Endocrinology Vol. 138, No. 7 2972-2983
Copyright © 1997 by The Endocrine Society
Insulin-Like Growth Factor Binding Protein-5 Binds to Plasminogen Activator Inhibitor-I2
Taek Jeong Nam,
Walker Busby, Jr. and
David R. Clemmons
Taek Jeong Nam,
Walker Busby, Jr. and
David R. Clemmons
Department of Medicine, University of North Carolina School of
Medicine, Chapel Hill, North Carolina 27599
Department of Medicine, University of North Carolina School of
Medicine, Chapel Hill, North Carolina 27599
Address all correspondence and requests for reprints to: David R. Clemmons, M.D., Division of Endocrinology, Department of Medicine CB 7170, University of North Carolina School of Medicine, Chapel Hill, North Carolina 27599-7170.
 |
Abstract
|
|---|
Insulin-like growth factor binding protein-5 (IGFBP-5) has been shown
to bind to the extracellular matrix (ECM) of both fibroblasts and
smooth muscle cells. The ECM-IGFBP-5 interaction is mediated in part by
binding to heparan sulfate containing proteoglycans. Because
proteoglycans may not be the only components of ECM that bind to
IGFBP-5, we have determined its ability to bind to other ECM proteins.
When a partially purified mixture of the proteins that were present in
fibroblast conditioned medium was purified by IGFBP-5 affinity
chromatography, a 55-kDa protein was eluted. Amino acid sequencing of
the amino terminal 28 amino acids showed that it was human plasminogen
activator inhibitor-1 (PAI-1). To determine if this interaction was
specific, purified human PAI-1 was incubated with IGFBP-5 and the
IGFBP-5/PAI-1 complex immunoprecipitated with anti-PAI-1 antiserum.
When the precipitate was analyzed by immunoblotting using anti-IGFBP-5
antiserum, the intensity of the IGFBP-5 band was substantially
increased compared with controls that did not contain human PAI-1. A
synthetic IGFBP-5 peptide that contained the amino acid sequence
between positions 201 and 218 inhibited IGFBP-5/PAI-1 interaction.
Coincubation of IGFBP-5 mutants that contained substitutions for
specific basic residues located between positions 201 and 218 with
PAI-1 indicated that some of these amino acids were important for
binding. Two mutants that contained neutral substitutions for specific
basic amino acids within the glycosaminoglycan binding domain had
reduced binding to PAI-1. In contrast, three other mutants that also
had substitutions for charged residues in the same region had no
reduction in binding. Heparin and heparan sulfate inhibited the
IGFBP-5/PAI-1 interaction; however, several other glycosaminoglycans
had no effect. PAI-1 was determined to be an important ECM component
for binding because approximately 27% of total ECM binding could be
inhibited with anti-PAI-1 antiserum. Competitive binding studies with
unlabeled IGFBP-5 showed that the dissociation constant of PAI-1 for
IGFBP-5 was 9.1 x 10-8 M. In summary,
IGFBP-5 binds specifically to plasminogen activator inhibitor-1.
Because this is present in the extracellular matrix of several cell
types, it may be one of the important binding components of ECM. PAI-1
binding partially protects IGFBP-5 from proteolysis, suggesting that it
is one of the ECM components that is involved in mediating this effect.
|
Introduction
|
|---|
THE INSULIN-LIKE growth factors (IGFs) are
present in physiological fluids bound to IGF binding proteins (IGFBPs)
(1, 2). Six IGF binding proteins with high affinities for IGF-I and II
have been identified (3). Although IGFBP-5 binds to the IGFs with high
affinity in interstitial fluids, when it is associated with
extracellular matrix its affinity is decreased 8- to 15-fold (4). This
lower affinity results in better equilibration of IGF-I with its
receptor and potentiation of IGF-I stimulated fibroblast (5) or smooth
muscle cell growth (6). Extracellular fluids also contain proteases
that degrade IGFBP-5 and cleave it into non-IGF binding fragments (7).
However, when IGFBP-5 is associated with ECM, it is protected from
proteolysis (5, 8); therefore, ECM binding may provide a mechanism for
localizing IGF-I in focal areas and provide better access to receptors.
This localization could be important in mediating the cellular response
to injury when IGF-I expression is increased in the local
microenvironment. Therefore, the factors in extracellular matrix that
account for IGFBP-5 localization have the potential to indirectly
regulate IGF-I actions.
We have previously shown that IGFBP-5 binds to heparin and heparan
sulfate (4, 9). Similarly, glycosaminoglycan containing proteoglycans,
such as tenascin, that are present in the ECM also bind IGFBP-5 (9).
This binding occurs through highly charged residues in IGFBP-5 that are
located between residues 201 and 218 (4, 9, 10). Similar to ECM
binding, glycosaminoglycan binding of IGFBP-5 also results in a
reduction in the affinity of IGFBP-5 for IGF-I (4) and in protection
from proteolysis (8). However, competitive binding studies using intact
ECM have shown that all of the ECM binding of IGFBP-5 could not be
accounted for by binding to proteoglycans (5). Therefore, these studies
were undertaken to determine if extracellular matrix proteins other
than proteoglycans might bind IGFBP-5.
|
Materials and Methods
|
|---|
Materials
Eagle’s MEM was purchased from Grand Island Biological Company
(Grand Island, NY, Life Technologies). Calf serum was purchased from
Colorado Serum (Colorado Laboratories, Denver, CO). BSA, ammonium
persulfate, sodium phosphate were obtained from Sigma Chemical Co. (St.
Louis, MO). 125I IGFBP-5 (30–45 µCi/µg) was prepared
using IODO-BEADS (Pierce Chemical Co., Rockford, IL) and 15 µg
IGFBP-5 according to manufacturer’s instructions (7, 11). Recombinant
human IGF-I was obtained from Bachem (Torrence, CA). Tris-sodium
duodecyl sulfate, chloramine-T, polyacrylamide, and prestained mol wt
standards were purchased from Bethesda Research Laboratories (BRL)
(Gaithersburg, MD). Tween 80 was obtained from Fisher Scientific
(Cleartown, NJ). Human vitronectin and antihuman vitronectin antiserum
were obtained from Telios (San Diego, CA).
Purification of PAI-I
Human dermal fibroblasts (GM-10) were purchased from Coriell
Institute (Camden, NJ). They were grown in EMEM supplemented with 10%
calf serum as described previously (7). Confluent monolayers were
washed twice in the serum-free medium, then 50 cc was added to
175-cm2 flasks (Falcon Labware, Division of
Becton-Dickinson), and the medium was collected after 24 h. This
was repeated until 1.2 liters of serum-free medium had been collected.
After purification by heparin sepharose affinity chromatography and
1 antichymotryptic peptide affinity chromatography as
described previously (12), the fractions containing IGFBP-5 protease
activity were applied to an IGFBP-5 affinity column. This column was
prepared by conjugating a synthetic 18 amino acid peptide that
contained the IGFBP-5 sequence between amino acids 201–218 to Affigel
10 (Bio-Rad Richmond, CA). This sequence is also present in IGFBP-3,
but it is not present in any other form of IGFBP. The peptide was
conjugated according to manufacturer’s instructions by incubating 5.6
mg of peptide with 7.0 cc of activated sepharose in 0.1 M
Na HEPES buffer, pH 7.5, for 4 h at 8 C. Unreacted sites were
blocked by a further incubation for 1 h with 0.25 M
Tris. The gel was washed, and the pool of activity that had been
purified from fibroblast conditioned medium was applied. The purified
active fractions that were eluted with 0.5 M NaCl were
concentrated and further analyzed by SDS-PAGE (12.5% gel). After
electrophoresis, the proteins were transferred to a nitrocellulose
membrane that was stained with Ponceau-S. A 55-kDa band was excised and
subjected to amino acid sequencing. Amino terminal amino acid
sequencing was conducted as described previously using Edman
degradation (13).
Preparation of human IGFBP-5
The human IGFBP-5 that was used in these studies was obtained by
purifying the protein from Chinese hamster ovary (CHO) cell conditioned
medium. CHO cells were transfected with an IGFBP-5 complementary DNA
(cDNA) that contained the entire protein coding region and had been
prepared as previously described (7). The cells were transfected, then
maintained in methotrexate (50 µg/ml) because the plasmid that was
used (pNUT) contains the gene that contains dihydrofolate reductase
resistance. After selection with methotrexate, 50 µg/ml, the highest
secreting clones were further subcultured in DMEM containing 10%
dialyzed FCS. The cells were grown to confluency and 50 ml of
serum-free medium were added to 175-cm tissue culture flasks and
incubated for 48 h. Eight hundred cubic centimeters of conditioned
medium were purified using a three-step purification procedure of
phenyl sepharose chromatography, IGF-I affinity chromatography, and
HPLC (C4 column) (Vydac, Hesperia, CA), as previously described (10).
The purified protein was homogenous by SDS-PAGE with silver staining
(14). The protein content was determined by amino acid composition
analysis (7).
