mL α₂M* with ¹²⁵I–FGF-2 for 2 hours at 37°C prior to addition to the collagen gel. In all cases, 100 ng¹²⁵I–FGF-2 alone was subjected to nonreducing SDS-PAGE and exposed simultaneously with the nondenaturing gels for which it served as 100% of total ¹²⁵I–FGF-2 added.

The FGFs are a family of highly homologous growth factors.5,44 Previous studies have determined that FGF-2 binds to α₂M and α₂M*16,28 and that FGF-1 can compete for this binding.16 To determine whether binding to α₂M or α₂M* is a conserved feature of the FGF family, we performed in vitro binding experiments with the following members of the FGF growth factor family: FGF-1, -2, -4, -5, -6, -7, -9, and -10. We screened these growth factors for binding to α₂M and α₂M* by radiolabeling each FGF with ¹²⁵I and incubating them with a fixed concentration of α₂M or α₂M*, 0.5 mg/mL (about 700 nM), as described in “Materials and methods.” The amount of each ¹²⁵I-FGF bound to either α₂M (Figure 1A) or α₂M* (Figure1B) was determined by reference to the total radioactivity of each¹²⁵I-FGF alone.

Interestingly, there were dramatic differences in the amount of each FGF that bound to α₂M and α₂M*. A substantial fraction of FGF-1, -2, -4, and -6 bound to both α₂M and α₂M*. However, only a very small fraction of FGF-5, -7, -9, and -10 bound to α₂M or α₂M*. To ensure that FGF-5, -7, -9, and -10 had adequate time to achieve equilibrium, these labeled FGFs were incubated with α₂M and α₂M* for 24 hours at 37°C. However, the binding was not improved following this prolonged period of incubation (data not shown).

Because FGF-5, -7, -9, and -10 displayed only a very slight interaction with α₂M or α₂M*, we chose to focus our binding studies on FGF-1, -2, -4, and -6 (Figure2). These radiolabeled ligands were incubated with α₂M or α₂M*, and the binding of each ¹²⁵I-FGF to α₂M and α₂M* was detected and quantitated as described in “Materials and methods.” A summary of the equilibriumK_ds derived from these experiments is presented in Table 1.

The K_d value we determined for the binding of FGF-2 to α₂M* is consistent with previous studies.28 In all cases where the FGF studied bound to α₂M or α₂M*, binding was of higher affinity to α₂M*. This result is in line with previous work, where differential binding affinities of cytokines and growth factors to α₂M and α₂M* is typically observed.28 45 In addition, we observed that members of the FGF family displayed varying ability to bind to α₂M or α₂M*. Of those FGFs that bound to α₂M and α₂M*, the binding affinities differed. The ranking (from greatest to least affinity) was FGF-2 > FGF-4 = FGF-6 > FGF-1.

Given that FGF-2 had the greatest affinity for α₂M and α₂M*, we chose to focus on it to further explore the ability of α₂M and α₂M* to regulate the activity of the FGF family. An important issue is the α₂M and α₂M* binding site for the FGFs. Typically, cytokines and growth factors bind to similar or identical sites on both α₂M and α₂M*.42,43,46 However, recent work in our laboratory has indicated that α₂M and α₂M* can employ distinct sites to bind certain growth factors, such as vascular endothelial growth factor.41Previous work has localized the α₂M binding site of TGF-β₁ to a stretch of the amino acids that includes the bait region.43 46 To investigate whether FGF-2 binds to α₂M and α₂M* at this TGF-β₁binding site, we incubated α₂M and α₂M* with ¹²⁵I–FGF-2 in the presence or absence of a 100-fold molar excess (relative to FGF-2) of TGF-β₁, as described in “Materials and methods.” Competition by TGF-β₁reduced the binding of FGF-2 to α₂M by 49% and to α₂M* by 43%.

Previous studies demonstrated the ability of α₂M to inhibit in vitro induction by FGF-2 of plasminogen activator release in endothelial cells.16 We sought to determine whether α₂M or α₂M* could inhibit the ability of FGF-2 to induce proliferation of FBHE cells. Initial studies were conducted to determine concentrations of FGF-2 that were effective in inducing proliferation in FBHE cells (data not shown). The optimal concentration for our studies was 0.5 ng/mL FGF-2. Therefore, we incubated 0.5 ng/mL FGF-2 with varying concentrations of α₂M or α₂M* for 2 hours at 37°C and added this mixture to quiescent FBHE cells. The α₂M* reduced the proliferation of FBHE cells in a dose-dependent manner, inhibiting about 70% of the proliferation observed with FGF-2 alone (Figure3). The α₂M was less inhibitory, reducing proliferation by about 20% at the highest concentration tested. Similar studies employing HUVECs demonstrated that α₂M* was also able to inhibit the ability of FGF-2 to stimulate HUVEC proliferation in a dose-dependent manner, inhibiting about 64% of the proliferation observed with FGF-2 (data not shown). Again, α₂M was less inhibitory, reducing proliferation by about 20% at the highest concentrations tested.