Coimmunoprecipitation, ligand blotting, and immunoblotting
The ability of IGFBP-5 to bind to the plasminogen activator
inhibitor-1 was analyzed by coimmunoprecipitation. It was determined
that the lowest nonspecific binding and the highest sensitivity of
detection could be obtained by immunoprecipitating with anti-PAI-1
antisera. The IGFBP-5 that had been coimmunoprecipitated was then
detected by immunoblotting (5) or by ligand blotting using
125I-IGF-I (15). PAI-1 and polyclonal anti-PAI-1 antiserum
were obtained from American Diagnostica (Greenwich, CT). In most
experiments, IGFBP-5 (100 ng/ml) was incubated with PAI-1 at 100 ng/ml,
and anti-PAI-1 was added at a 1:5000 dilution. The reagents were
allowed to incubate for 16 h at 4 C in 0.03 M sodium
phosphate, pH 7.4, containing 0.01 M EDTA, 0.1% Tween 80,
and 0.02% BSA. Protein A sepharose (3 mg) was added and incubated for
2 h at 4 C. The immune complexes were precipitated by centrifuging
at 3000 x g for 5 min. The pellets were rinsed with
1.0 ml of the buffer listed previously and 50 µl of 2x Laemmli
sample buffer was added. The sample was centrifuged (2000 x
g for 5 min) and the supernatant loaded onto a 12.5% SDS
polyacrylamide gel. The proteins were electrophoresed through a 12.5%
gel using nonreducing conditions, then transferred to Immobilon
membranes (Millipore, Bedford, MA) as previously described (15). After
transfer, the filters were immunoblotted using a 1:1000 dilution of
anti-IGFBP-5 antiserum. The immune complexes were detected using the
Protoblot system (Promega, Madison, WI) according to manufacturer’s
recommendation. The electrophoretic mobilities of the bands that were
detected were compared with known mol wt standards (Bethesda Research
Laboratories, Gaithersburg, MD). The photographic negatives were
scanned using a Hoffer GS-300 scanner (Hoffer Scientific Instruments,
San Francisco, CA), and the results are expressed in scanning units.
For more sensitive detection of IGFBP-5, ligand blotting was performed.
Five hundred thousand counts per minute of 125I-IGF-I [150
µCi/µg (16)] was added to the filters and incubated overnight
using conditions described previously (17). The filters were washed,
and autoradiography was used to detect bound 125I-IGF-I.
For some experiments, quantitation of the amount of IGFBP-5 was
determined by analysis autoradiography band intensity using a Phor
Imager and Image Quant SF software (Molecular Dynamics, Sunnyvale, CA).
The results are expressed as arbitrarily defined scanning units.
To enhance the sensitivity detection of IGFBP-5 after
coimmunoprecipitation, 125I-IGFBP-5 (30 µCi/µg)
(260,000 cpm/ml) that had been prepared as described previously (12),
was incubated and was used with PAI-1 (200 ng/ml), and a 1:10,000
dilution of anti-PAI-1 antiserum in 0.25 ml of 0.03 M
sodium phosphate containing 0.01 M EDTA, 0.1% Tween 80,
0.1% BSA, pH 7.0. After an overnight incubation at 4 C, the immune
complexes were precipitated by adding rabbit IgG (1.0 µl) and
protein-A sepharose (10 µl of a 20 mg/ml solution) then centrifuged.
In some experiments, duplicate tubes contained unlabeled IGFBP-5 or
IGFBP-5 mutants (100 ng/ml). Similarly, in some experiments
unlabeled IGFBP-5 peptides that had been prepared as described
previously (10) were used. Fifty microliters of Lammeli sample buffer
were added, and the samples were heated to 65 C then electrophoresed
through a 12.5% SDS gel and transferred to Immobilon filters. The
precipitated 125I-IGFBP-5 was detected by autoradiography.
In some experiments, vitronectin (200 ng/ml) was incubated with
125I-IGFBP-5 and PAI-1 then the PAI-1 immunoprecipitated as
described previously. Similarly, in some experiments antivitronectin
antiserum (1:5000 dilution) was added with PAI-1 or vitronectin and the
complexes immunoprecipitated. The amount of 125I-IGFBP-5
that was immunoprecipitated was determined by SDS-PAGE with
autoradiography.
Measurement of binding affinity
To determine the affinity of the PAI-1/IGFBP-5 interaction,
increasing concentrations of unlabeled IGFBP-5 (10–1000 ng/ml) were
mixed with 125I-IGFBP-5 (50,000 cpm/ml), and a constant
amount of unlabeled PAI-1 (100 ng/ml). After an overnight incubation,
the anti-PAI-1 antiserum was added (1:5000 dilution), and the bound
125I-IGFBP-5 was coimmunoprecipitated with protein A
sepharose as previously described, then counted in a gamma
spectrometer.
IGFBP-5 proteolysis
The effect of PAI-1 and vitronectin on IGFBP-5 proteolysis was
determined by incubating IGFBP-5 with an IGFBP-5 protease that had been
purified by heparin sepharose chromatography from fibroblast
conditioned medium (12). The partially purified protein was added to
0.25 ml of 0.05 M Tris, pH 7.0, with IGFBP-5 (50 ng/ml) and
vitronectin (10 µg/ml) or PAI-1 (10 µg/ml). After 14 h at 37
C, the products of the reaction were analyzed by immunoblotting.
Preparation of synthetic peptides and IGFBP-5 mutants
Synthetic peptides containing 11–20 amino acids from four
different regions of IGFBP-5 were synthesized and purified as described
previously (10). Their sequences are as follows: A,
DRKGFYKRKQCKPSRGRKR; B, AVKKDRRKKLT; C, ALLHGRGVCLNEKS; D,
RPKHTRISELKAE. To determine their capacity to interfere with IGFBP-5
binding to PAI-1, they were added using concentrations between 0.1 and
5.0 µg/ml. IGFBP-5 mutants containing the substitutions for basic
amino acids were prepared using in vitro mutagenesis as
previously described (9, 10). They were purified to homogeneity and
their protein content determined by comparing their HPLC peak areas to
an IGFBP-5 standard whose concentration had been determined by amino
acid composition analysis (9). They were each shown to have an affinity
for IGF-I that was similar to the wild-type, nonmutated protein
(10).
|
Results
|
|---|
Analysis of the proteins that were purified from fibroblast
conditioned medium (using the three purification steps outlined in
Materials and Methods) by SDS-PAGE and Ponceau-S staining
showed that a 55-kDa protein had been eluted from the IGFBP-5 peptide
affinity column. When this band was sequenced, a single sequence that
corresponded to positions 1–28 of human plasminogen activator
inhibitor-1 was obtained. Therefore, coimmunoprecipitation studies were
undertaken to determine if PAI-I could bind to IGFBP-5 at lower
concentrations.
Coincubation of PAI-1 (100 ng/ml) and IGFBP-5 with anti-PAI-1 resulted
in coimmunoprecipitation of IGFBP-5 (Fig. 1A
). To prove
that coimmunoprecipitation of the IGFBP-5/PAI-1 complex was specific,
anti-PAI-1 antibody and protein-A sepharose were added to tubes that
received IGFBP-5 alone. This resulted in minimal precipitation of
IGFBP-5. To analyze this interaction by a more sensitive method,
125I-IGFBP-5 was used as a ligand. 125I-IGFBP-5
was incubated with unlabeled PAI-1 and anti-PAI-1 antibody. As shown in
Fig. 1B, the addition of unlabeled PAI-1 resulted in a large increase
in the amount of 125I-IGFBP-5 that was coimmunoprecipitated
when compared with the sample that did not contain PAI-1. The addition
of excess unlabeled IGFBP-5 greatly reduced the amount of
125I-IGFBP-5 that was coimmunoprecipitated, indicating that
the immunoprecipitation was specific. Similarly, if anti-PAI-1 was
omitted, minimal 125I-IGFBP-5 was precipitated by protein A
sepharose.
View larger version (33K):
[in this window]
[in a new window]
|
Figure 1. Coimmunoprecipitation of IGFBP-5 and PAI-1. A,
IGFBP-5, 50 ng/ml, was incubated with PAI-1, 100 ng/ml, and anti-PAI-1
antibody, 1:5000 dilution, for 16 h at 4 C. At that time,
immunoprecipitation was conducted by adding protein A sepharose, as
described in Materials and Methods. The pellets were
dissolved in Laemmli sample buffer and the products electrophoresed
through a 12.5% gel, then transferred to Immobilon membranes and
immunoblotted using a 1:1000 dilution of anti-IGFBP-5 antiserum. The
results show that, in the presence of PAI-1, IGFBP-5 could be
coimmunoprecipitated, whereas the anti-PAI-1 antibody did not
immunoprecipitate IGFBP-5 without PAI-1 being added. PAI-1 did not
react with the IGFBP-5 antibody. B, 125I-IGFBP-5, 50,000
cpm/ml, was incubated with anti-PAI-1 antiserum (1:5000) and PAI-1 (100
ng/ml). In the presence of PAI-1, an abundant 125I-IGFBP-5
band was detected. In the absence of the PAI-1 antibody or unlabeled
PAI-1, minimal 125I-IGFBP-5 was immunoprecipitated. The
asterisk denotes an incubation mixture that contained
excess unlabeled IGFBP-5 (100 ng/ml).
|
|
To determine whether this interaction was influenced by
glycosaminoglycans, which are known to bind to IGFBP-5, soluble GAGs
were added with PAI-1 and 125I-IGFBP-5, then
coimmunoprecipitation performed. Heparin and heparan sulfate completely
inhibited the coimmunoprecipitation reaction (Fig. 2).
Chondroitin sulfate A and B had some effect but were less potent.
Because charged glycosaminoglycans were shown to be important for this
interaction, the effect of increasing concentrations of heparin on the
IGFBP-5/PAI-1 interaction was determined. Heparin concentrations of 1
and 10 µg/ml resulted in complete inhibition of the IGFBP-5 binding
to PAI-1 (data not shown). Heparan sulfate and heparin contain O-linked
sulfate groups in the 2 or 3 position of the iduronic acid ring.
Chondroitin sulfate B (dermatan sulfate) that had an intermediate
effect also contains sulfate groups in the same positions, but it
contains fewer sulfate groups per molecule (Fig. 2). In contrast,
chondroitin sulfate A and C do not have O-linked sulfates in the 2 or 3
position and they had significantly reduced effects on the
PAI-1/IGFBP-5 binding. This suggests that the presence of O-linked
sulfate groups in these specific positions of the glycosaminoglycan
structure is required for maximal inhibition of binding.
View larger version (45K):
[in this window]
[in a new window]
|
Figure 2. The effects of glycosaminoglycans on IGFBP-5/PAI-1
association. 125I-IGFBP-5 was incubated without, lane 1, or
with 100 ng/ml of PAI-1 (lanes 2–7), then immunoprecipitated as
described in Materials and Methods. After
immunoprecipitation, the pellets were dissolved in Laemmli sample
buffer and the products electrophoresed, then transferred to Immobilon
membranes and autoradiography performed directly; lanes 2–7,
125I-IGFBP-5; lanes 3–7, 125I-IGFBP-5 plus
various glycosaminoglycans (0.1 µg/ml). Lane 3, heparin; lane 4,
heparan sulfate; lane 5, chondroitin sulfate A; lane 6, chondroitin
sulfate B; lane 7, chondroitin sulfate C. The radiolabeled band shown
in lane 1 that migrates with an Mr estimate of 16 kDa is a
fragment of radiolabeled IGFBP-5.
|
|
Because the binding was inhibited by glycosaminoglycans, this suggested
that a region of charged amino acids within IGFBP-5 might be required
for binding. To test this hypothesis, synthetic peptides containing
regions of IGFBP-5 sequence with several charged amino acids were
incubated with 125I-IGFBP-5 and PAI-1 and
coimmunoprecipitated. As shown in Fig. 3, peptide A,
which contained the IGFBP-5 region from amino acid 201 to 218, was an
effective competitive inhibitor. Peptide B, whose amino acid sequence
contains a similar number of charged residues from a different region
of the IGFBP-5 molecule (10), had a reduced effect. The effect of
peptide C was variable and in some experiments, it was a much less
effective inhibitor of coimmunoprecipitation (data not shown). Peptide
D had no effect on PAI-1 binding. Both peptides C and D contain fewer
charged residues than peptides A or B.
View larger version (78K):
[in this window]
[in a new window]
|
Figure 3. The effects of IGFBP-5 peptides on PAI-1 and
IGFBP-5 binding. Radiolabeled IGFBP-5, 50,000 cpm/ml, was incubated
with unlabelled PAI-1, 100 ng/ml, and 100 µg/ml of one of four
peptides, lanes 2–5, that contained various sequences within the
IGFBP-5 molecule (9 ). After a 14-h incubation, the products were
coimmunoprecipitated and processed by SDS-PAGE and transferred to
Immobilon. The figure is an autoradiogram showing the amount of
125I-IGFBP-5 that was coimmunoprecipitated. Lanes 1–5,
125I-IGFBP-5; lane 2, peptide A; lane 3, peptide B; lane 4,
peptide C; lane 5, peptide D. The radiolabeled band with an
Mr estimated of 18 kDa shown in lanes 1, 4, and 5 is a
fragment of radiolabeled IGFBP-5.
|
|
To further determine the specific amino acids that were involved in the
PAI-1/IGFBP-5 interaction, coprecipitation was conducted using IGFBP-5
mutants that contain neutral substitutions for basic amino acids
between residues 201 and 218. A form of IGFBP-5 that contained four
substitutions K211N, R214A, K217A, and R218A had a 24% reduction in
its ability to bind to PAI-1 (Fig. 4). A form that
contained three substitutions, R202A, K206A, and R207A, also had
decreased binding. A form that contained four substitutions at
positions R201A, K20N, K206N, and K208N had a 71% reduction in
detectable PAI-1 binding. In contrast, two other mutants that contained
substitutions for charged residues in the region of IGFBP-5 between
positions 201 and 218 and K134A/R136A had nearly normal PAI-1 binding.
This suggests that both the positional location and the number of
charged residues that are altered determine binding affinity. In
general, the mutant forms that had reduced binding for
glycosaminoglycans (9) had some alteration in PAI-1 binding, but there
were distinct differences. For example, the R201A, K202N, K206N, K208N
mutant has only minimally reduced GAG binding (9). This suggests that
the GAG binding site and the PAI-1 binding site, while both charge
dependent, have different configurations.
View larger version (68K):
[in this window]
[in a new window]
|
Figure 4. Coimmunoprecipitation of IGFBP-5 mutants and
PAI-1. One hundred nanograms per milliliter of each IGFBP-5 mutant was
incubated with 100 ng/ml of PAI-1 and the products immunoprecipitated
using anti-PAI-1 antiserum as described previously. After SDS-PAGE, the
products were transferred to an Immobilon membrane, then ligand blotted
using 125I-IGF-I as described in methods. Lane 1, wild-type
IGFBP-5, no PAI-1; lanes 2–8, PAI-1 plus IGFBP-5 or mutants. Lane 2,
wild-type IGFBP-5; lane 3, K134A/R136A; lane 4, R202A/K206A/R207A; lane
5, R201A/K202N/K206N/K208N; lane 6, K211N; lane 7, R201N/K202N; lane 8,
K211N/R214A/K217A/R218A. The results of Phor Image analysis expressed
as scanning units were: lane 1, 2,237; lane 2, 190,511; lane 3,
204,125; lane 4, 170,867; lane 5, 76,748; lane 6, 250,813; lane 7,
223,452; lane 8, 146,656.
|
|
The effect of PAI-1 binding to IGFBP-5 on its susceptibility to
proteolytic cleavage was determined. When IGFBP-5 was incubated with an
IGFBP-5 protease obtained from fibroblast conditioned medium, there was
time-dependent cleavage of the substrate into a 22-kDa band. The
addition of PAI-1 to the incubation buffer resulted in a 1.6-fold
increase in intact IGFBP-5 band intensity (Fig. 5). The
addition of vitronectin (an ECM protein that stabilizes PAI-1) with
PAI-1 did not result in additional inhibition of IGFBP-5 proteolysis.
Neither tissue plasminogen activator (TPA) or urokinase can
significantly degrade IGFBP-5; therefore, it is unlikely that PAI-1 is
acting to directly inhibit a TPA-like protease. The binding of
125I-IGF-I to IGFBP-5 was determined in the presence and
absence of PAI-1 and the results analyzed by the method of Scatchard:
the addition of PAI-1 had no effect on the affinity of IGFBP-5 for
IGF-I (data not shown).
View larger version (85K):
[in this window]
[in a new window]
|
Figure 5. The effect of PAI-1 on IGFBP-5 proteolysis.
IGFBP-5, 50 ng/ml, was incubated with IGFBP-5 protease that had been
purified from human fibroblast conditioned medium. After a 14-h
incubation, the products were electrophoresed and analyzed by
immunoblotting. Lanes 1–4, IGFBP-5 plus protease; lane 2, vitronectin,
10 µg/ml; lane 3, PAI-1, 10 µg/ml; lane 4, PAI-1 plus vitronectin.
The 22-kDa band, which is the major cleavage product of this protease,
is shown by an arrow. Scanning unit values for the
intact IGFBP-5 bands were; lane 1, 87,644; lane 2, 107,781; lane 3,
144,811; and lane 4, 123,322.
|
|
Because multiple extracellular matrix proteins may bind to IGFBP-5
simultaneously, we determined if vitronectin, a known binding protein
for PAI-1, could alter the PAI-1/IGFBP-5 interaction. As shown in Fig. 6, when coimmunoprecipitation was performed using
anti-PAI-1 antibodies, increasing concentrations of vitronectin
inhibited the amount of IGFBP-5 that could be coimmunoprecipitated with
PAI-1 by 33% (Fig. 6, compare lane 2 to lane 3). This suggests that
vitronectin is capable of binding to PAI-1 and preventing the
IGFBP-5/PAI-1 complex formation. When the experiment was repeated with
antivitronectin antibodies, vitronectin and 125I-IGFBP5
coprecipitated (Fig. 6, panel B). PAI-1 (100 ng/ml) interfered with the
interaction and reduced the amount of 125I-IGFBP-5 that
could be coprecipitated to basal levels.