We then sought to determine whether α₂M*, which was able to significantly inhibit endothelial cell proliferation in vitro (Figure 3), could also regulate the ability of FGF-2 to induce the differentiation of endothelial cells, such as formation of endothelial tubules. Matrigel is a solubilized basement membrane matrix extracted from Engelbreth-Holm-Swarm mouse tumor47 that has been demonstrated to induce endothelial tube formation by a variety of endothelial cells, including FBHE cells.48Twenty-four–well plates were coated with Matrigel as described in “Materials and methods,” and 40 000 FBHE cells per well were plated in either media alone, media with varying concentrations of FGF-2, or media containing varying concentrations of FGF-2 incubated with 0.5 mg/mL α₂M* for 2 hours at 37°C prior to addition to the cells. Initially, under all conditions, FBHE cells formed endothelial tubules on Matrigel. However, by 24 hours, cells not treated with FGF-2 began to regress. After 96 hours of incubation, only those FBHE cells incubated on Matrigel in the presence of at least 0.5 ng/mL FGF-2 remained organized into endothelial tubules (Figure4A,B). We then examined the effect of 0.5 mg/mL α₂M* on the activity of 0.625 ng/mL FGF-2 (Figure4C). Interestingly, the presence of α₂M* had no effect on either the timing of endothelial tubule formation or the survival of the endothelial tubules in the presence of FGF-2. Increasing the concentration of α₂M* had no effect on the activity of FGF-2, and FBHE cells treated with α₂M* alone behaved as those cells treated with media alone (data not shown).

Matrigel is a complex mixture of basement membrane components and growth factors.47 To ensure that the inability of α₂M* to inhibit the induction of endothelial tubules by FGF-2 was not due to the presence of other growth factors present in Matrigel, we employed collagen type I gels, which have been demonstrated to support the formation of endothelial tubules by HUVECs.49 Forty-eight–well plates were coated with collagen gel as described in “Materials and methods,” and 40 000 FBHE cells or 30 000 HUVECs per well were plated in either media alone, media with varying concentrations of FGF-2, or media containing varying concentrations of FGF-2 incubated with 0.5 mg/mL α₂M* for 2 hours at 37°C prior to addition to the cells. FBHE cells did not organize into endothelial tubules on collagen gels under any of the conditions tested—nor did HUVECs plated onto collagen in the absence of FGF-2 (Figure5A). However, in the presence of at least 50 ng/mL FGF-2, HUVECs were stimulated to organize into endothelial tubules (Figure 5B). The α₂M* was unable to inhibit the ability of FGF-2 to induce the formation of endothelial tubules by HUVECs (Figure 5C). Again, increasing the concentration of α₂M* had no effect on the activity of FGF-2, and HUVECs treated with α₂M* alone behaved as those HUVECs treated with media alone (data not shown).

We then sought to understand why α₂M* appeared to have no effect on the ability of FGF-2 to promote endothelial tubule formation (Figures 4 and 5), given that α₂M* profoundly reduces the ability of FGF-2 to induce endothelial cell proliferation (Figure 3). One possibility is that FGF-2 can partition away from its complex with α₂M* and bind to components of the Matrigel basement membrane or to collagen gels, thereby evading inhibition by α₂M*. We tested whether FGF-2 was able to bind to Matrigel by coating 24-wells plates with Matrigel, as described in “Materials and methods,” and then incubating radiolabeled FGF-2 alone with the Matrigel for 48 hours at 37°C or radiolabeled FGF-2 incubated with α₂M* for 2 hours at 37°C prior to addition to Matrigel. The Matrigel layers were washed to remove any unbound ¹²⁵I–FGF-2 and solubilized, followed by electrophoresis on nonreducing 15% SDS-PAGE, and then compared with the total radioactivity of the ¹²⁵I–FGF-2 alone. The percentage of ¹²⁵I–FGF-2 bound to Matrigel in the presence or absence of α₂M* was quantitated as described in “Materials and methods” (Figure 6). A substantial fraction, 62% ± 11%, of the FGF-2 added in the solution phase over Matrigel layers partitioned out of solution and bound to Matrigel. The percentage of FGF-2 that bound to Matrigel was reduced by approximately 50% (to 30% ± 10% of the¹²⁵I–FGF-2 added) in the presence of α₂M*.