View larger version (38K):
[in this window]
[in a new window]
|
Figure 6. The effect of vitronectin on the binding of PAI-1
to IGFBP-5. A, 125I-IGFBP-5 and PAI-1 (100 ng/ml) were
incubated with anti-PAI-1 antisera and increasing concentrations of
vitronectin. After a 6-h incubation, the products were
immunoprecipitated, electrophoresed, and the gel analyzed by
autoradiography. Lane 1, 125I-IGFBP-5 alone; lanes 2–5,
125I-IGFBP-5 plus PAI-1; lane 3, vitronectin (1 µg/ml);
lane 4, vitronectin (0.1 µg/ml); lane 5, vitronectin (0.01 µg/ml).
Phosphor Image intensity units were: lane 1, 46,673; lane 2, 98160;
lane 3, 65,869; lane 4, 75077; and lane 5, 81,264. B,
125I-IGFBP-5 and vitronectin (100 ng/ml) were incubated
with antivitronectin antiserum (1:5000), then the products
immunoprecipitated as in panel A. Lane 1, 125I-IGFBP-5
alone; lane 2, vitronectin (100 ng/ml); lane 3, PAI-1 (100 ng/ml); lane
4, PAI-1 (1000 ng/ml). Phosphor Image intensity units were lane 1,
31,342; lane 2, 86,192; lane 3, 44,360; and lane 4, 36,444.
|
|
To assess the physiological significance of PAI-1/IGFBP-5 binding,
extracellular matrix was prepared from human fibroblast cultures as
previously described (18). 125I-IGFBP-5 was incubated
overnight with the ECM extracts and increasing concentrations of
anti-PAI-1 antibody. This antibody recognizes ECM bound PAI-1. The
addition of increasing amounts of anti-PAI-1 to the incubation mixture
partially blocked ability of 125I-IGFBP-5 to bind to the
ECM (Table 1). In contrast, an equal concentration
nonimmune IgG had no effect. Approximately 27% of the total
125I-IGFBP-5 binding could be inhibited by this antibody.
This suggests that one of the components of ECM that binds to IGFBP-5
is PAI-1. However, because glycosaminoglycans inhibit IGFBP-5 binding
to PAI-1 and are abundant in ECM, they may be a more important binding
determinant. To determine the affinity of IGFBP-5 for PAI-1,
competitive binding assays using unlabeled and 125I-IGFBP-5
were performed. As shown in Fig. 7, the results are
consistent with a single site model with relatively high affinity. When
the data from three experiments were analyzed by the method of
Scatchard, the affinity (Kd) was determined to be 9.1
x 10-8 M.
View larger version (15K):
[in this window]
[in a new window]
|
Figure 7. Scatchard plot. 125I-IGFBP-5 (30,000
cpm/tube) was incubated with increasing concentrations of unlabeled
IGFBP-5 (5–1000 ng/ml) and PAI-1, 100 ng/ml. The products were
immunoprecipitated as described in Materials and Methods
and bound 125I-IGFBP-5 determined directly by gamma
counting. Nonspecific binding was defined as the cpm bound in the
presence of 10 µg/ml of unlabeled IGFBP-5 and was less than 4% of
the total binding.
|
|
|
Discussion
|
|---|
IGFBP-5 is a potent modulator of IGF-I actions, and the factors
that control its abundance in the pericellular environment may be
important determinants of how this growth factor functions to modify
cellular responses. The PAI-1/IGFBP-5 interaction may be particularly
relevant to controlling IGFBP-5 abundance in the ECM. In this paper, we
demonstrate that IGFBP-5 binds to PAI-1 and that the binding
interaction has a relatively high affinity. An affinity constant in the
range of 10-9 M was estimated by Scatchard
analysis. Additionally, we show that anti-PAI-1 antiserum inhibits
IGFBP-5 binding to ECM. This finding suggests that PAI-1 may be one of
the major binding sites for IGFBP-5 within the extracellular matrix.
However, anti-PAI-1 antiserum caused only a 27% reduction in PAI-1
binding that supports the conclusion that other ECM components are also
important determinants of IGFBP-5 binding. Because glycosaminoglycans
can inhibit the IGFBP-5-PAI-1 interaction, they may represent the
predominant ECM binding component. Because PAI-1 is synthesized by
multiple connective tissue cell types, including cell types that are
involved in the response to injury, such as fibroblasts and smooth
muscle cells, the PAI-1-IGFBP-5 interaction may be a determinant of the
amounts of IGFBP-5 and IGF-I that are localized in ECM and are
available to stimulate the responses of these cells during the repair
process.
Previously, we have shown that the amount of IGFBP-5 that is localized
in the extracellular matrix positively modulates both the cell growth
and DNA synthesis responses to IGF-I(5). Specifically, the mitogenic
response of fibroblasts to IGF-I is enhanced if increased amounts of
IGFBP-5 are present in the ECM. Furthermore, ECM localization protects
IGFBP-5 from proteolysis, thus resulting in increased IGF-I being
present at this site (7). A further understanding of why binding of
IGF-I to IGFBP-5 in the ECM results in enhancement rather than
inhibition of its action was made clearer by the observation that,
after ECM localization of this binding protein, its affinity for IGF-I
is reduced 8-fold (5, 9). Because the affinity of non-ECM associated
IGFBP-5 is 20- to 30-fold greater than the affinity of the type I IGF
receptor, this reduction in affinity would allow IGF-I that is
localized bound to ECM to be in a more favorable equilibrium with
receptors on cell surfaces. Therefore, ECM localization of IGFBP-5 and
IGF-I results in protection from proteolysis and maintenance of a
reservoir of IGF-I that can be released to receptors at critical time
points during cell cycle progression. Because binding to PAI-1 does not
lower IGFBP-5 affinity for IGF-I, this suggests that there may be
differences in the affinity of ECM-associated IGFBP-5 and that the net
reduction in affinity may be dependent upon whether binding is to
predominantly glycosaminoglycans or to other proteins such as PAI-1.
Therefore, the factors that regulate the distribution of IGFBP-5 among
the various ECM binding proteins may indirectly alter IGFBP-5 affinity
and thereby modulate IGF-I action.
The variables that regulate the synthesis of IGFBP-5 are also important
determinants of its abundance in ECM. We have recently shown that IGF-I
stimulates IGFBP-5 transcription by smooth muscle cells (19), and Dong
and Canalis have confirmed this finding in osteoblasts (20).
Furthermore, they demonstrated that retinoic acid is also a potent
stimulant of IGFBP-5 synthesis by bone cells and results in increased
amounts of IGFBP-5 in the ECM (20). Hakeda et al. (21) have
shown that prostaglandin-E and PTH stimulate IGFBP-5 release from
osteoblast ECM and this could modulate IGF-I levels in the pericellular
environment.
In addition to the factors that increase IGFBP-5 synthesis, we have
shown that several connective tissue cell types release an IGFBP-5
protease into their conditioned medium (7, 12, 19). The factors that
regulate the activity of this protease may be important determinants of
IGFBP-5 abundance in interstitial fluids and in the ECM. These studies
show that PAI-1 binding to IGFBP-5 alters its susceptibility to
proteolysis in the interstitial fluid. However, because multiple
substances, such as PAI-1 (9) and vitronectin may bind to IGFBP-5 in
ECM, other ECM proteins may function coordinately with PAI-1 or
vitronectin to stabilize IGFBP-5. The synthesis of PAI-1 may also be
increased coordinately with IGFBP-5. Specifically, Padaytly et
al. and Afossa et al. have shown that IGF-I increases
PAI-1 synthesis, both in vivo and in vitro;
therefore, it is possible that PAI-1 and IGFBP-5 synthesis are
coordinately regulated by IGF-I in selected cell types (22, 23).
Conversely, factors such as thrombin (24) that cause release of PAI-1
from ECM might function to decrease ECM-associated IGFBP-5. Heparin
functions to inhibit the IGFBP-5/PAI-1 interaction and to reduce the
abundance of PAI-1 or IGFBP-5 within ECM (25, 26, 27); therefore,
heparin inhibition of IGFBP-5 binding to ECM could be a component of
the mechanism by which heparin functions to inhibit smooth muscle cell
replication.
IGF-I synthesis and peptide abundance is also often increased at sites
of injury (28, 29). Similarly PAI-1 expression is increased after
injury. The 5' flanking sequences of both the IGFBP-5 and PAI-1 genes
contain the Egr-1 injury response element (30), which mediates an
increase in the transcription of specific genes after injury.
Therefore, it is possible that PAI-1 and IGFBP-5 synthesis may be
directly increased in response to injury. Because IGF-I synthesis also
increases after injury, this increase in IGF-I could further enhance
IGFBP-5 expression by a mechanism that is distinct from Egr-1 (19).
PAI-1 is a component of smooth muscle cell extracellular matrix. It is
also synthesized by endothelium in response to injury; therefore,
because it binds tightly to IGFBP-5, it could provide a mechanism for
stabilizing large amounts of IGFBP-5 in the newly developing ECM
during neointima formation. PAI-1 also has other interesting properties
because it binds vitronectin (31, 32) and the PAI-1 binding domain is
also the region of vitronectin that interacts with the vitronectin
receptor (31). This suggests other potential mechanisms for enhancing
IGF-I action. Proteases, such as thrombin, could release vitronectin
from the ECM and allow more vitronectin to interact with its receptor.
Similarly, if they degrade IGFBP-5, they would allow release of IGF-I
to receptors. Simultaneous vitronectin receptor and IGF receptor
occupancy has been shown to be a potent stimulant of smooth muscle cell
migration (33), a process that is an important component of neointina
formation.
The exact site of molecular interaction between PAI-1 and IGFBP-5
remains to be determined. However, it is clear that ionic residues are
involved. Similarly, glycosaminoglycans may interfere with this
interaction, and there may be a competitive equilibrium between
proteoglycans that are present in the extracellular matrix and PAI-1.
Because plasminogen is not the major protease in pSMC or fibroblast
medium that degrades IGFBP-5, it is not surprising that PAI-1 binding
of IGFBP-5 did not completely protect it from proteolysis. However,
plasmin can degrade IGFBP-5, so PAI-1 could play an indirect role in
maintaining intact IGFBP-5 by inhibiting plasmin formation (34). The
charged residue motif that binds PAI-1 is clearly distinct from the
heparin binding site (9), although overlapping residues are used. This
suggests that the highly charged regions of IGFBP-5 that are surface
exposed recognize distinct substrates differently and that there may be
specificity for substrate binding. The exact factors that control the
surface exposure of distinct sites in IGFBP-5 are unknown but will be
interesting points for future study.
THE INSULIN-LIKE growth factors (IGFs) are
present in physiological fluids bound to IGF binding proteins (IGFBPs)
(1, 2). Six IGF binding proteins with high affinities for IGF-I and II
have been identified (3). Although IGFBP-5 binds to the IGFs with high
affinity in interstitial fluids, when it is associated with
extracellular matrix its affinity is decreased 8- to 15-fold (4). This
lower affinity results in better equilibration of IGF-I with its
receptor and potentiation of IGF-I stimulated fibroblast (5) or smooth
muscle cell growth (6). Extracellular fluids also contain proteases
that degrade IGFBP-5 and cleave it into non-IGF binding fragments (7).
However, when IGFBP-5 is associated with ECM, it is protected from
proteolysis (5, 8); therefore, ECM binding may provide a mechanism for
localizing IGF-I in focal areas and provide better access to receptors.
This localization could be important in mediating the cellular response
to injury when IGF-I expression is increased in the local
microenvironment. Therefore, the factors in extracellular matrix that
account for IGFBP-5 localization have the potential to indirectly
regulate IGF-I actions.
We have previously shown that IGFBP-5 binds to heparin and heparan
sulfate (4, 9). Similarly, glycosaminoglycan containing proteoglycans,
such as tenascin, that are present in the ECM also bind IGFBP-5 (9).
This binding occurs through highly charged residues in IGFBP-5 that are
located between residues 201 and 218 (4, 9, 10). Similar to ECM
binding, glycosaminoglycan binding of IGFBP-5 also results in a
reduction in the affinity of IGFBP-5 for IGF-I (4) and in protection
from proteolysis (8). However, competitive binding studies using intact
ECM have shown that all of the ECM binding of IGFBP-5 could not be
accounted for by binding to proteoglycans (5). Therefore, these studies
were undertaken to determine if extracellular matrix proteins other
than proteoglycans might bind IGFBP-5.
|
Materials and Methods
|
|---|
Materials
Eagle’s MEM was purchased from Grand Island Biological Company
(Grand Island, NY, Life Technologies). Calf serum was purchased from
Colorado Serum (Colorado Laboratories, Denver, CO). BSA, ammonium
persulfate, sodium phosphate were obtained from Sigma Chemical Co. (St.
Louis, MO). 125I IGFBP-5 (30–45 µCi/µg) was prepared
using IODO-BEADS (Pierce Chemical Co., Rockford, IL) and 15 µg
IGFBP-5 according to manufacturer’s instructions (7, 11). Recombinant
human IGF-I was obtained from Bachem (Torrence, CA). Tris-sodium
duodecyl sulfate, chloramine-T, polyacrylamide, and prestained mol wt
standards were purchased from Bethesda Research Laboratories (BRL)
(Gaithersburg, MD). Tween 80 was obtained from Fisher Scientific
(Cleartown, NJ). Human vitronectin and antihuman vitronectin antiserum
were obtained from Telios (San Diego, CA).
Purification of PAI-I
Human dermal fibroblasts (GM-10) were purchased from Coriell
Institute (Camden, NJ). They were grown in EMEM supplemented with 10%
calf serum as described previously (7). Confluent monolayers were
washed twice in the serum-free medium, then 50 cc was added to
175-cm2 flasks (Falcon Labware, Division of
Becton-Dickinson), and the medium was collected after 24 h. This
was repeated until 1.2 liters of serum-free medium had been collected.
After purification by heparin sepharose affinity chromatography and
1 antichymotryptic peptide affinity chromatography as
described previously (12), the fractions containing IGFBP-5 protease
activity were applied to an IGFBP-5 affinity column. This column was
prepared by conjugating a synthetic 18 amino acid peptide that
contained the IGFBP-5 sequence between amino acids 201–218 to Affigel
10 (Bio-Rad Richmond, CA). This sequence is also present in IGFBP-3,
but it is not present in any other form of IGFBP. The peptide was
conjugated according to manufacturer’s instructions by incubating 5.6
mg of peptide with 7.0 cc of activated sepharose in 0.1 M
Na HEPES buffer, pH 7.5, for 4 h at 8 C. Unreacted sites were
blocked by a further incubation for 1 h with 0.25 M
Tris. The gel was washed, and the pool of activity that had been
purified from fibroblast conditioned medium was applied. The purified
active fractions that were eluted with 0.5 M NaCl were
concentrated and further analyzed by SDS-PAGE (12.5% gel). After
electrophoresis, the proteins were transferred to a nitrocellulose
membrane that was stained with Ponceau-S. A 55-kDa band was excised and
subjected to amino acid sequencing. Amino terminal amino acid
sequencing was conducted as described previously using Edman
degradation (13).
Preparation of human IGFBP-5
The human IGFBP-5 that was used in these studies was obtained by
purifying the protein from Chinese hamster ovary (CHO) cell conditioned
medium. CHO cells were transfected with an IGFBP-5 complementary DNA
(cDNA) that contained the entire protein coding region and had been
prepared as previously described (7). The cells were transfected, then
maintained in methotrexate (50 µg/ml) because the plasmid that was
used (pNUT) contains the gene that contains dihydrofolate reductase
resistance. After selection with methotrexate, 50 µg/ml, the highest
secreting clones were further subcultured in DMEM containing 10%
dialyzed FCS. The cells were grown to confluency and 50 ml of
serum-free medium were added to 175-cm tissue culture flasks and
incubated for 48 h. Eight hundred cubic centimeters of conditioned
medium were purified using a three-step purification procedure of
phenyl sepharose chromatography, IGF-I affinity chromatography, and
HPLC (C4 column) (Vydac, Hesperia, CA), as previously described (10).
The purified protein was homogenous by SDS-PAGE with silver staining
(14). The protein content was determined by amino acid composition
analysis (7).
Coimmunoprecipitation, ligand blotting, and immunoblotting
The ability of IGFBP-5 to bind to the plasminogen activator
inhibitor-1 was analyzed by coimmunoprecipitation. It was determined
that the lowest nonspecific binding and the highest sensitivity of
detection could be obtained by immunoprecipitating with anti-PAI-1
antisera. The IGFBP-5 that had been coimmunoprecipitated was then
detected by immunoblotting (5) or by ligand blotting using
125I-IGF-I (15). PAI-1 and polyclonal anti-PAI-1 antiserum
were obtained from American Diagnostica (Greenwich, CT). In most
experiments, IGFBP-5 (100 ng/ml) was incubated with PAI-1 at 100 ng/ml,
and anti-PAI-1 was added at a 1:5000 dilution. The reagents were
allowed to incubate for 16 h at 4 C in 0.03 M sodium
phosphate, pH 7.4, containing 0.01 M EDTA, 0.1% Tween 80,
and 0.02% BSA. Protein A sepharose (3 mg) was added and incubated for
2 h at 4 C. The immune complexes were precipitated by centrifuging
at 3000 x g for 5 min. The pellets were rinsed with
1.0 ml of the buffer listed previously and 50 µl of 2x Laemmli
sample buffer was added. The sample was centrifuged (2000 x
g for 5 min) and the supernatant loaded onto a 12.5% SDS
polyacrylamide gel. The proteins were electrophoresed through a 12.5%
gel using nonreducing conditions, then transferred to Immobilon
membranes (Millipore, Bedford, MA) as previously described (15). After
transfer, the filters were immunoblotted using a 1:1000 dilution of
anti-IGFBP-5 antiserum. The immune complexes were detected using the
Protoblot system (Promega, Madison, WI) according to manufacturer’s
recommendation. The electrophoretic mobilities of the bands that were
detected were compared with known mol wt standards (Bethesda Research
Laboratories, Gaithersburg, MD). The photographic negatives were
scanned using a Hoffer GS-300 scanner (Hoffer Scientific Instruments,
San Francisco, CA), and the results are expressed in scanning units.