Because it is possible that α₂M* may bind to Matrigel carrying bound ¹²⁵I–FGF-2, we determined whether the FGF-2 bound to the Matrigel layer was free or bound to α₂M*. We incubated ¹²⁵I-α₂M* with Matrigel for 48 hours at 37°C and determined the amount of α₂M* that had bound to the Matrigel. The amount of α₂M* was 15% ± 4% of the total ¹²⁵I-α₂M* added. We then incubated ¹²⁵I–FGF-2 with α₂M* for 2 hours at 37°C and added this mixture to Matrigel for 48 hours at 37°C. The radioactivity associated with the α₂M* band in the Matrigel fraction was quantified with a Phosphorimager. The percentage of ¹²⁵I–FGF-2 bound to α₂M* on the Matrigel layer was 6% ± 1% (n = 3) of the total¹²⁵I–FGF-2.

Similarly, we tested whether FGF-2 was able to partition away from its complex with α₂M* and bind to collagen gels by coating 48-well plates with collagen type I, as described in “Materials and methods,” and then incubating radiolabeled FGF-2 alone with the collagen gel for 24 hours at 37°C or radiolabeled FGF-2 incubated with α₂M* for 2 hours at 37°C prior to addition to collagen gel. The collagen gels were washed to remove any unbound¹²⁵I–FGF-2 and solubilized, followed by electrophoresis on nonreducing 5% SDS-PAGE, and then compared with the total radioactivity of the ¹²⁵I–FGF-2 alone. The percentage of¹²⁵I–FGF-2 bound to collagen gel in the presence or absence of α₂M* was quantified as described in “Materials and methods.” We found that 76% of the FGF-2 bound to α₂M* added in the solution phase over collagen gels partitioned out of solution and bound to collagen gels.

In the present study, we have shown that FGF-1, -2, -4, and -6 bind to α₂M and α₂M* in solution. The ability of α₂M and α₂M* to bind to certain members of the FGF family and not to others was unexpected. The FGFs are highly homologous in certain regions, which provides the basis for assigning membership to this family.5 However, of the FGFs we tested, only for FGF-1, -2, -4, and -6 was a significant fraction of the FGF complexed to α₂M and α₂M* at equilibrium. FGF-5, -7, -9, and -10 bound insignificantly to α₂M or α₂M*.

A binding site for α₂M and α₂M* may therefore be present on certain FGFs but missing on other FGFs, accounting for their differences in binding activity. Binding to α₂M and α₂M* may provide an additional level of regulation of FGF family growth factor activities. Analyses of the evolutionary relationships between the FGFs show that FGF-1 and FGF-2 are more closely related to each other than to other FGFs. Similarly, FGF-4 and FGF-6 are most closely related to each other.3,50 Thus, it is not surprising that ability to bind to α₂M or α₂M* is shared by these pairs of FGFs. However, FGF-7 and -10 are more closely related to FGF-1/FGF-2 than FGF-4/FGF-6, yet they do not interact significantly with α₂M.3 This may support the concept of a latent binding site for α₂M that is active only in certain members of the FGF family.

The FGFs that we investigated bind to α₂M* with higher affinity than to α₂M. Such differential binding to α₂M or to α₂M* has been observed for many growth factors and cytokines.28 Interestingly, we found that TGF-β₁ competed with FGF-2 for binding to α₂M and α₂M*. Previous studies have demonstrated that TGF-β and PDGF bind to a 20-kd fragment of α₂M that includes the bait region.43 TGF-β and PDGF also bind to a synthetic peptide comprising a 16–amino acid portion of this same region.46 The higher affinity of binding of the FGFs to α₂M*, as compared with α₂M, must presumably be due to the major conformational change that occurs upon the conversion of α₂M to α₂M* by proteinases or small-amine nucleophiles, such as methylamine. This conformational change may permit more ready access of the FGF to the binding site in α₂M*. The hydrophobic nature of the binding site46 suggests that the FGFs are interacting with α₂M and α₂M* via a mechanism that is primarily hydrophobic in nature and, therefore, exquisitely sensitive to changes in temperature and pH. Further studies exploring these questions and whether physiologic pH and temperature variations may dramatically alter the ability of α₂M and α₂M* to modulate the activity of the FGFs are ongoing in our laboratory.