For more sensitive detection of IGFBP-5, ligand blotting was performed.
Five hundred thousand counts per minute of 125I-IGF-I [150
µCi/µg (16)] was added to the filters and incubated overnight
using conditions described previously (17). The filters were washed,
and autoradiography was used to detect bound 125I-IGF-I.
For some experiments, quantitation of the amount of IGFBP-5 was
determined by analysis autoradiography band intensity using a Phor
Imager and Image Quant SF software (Molecular Dynamics, Sunnyvale, CA).
The results are expressed as arbitrarily defined scanning units.
To enhance the sensitivity detection of IGFBP-5 after
coimmunoprecipitation, 125I-IGFBP-5 (30 µCi/µg)
(260,000 cpm/ml) that had been prepared as described previously (12),
was incubated and was used with PAI-1 (200 ng/ml), and a 1:10,000
dilution of anti-PAI-1 antiserum in 0.25 ml of 0.03 M
sodium phosphate containing 0.01 M EDTA, 0.1% Tween 80,
0.1% BSA, pH 7.0. After an overnight incubation at 4 C, the immune
complexes were precipitated by adding rabbit IgG (1.0 µl) and
protein-A sepharose (10 µl of a 20 mg/ml solution) then centrifuged.
In some experiments, duplicate tubes contained unlabeled IGFBP-5 or
IGFBP-5 mutants (100 ng/ml). Similarly, in some experiments unlabeled
IGFBP-5 peptides that had been prepared as described previously (10)
were used. Fifty microliters of Lammeli sample buffer were added, and
the samples were heated to 65 C then electrophoresed through a 12.5%
SDS gel and transferred to Immobilon filters. The precipitated
125I-IGFBP-5 was detected by autoradiography. In some
experiments, vitronectin (200 ng/ml) was incubated with
125I-IGFBP-5 and PAI-1 then the PAI-1 immunoprecipitated as
described previously. Similarly, in some experiments antivitronectin
antiserum (1:5000 dilution) was added with PAI-1 or vitronectin and the
complexes immunoprecipitated. The amount of 125I-IGFBP-5
that was immunoprecipitated was determined by SDS-PAGE with
autoradiography.
Measurement of binding affinity
To determine the affinity of the PAI-1/IGFBP-5 interaction,
increasing concentrations of unlabeled IGFBP-5 (10–1000 ng/ml) were
mixed with 125I-IGFBP-5 (50,000 cpm/ml), and a constant
amount of unlabeled PAI-1 (100 ng/ml). After an overnight incubation,
the anti-PAI-1 antiserum was added (1:5000 dilution), and the bound
125I-IGFBP-5 was coimmunoprecipitated with protein A
sepharose as previously described, then counted in a gamma
spectrometer.
IGFBP-5 proteolysis
The effect of PAI-1 and vitronectin on IGFBP-5 proteolysis was
determined by incubating IGFBP-5 with an IGFBP-5 protease that had been
purified by heparin sepharose chromatography from fibroblast
conditioned medium (12). The partially purified protein was added to
0.25 ml of 0.05 M Tris, pH 7.0, with IGFBP-5 (50 ng/ml) and
vitronectin (10 µg/ml) or PAI-1 (10 µg/ml). After 14 h at 37
C, the products of the reaction were analyzed by immunoblotting.
Preparation of synthetic peptides and IGFBP-5 mutants
Synthetic peptides containing 11–20 amino acids from four
different regions of IGFBP-5 were synthesized and purified as described
previously (10). Their sequences are as follows: A,
DRKGFYKRKQCKPSRGRKR; B, AVKKDRRKKLT; C, ALLHGRGVCLNEKS; D,
RPKHTRISELKAE. To determine their capacity to interfere with IGFBP-5
binding to PAI-1, they were added using concentrations between 0.1 and
5.0 µg/ml. IGFBP-5 mutants containing the substitutions for basic
amino acids were prepared using in vitro mutagenesis as
previously described (9, 10). They were purified to homogeneity and
their protein content determined by comparing their HPLC peak areas to
an IGFBP-5 standard whose concentration had been determined by amino
acid composition analysis (9). They were each shown to have an affinity
for IGF-I that was similar to the wild-type, nonmutated protein
(10).
|
Results
|
|---|
Analysis of the proteins that were purified from fibroblast
conditioned medium (using the three purification steps outlined in
Materials and Methods) by SDS-PAGE and Ponceau-S staining
showed that a 55-kDa protein had been eluted from the IGFBP-5 peptide
affinity column. When this band was sequenced, a single sequence that
corresponded to positions 1–28 of human plasminogen activator
inhibitor-1 was obtained. Therefore, coimmunoprecipitation studies were
undertaken to determine if PAI-I could bind to IGFBP-5 at lower
concentrations.
Coincubation of PAI-1 (100 ng/ml) and IGFBP-5 with anti-PAI-1 resulted
in coimmunoprecipitation of IGFBP-5 (Fig. 1A
). To prove that coimmunoprecipitation
of the IGFBP-5/PAI-1 complex was specific, anti-PAI-1 antibody and
protein-A sepharose were added to tubes that received IGFBP-5 alone.
This resulted in minimal precipitation of IGFBP-5. To analyze this
interaction by a more sensitive method, 125I-IGFBP-5 was
used as a ligand. 125I-IGFBP-5 was incubated with unlabeled
PAI-1 and anti-PAI-1 antibody. As shown in Fig. 1B, the addition of
unlabeled PAI-1 resulted in a large increase in the amount of
125I-IGFBP-5 that was coimmunoprecipitated when compared
with the sample that did not contain PAI-1. The addition of excess
unlabeled IGFBP-5 greatly reduced the amount of
125I-IGFBP-5 that was coimmunoprecipitated, indicating that
the immunoprecipitation was specific. Similarly, if anti-PAI-1 was
omitted, minimal 125I-IGFBP-5 was precipitated by protein A
sepharose.
To determine whether this interaction was influenced by
glycosaminoglycans, which are known to bind to IGFBP-5, soluble GAGs
were added with PAI-1 and 125I-IGFBP-5, then
coimmunoprecipitation performed. Heparin and heparan sulfate completely
inhibited the coimmunoprecipitation reaction (Fig. 2). Chondroitin sulfate A and B had some
effect but were less potent. Because charged glycosaminoglycans were
shown to be important for this interaction, the effect of increasing
concentrations of heparin on the IGFBP-5/PAI-1 interaction was
determined. Heparin concentrations of 1 and 10 µg/ml resulted in
complete inhibition of the IGFBP-5 binding to PAI-1 (data not shown).
Heparan sulfate and heparin contain O-linked sulfate groups in the 2 or
3 position of the iduronic acid ring. Chondroitin sulfate B (dermatan
sulfate) that had an intermediate effect also contains sulfate groups
in the same positions, but it contains fewer sulfate groups per
molecule (Fig. 2). In contrast, chondroitin sulfate A and C do not have
O-linked sulfates in the 2 or 3 position and they had significantly
reduced effects on the PAI-1/IGFBP-5 binding. This suggests that the
presence of O-linked sulfate groups in these specific positions of the
glycosaminoglycan structure is required for maximal inhibition of
binding.
Because the binding was inhibited by glycosaminoglycans, this suggested
that a region of charged amino acids within IGFBP-5 might be required
for binding. To test this hypothesis, synthetic peptides containing
regions of IGFBP-5 sequence with several charged amino acids were
incubated with 125I-IGFBP-5 and PAI-1 and
coimmunoprecipitated. As shown in Fig. 3, peptide A, which contained the IGFBP-5 region from amino acid 201 to
218, was an effective competitive inhibitor. Peptide B, whose amino
acid sequence contains a similar number of charged residues from a
different region of the IGFBP-5 molecule (10), had a reduced effect.
The effect of peptide C was variable and in some experiments, it was a
much less effective inhibitor of coimmunoprecipitation (data not
shown). Peptide D had no effect on PAI-1 binding. Both peptides C and D
contain fewer charged residues than peptides A or B.