α₂M may bind a significant fraction of FGF-1, -2, -4, -6 in vivo despite the high-nanomolar equlibriumK_ds for their interaction because α₂M is present in micromolar concentrations in vascular fluids and in the extravascular space, especially in states of increased vascular permeability.51-56 α₂M may also serve as a carrier in serum for FGF-1, -2, -4, and -6, a role played by heparan sulfate proteoglycans on the cell surface or in the ECM.57 At sites of ongoing proteolysis, such as tissue remodeling during angiogenesis,58 α₂M encounters large amounts of activated proteinases that interact with it and convert it to α₂M*, which is rapidly cleared from such sites by binding to lipoprotein-related protein.24,59Cytokines and growth factors bound to α₂M* can thereby also be cleared,29 possibly allowing α₂M* to directly down-regulate FGF activity.

Among many other activities, FGF-2 plays an important role in angiogenesis.60,61 Angiogenesis is a complicated process that can be divided into a series of events, including digestion of the basement membrane surrounding a blood vessel, migration and proliferation of endothelial cells, and endothelial tube formation.37,62,63 We chose to study whether α₂M might regulate angiogenesis by investigating the ability of α₂M to affect certain angiogenic properties of FGF-2. Previous studies demonstrated that α₂M inhibited FGF-2–stimulated plasminogen activator release.16 The present study demonstrates that α₂M* is able to inhibit FGF-2–stimulated endothelial cell proliferation in a dose-dependent fashion, whereas α₂M is less able to inhibit this activity of FGF-2.

Because α₂M* may be present in high concentrations in inflammatory lesions due to proteinase inhibition by α₂M, FGF-2 must be able to escape inhibition by α₂M* when and where appropriate. We demonstrate here that α₂M* is unable to block the ability of FGF-2 to induce the formation and survival of endothelial tubules by FBHE cells on Matrigel basement membranes or by HUVECs on collagen gels. This appears to be because FGF-2 partitions away from α₂M*–FGF-2 complexes to bind to Matrigel basement membranes or to the collagen gel. Matrigel contains heparan sulfate proteoglycans47 that would provide binding sites for FGF-2 and allow presentation of FGF-2 to FGF receptor tyrosine kinase on the surface of the FBHE cells, enabling FGF-2 to evade α₂M*-mediated inhibition. The presence of heparan sulfate proteoglycans in the ECM may therefore serve to protect FGF-2 from clearance by α₂M*. In addition, our studies indicate that FGF-2 may also interact directly with other important components of the basement membrane, such as collagen, again enabling FGF-2 to evade α₂M*-mediated inhibition.

We show here that the regulation of FGF activity depends not only on the presence of fluid-phase inhibitors such as α₂M but also, critically, on the presence of ECM. In the absence of ECM, α₂M*, and to some extent α₂M, can dramatically inhibit the activity of FGF-2. However, the ECM is able to protect FGF-2 from inhibition by α₂M and α₂M*. These findings have implications for the proposed regulation in vivo by α₂M and α₂M* of other growth factors and cytokines that have been demonstrated to bind to α₂M or α₂M* in vitro. We suggest that the activities of growth factors or cytokines that bind to α₂M or α₂M* must be examined in an assay system that includes other important contributors to cellular function and activity. We show here, employing an in vitro model of the ECM, that regulation of FGF activity by α₂M is dependent on the extracellular environment. Importantly, this work suggests that, in vivo, mechanisms exist to allow growth factors to evade inhibition by α₂M and α₂M* in the presence of the ECM. Knowledge of the detailed kinetics of the interactions between α₂M, proteinases, growth factors, and the ECM may further elucidate the role that α₂M plays in these processes.

The ability of α₂M to affect cellular growth and differentiation through the binding and regulation of growth factors and cytokines may depend critically on its well-known regulatory role in proteinase activity. Inhibition of proteinase activity by α₂M, therefore, may permit α₂M to localize and modulate the activity of the many cytokines and growth factors and link proteinase activity, required for tissue remodeling, to the activity of growth factors that control these processes.

We thank Dr George Cianciolo for his critical reading of the manuscript and Marie Thomas for her assistance in preparation of the manuscript. We also gratefully acknowledge Susan Heffelfinger (Department of Pathology and Laboratory Medicine, University of Cincinnati, OH) for her expert advice concerning collagen gels and endothelial tubule formation.