To further determine the specific amino acids that were involved in the
PAI-1/IGFBP-5 interaction, coprecipitation was conducted using IGFBP-5
mutants that contain neutral substitutions for basic amino acids
between residues 201 and 218. A form of IGFBP-5 that contained four
substitutions K211N, R214A, K217A, and R218A had a 24% reduction in
its ability to bind to PAI-1 (Fig. 4). A
form that contained three substitutions, R202A, K206A, and R207A, also
had decreased binding. A form that contained four substitutions at
positions R201A, K20N, K206N, and K208N had a 71% reduction in
detectable PAI-1 binding. In contrast, two other mutants that contained
substitutions for charged residues in the region of IGFBP-5 between
positions 201 and 218 and K134A/R136A had nearly normal PAI-1 binding.
This suggests that both the positional location and the number of
charged residues that are altered determine binding affinity. In
general, the mutant forms that had reduced binding for
glycosaminoglycans (9) had some alteration in PAI-1 binding, but there
were distinct differences. For example, the R201A, K202N, K206N, K208N
mutant has only minimally reduced GAG binding (9). This suggests that
the GAG binding site and the PAI-1 binding site, while both charge
dependent, have different configurations.
The effect of PAI-1 binding to IGFBP-5 on its susceptibility to
proteolytic cleavage was determined. When IGFBP-5 was incubated with an
IGFBP-5 protease obtained from fibroblast conditioned medium, there was
time-dependent cleavage of the substrate into a 22-kDa band. The
addition of PAI-1 to the incubation buffer resulted in a 1.6-fold
increase in intact IGFBP-5 band intensity (Fig. 5). The addition of vitronectin (an ECM
protein that stabilizes PAI-1) with PAI-1 did not result in additional
inhibition of IGFBP-5 proteolysis. Neither tissue plasminogen activator
(TPA) or urokinase can significantly degrade IGFBP-5; therefore, it is
unlikely that PAI-1 is acting to directly inhibit a TPA-like protease.
The binding of 125I-IGF-I to IGFBP-5 was determined in the
presence and absence of PAI-1 and the results analyzed by the method of
Scatchard: the addition of PAI-1 had no effect on the affinity of
IGFBP-5 for IGF-I (data not shown).
Because multiple extracellular matrix proteins may bind to IGFBP-5
simultaneously, we determined if vitronectin, a known binding protein
for PAI-1, could alter the PAI-1/IGFBP-5 interaction. As shown in Fig. 6, when coimmunoprecipitation was
performed using anti-PAI-1 antibodies, increasing concentrations of
vitronectin inhibited the amount of IGFBP-5 that could be
coimmunoprecipitated with PAI-1 by 33% (Fig. 6, compare lane 2 to lane
3). This suggests that vitronectin is capable of binding to PAI-1 and
preventing the IGFBP-5/PAI-1 complex formation. When the experiment was
repeated with antivitronectin antibodies, vitronectin and
125I-IGFBP5 coprecipitated (Fig. 6, panel B). PAI-1 (100
ng/ml) interfered with the interaction and reduced the amount of
125I-IGFBP-5 that could be coprecipitated to basal
levels.
To assess the physiological significance of PAI-1/IGFBP-5 binding,
extracellular matrix was prepared from human fibroblast cultures as
previously described (18). 125I-IGFBP-5 was incubated
overnight with the ECM extracts and increasing concentrations of
anti-PAI-1 antibody. This antibody recognizes ECM bound PAI-1. The
addition of increasing amounts of anti-PAI-1 to the incubation mixture
partially blocked ability of 125I-IGFBP-5 to bind to the
ECM (Table 1). In contrast, an equal
concentration nonimmune IgG had no effect. Approximately 27% of the
total 125I-IGFBP-5 binding could be inhibited by this
antibody. This suggests that one of the components of ECM that binds to
IGFBP-5 is PAI-1. However, because glycosaminoglycans inhibit IGFBP-5
binding to PAI-1 and are abundant in ECM, they may be a more important
binding determinant. To determine the affinity of IGFBP-5 for PAI-1,
competitive binding assays using unlabeled and 125I-IGFBP-5
were performed. As shown in Fig. 7, the
results are consistent with a single site model with relatively high
affinity. When the data from three experiments were analyzed by the
method of Scatchard, the affinity (Kd) was determined to be
9.1 x 10-8 M.
|
Discussion
|
|---|
IGFBP-5 is a potent modulator of IGF-I actions, and the factors
that control its abundance in the pericellular environment may be
important determinants of how this growth factor functions to modify
cellular responses. The PAI-1/IGFBP-5 interaction may be particularly
relevant to controlling IGFBP-5 abundance in the ECM. In this paper, we
demonstrate that IGFBP-5 binds to PAI-1 and that the binding
interaction has a relatively high affinity. An affinity constant in the
range of 10-9 M was estimated by Scatchard
analysis. Additionally, we show that anti-PAI-1 antiserum inhibits
IGFBP-5 binding to ECM. This finding suggests that PAI-1 may be one of
the major binding sites for IGFBP-5 within the extracellular matrix.
However, anti-PAI-1 antiserum caused only a 27% reduction in PAI-1
binding that supports the conclusion that other ECM components are also
important determinants of IGFBP-5 binding. Because glycosaminoglycans
can inhibit the IGFBP-5-PAI-1 interaction, they may represent the
predominant ECM binding component. Because PAI-1 is synthesized by
multiple connective tissue cell types, including cell types that are
involved in the response to injury, such as fibroblasts and smooth
muscle cells, the PAI-1-IGFBP-5 interaction may be a determinant of the
amounts of IGFBP-5 and IGF-I that are localized in ECM and are
available to stimulate the responses of these cells during the repair
process.
Previously, we have shown that the amount of IGFBP-5 that is localized
in the extracellular matrix positively modulates both the cell growth
and DNA synthesis responses to IGF-I(5). Specifically, the mitogenic
response of fibroblasts to IGF-I is enhanced if increased amounts of
IGFBP-5 are present in the ECM. Furthermore, ECM localization protects
IGFBP-5 from proteolysis, thus resulting in increased IGF-I being
present at this site (7). A further understanding of why binding of
IGF-I to IGFBP-5 in the ECM results in enhancement rather than
inhibition of its action was made clearer by the observation that,
after ECM localization of this binding protein, its affinity for IGF-I
is reduced 8-fold (5, 9). Because the affinity of non-ECM associated
IGFBP-5 is 20- to 30-fold greater than the affinity of the type I IGF
receptor, this reduction in affinity would allow IGF-I that is
localized bound to ECM to be in a more favorable equilibrium with
receptors on cell surfaces. Therefore, ECM localization of IGFBP-5 and
IGF-I results in protection from proteolysis and maintenance of a
reservoir of IGF-I that can be released to receptors at critical time
points during cell cycle progression. Because binding to PAI-1 does not
lower IGFBP-5 affinity for IGF-I, this suggests that there may be
differences in the affinity of ECM-associated IGFBP-5 and that the net
reduction in affinity may be dependent upon whether binding is to
predominantly glycosaminoglycans or to other proteins such as PAI-1.
Therefore, the factors that regulate the distribution of IGFBP-5 among
the various ECM binding proteins may indirectly alter IGFBP-5 affinity
and thereby modulate IGF-I action.
The variables that regulate the synthesis of IGFBP-5 are also important
determinants of its abundance in ECM. We have recently shown that IGF-I
stimulates IGFBP-5 transcription by smooth muscle cells (19), and Dong
and Canalis have confirmed this finding in osteoblasts (20).
Furthermore, they demonstrated that retinoic acid is also a potent
stimulant of IGFBP-5 synthesis by bone cells and results in increased
amounts of IGFBP-5 in the ECM (20). Hakeda et al. (21) have
shown that prostaglandin-E and PTH stimulate IGFBP-5 release from
osteoblast ECM and this could modulate IGF-I levels in the pericellular
environment.
In addition to the factors that increase IGFBP-5 synthesis, we have
shown that several connective tissue cell types release an IGFBP-5
protease into their conditioned medium (7, 12, 19). The factors that
regulate the activity of this protease may be important determinants of
IGFBP-5 abundance in interstitial fluids and in the ECM. These studies
show that PAI-1 binding to IGFBP-5 alters its susceptibility to
proteolysis in the interstitial fluid. However, because multiple
substances, such as PAI-1 (9) and vitronectin may bind to IGFBP-5 in
ECM, other ECM proteins may function coordinately with PAI-1 or
vitronectin to stabilize IGFBP-5. The synthesis of PAI-1 may also be
increased coordinately with IGFBP-5. Specifically, Padaytly et
al. and Afossa et al. have shown that IGF-I increases
PAI-1 synthesis, both in vivo and in vitro;
therefore, it is possible that PAI-1 and IGFBP-5 synthesis are
coordinately regulated by IGF-I in selected cell types (22, 23).
Conversely, factors such as thrombin (24) that cause release of PAI-1
from ECM might function to decrease ECM-associated IGFBP-5. Heparin
functions to inhibit the IGFBP-5/PAI-1 interaction and to reduce the
abundance of PAI-1 or IGFBP-5 within ECM (25, 26, 27); therefore,
heparin inhibition of IGFBP-5 binding to ECM could be a component of
the mechanism by which heparin functions to inhibit smooth muscle cell
replication.
IGF-I synthesis and peptide abundance is also often increased at sites
of injury (28, 29). Similarly PAI-1 expression is increased after
injury. The 5' flanking sequences of both the IGFBP-5 and PAI-1 genes
contain the Egr-1 injury response element (30), which mediates an
increase in the transcription of specific genes after injury.
Therefore, it is possible that PAI-1 and IGFBP-5 synthesis may be
directly increased in response to injury. Because IGF-I synthesis also
increases after injury, this increase in IGF-I could further enhance
IGFBP-5 expression by a mechanism that is distinct from Egr-1 (19).
PAI-1 is a component of smooth muscle cell extracellular matrix. It is
also synthesized by endothelium in response to injury; therefore,
because it binds tightly to IGFBP-5, it could provide a mechanism for
stabilizing large amounts of IGFBP-5 in the newly developing ECM during
neointima formation. PAI-1 also has other interesting properties
because it binds vitronectin (31, 32) and the PAI-1 binding domain is
also the region of vitronectin that interacts with the vitronectin
receptor (31). This suggests other potential mechanisms for enhancing
IGF-I action. Proteases, such as thrombin, could release vitronectin
from the ECM and allow more vitronectin to interact with its receptor.
Similarly, if they degrade IGFBP-5, they would allow release of IGF-I
to receptors. Simultaneous vitronectin receptor and IGF receptor
occupancy has been shown to be a potent stimulant of smooth muscle cell
migration (33), a process that is an important component of neointina
formation.
The exact site of molecular interaction between PAI-1 and IGFBP-5
remains to be determined. However, it is clear that ionic residues are
involved. Similarly, glycosaminoglycans may interfere with this
interaction, and there may be a competitive equilibrium between
proteoglycans that are present in the extracellular matrix and PAI-1.
Because plasminogen is not the major protease in pSMC or fibroblast
medium that degrades IGFBP-5, it is not surprising that PAI-1 binding
of IGFBP-5 did not completely protect it from proteolysis. However,
plasmin can degrade IGFBP-5, so PAI-1 could play an indirect role in
maintaining intact IGFBP-5 by inhibiting plasmin formation (34). The
charged residue motif that binds PAI-1 is clearly distinct from the
heparin binding site (9), although overlapping residues are used. This
suggests that the highly charged regions of IGFBP-5 that are surface
exposed recognize distinct substrates differently and that there may be
specificity for substrate binding. The exact factors that control the
surface exposure of distinct sites in IGFBP-5 are unknown but will be
interesting points for future study.
View larger version (33K):
[in this window]
[in a new window]
|
Figure 11. Coimmunoprecipitation of IGFBP-5 and PAI-1. A,
IGFBP-5, 50 ng/ml, was incubated with PAI-1, 100 ng/ml, and anti-PAI-1
antibody, 1:5000 dilution, for 16 h at 4 C. At that time,
immunoprecipitation was conducted by adding protein A sepharose, as
described in Materials and Methods. The pellets were
dissolved in Laemmli sample buffer and the products electrophoresed
through a 12.5% gel, then transferred to Immobilon membranes and
immunoblotted using a 1:1000 dilution of anti-IGFBP-5 antiserum. The
results show that, in the presence of PAI-1, IGFBP-5 could be
coimmunoprecipitated, whereas the anti-PAI-1 antibody did not
immunoprecipitate IGFBP-5 without PAI-1 being added. PAI-1 did not
react with the IGFBP-5 antibody. B, 125I-IGFBP-5, 50,000
cpm/ml, was incubated with anti-PAI-1 antiserum (1:5000) and PAI-1 (100
ng/ml). In the presence of PAI-1, an abundant 125I-IGFBP-5
band was detected. In the absence of the PAI-1 antibody or unlabeled
PAI-1, minimal 125I-IGFBP-5 was immunoprecipitated. The
asterisk denotes an incubation mixture that contained
excess unlabeled IGFBP-5 (100 ng/ml).
|
|
View larger version (45K):
[in this window]
[in a new window]
|
Figure 21. The effects of glycosaminoglycans on IGFBP-5/PAI-1
association. 125I-IGFBP-5 was incubated without, lane 1, or
with 100 ng/ml of PAI-1 (lanes 2–7), then immunoprecipitated as
described in Materials and Methods. After
immunoprecipitation, the pellets were dissolved in Laemmli sample
buffer and the products electrophoresed, then transferred to Immobilon
membranes and autoradiography performed directly; lanes 2–7,
125I-IGFBP-5; lanes 3–7, 125I-IGFBP-5 plus
various glycosaminoglycans (0.1 µg/ml). Lane 3, heparin; lane 4,
heparan sulfate; lane 5, chondroitin sulfate A; lane 6, chondroitin
sulfate B; lane 7, chondroitin sulfate C. The radiolabeled band shown
in lane 1 that migrates with an Mr estimate of 16 kDa is a
fragment of radiolabeled IGFBP-5.
|
|
View larger version (78K):
[in this window]
[in a new window]
|
Figure 31. The effects of IGFBP-5 peptides on PAI-1 and
IGFBP-5 binding. Radiolabeled IGFBP-5, 50,000 cpm/ml, was incubated
with unlabelled PAI-1, 100 ng/ml, and 100 µg/ml of one of four
peptides, lanes 2–5, that contained various sequences within the
IGFBP-5 molecule (9 ). After a 14-h incubation, the products were
coimmunoprecipitated and processed by SDS-PAGE and transferred to
Immobilon. The figure is an autoradiogram showing the amount of
125I-IGFBP-5 that was coimmunoprecipitated. Lanes 1–5,
125I-IGFBP-5; lane 2, peptide A; lane 3, peptide B; lane 4,
peptide C; lane 5, peptide D. The radiolabeled band with an
Mr estimated of 18 kDa shown in lanes 1, 4, and 5 is a
fragment of radiolabeled IGFBP-5.
|
|
View larger version (68K):
[in this window]
[in a new window]
|
Figure 41. Coimmunoprecipitation of IGFBP-5 mutants and
PAI-1. One hundred nanograms per milliliter of each IGFBP-5 mutant was
incubated with 100 ng/ml of PAI-1 and the products immunoprecipitated
using anti-PAI-1 antiserum as described previously. After SDS-PAGE, the
products were transferred to an Immobilon membrane, then ligand blotted
using 125I-IGF-I as described in methods. Lane 1, wild-type
IGFBP-5, no PAI-1; lanes 2–8, PAI-1 plus IGFBP-5 or mutants. Lane 2,
wild-type IGFBP-5; lane 3, K134A/R136A; lane 4, R202A/K206A/R207A; lane
5, R201A/K202N/K206N/K208N; lane 6, K211N; lane 7, R201N/K202N; lane 8,
K211N/R214A/K217A/R218A. The results of Phor Image analysis expressed
as scanning units were: lane 1, 2,237; lane 2, 190,511; lane 3,
204,125; lane 4, 170,867; lane 5, 76,748; lane 6, 250,813; lane 7,
223,452; lane 8, 146,656.
|
|
View larger version (85K):
[in this window]
[in a new window]
|
Figure 51. The effect of PAI-1 on IGFBP-5 proteolysis.
IGFBP-5, 50 ng/ml, was incubated with IGFBP-5 protease that had been
purified from human fibroblast conditioned medium. After a 14-h
incubation, the products were electrophoresed and analyzed by
immunoblotting. Lanes 1–4, IGFBP-5 plus protease; lane 2, vitronectin,
10 µg/ml; lane 3, PAI-1, 10 µg/ml; lane 4, PAI-1 plus vitronectin.
The 22-kDa band, which is the major cleavage product of this protease,
is shown by an arrow. Scanning unit values for the
intact IGFBP-5 bands were; lane 1, 87,644; lane 2, 107,781; lane 3,
144,811; and lane 4, 123,322.
|
|
View larger version (38K):
[in this window]
[in a new window]
|
Figure 61. The effect of vitronectin on the binding of PAI-1
to IGFBP-5. A, 125I-IGFBP-5 and PAI-1 (100 ng/ml) were
incubated with anti-PAI-1 antisera and increasing concentrations of
vitronectin. After a 6-h incubation, the products were
immunoprecipitated, electrophoresed, and the gel analyzed by
autoradiography. Lane 1, 125I-IGFBP-5 alone; lanes 2–5,
125I-IGFBP-5 plus PAI-1; lane 3, vitronectin (1 µg/ml);
lane 4, vitronectin (0.1 µg/ml); lane 5, vitronectin (0.01 µg/ml).
Phosphor Image intensity units were: lane 1, 46,673; lane 2, 98160;
lane 3, 65,869; lane 4, 75077; and lane 5, 81,264. B,
125I-IGFBP-5 and vitronectin (100 ng/ml) were incubated
wi | |
Top
Abstract
Introduction
