(Received for publication, January 9, 1997, and in revised form, March 10, 1997)
From the Department of Vascular Biology, The Scripps
Research Institute, La Jolla, California 92037 and ¶ The
Burnham Institute, La Jolla, California 92037
Vitronectin (Vn) is a major adhesive glycoprotein
in blood. However, many of the functions of Vn are regulated by its
conformational state and degree of multimerization. Here, the ability
of native and denatured Vn to bind to integrin adhesion receptors was
compared. Three lines of evidence suggest that the native, plasma form
of Vn is not an adhesive glycoprotein. (i) Antibodies that bind in close proximity to the cell adhesion domain of Vn fail to bind to
native Vn present in unfractionated plasma. (ii) Denatured Vn binds to
both glycoprotein IIb/IIIa and v
3
in a dose-dependent manner. In contrast, native Vn is
unable to bind either integrin. (iii) Thermal denaturation of native
Vn, or its complexation with type 1 plasminogen activator inhibitor,
exposed the cell adhesion domain of Vn. Thus, while plasma Vn is unable
to bind integrins and is not an adhesive glycoprotein, the
conformationally altered from of the protein binds avidly to both
v
3 and glycoprotein IIb/IIIa. The data
presented here indicate that such conformational changes in Vn are
likely to occur in areas of tissue injury and thrombosis.
Vitronectin (Vn),1 an adhesive glycoprotein present in blood and in a variety of tissues, belongs to a group of molecules that contain the RGD cell adhesion sequence and play key roles in the attachment of cells to their surrounding matrix (1, 2). In addition, Vn has several other functions in the complement, coagulation, and fibrinolytic system. For example, Vn is identical to the S protein of the complement system and thus inhibits complement-mediated cell lysis (3, 4). Vn competes with heparin binding to antithrombin III, thereby preventing the rapid inactivation of thrombin and factor Xa by this protease inhibitor (5, 6). Vn also binds to and stabilizes the biological activity of type 1 plasminogen activator inhibitor (PAI-1), the physiological inhibitor of both tissue- and urinary-type plasminogen activators (7, 8). Thus, Vn provides unique regulatory links between cell adhesion and proteolytic enzyme cascades, although it is not clear how the multiple functions of Vn are coordinately regulated.
Vn is a conformationally labile molecule, and this lability is likely to play a large role in regulating the functions of the protein. This notion is based on the observation that different preparations of Vn exhibit differing abilities to bind to the glycosaminoglycan heparin. Only 2% of the Vn in plasma is capable of binding to heparin-Sepharose, whereas this fraction is increased to 7% by the generation of serum (9). In addition, the formation of complexes of Vn with PAI-1 (10), thrombin-antithrombin III (11), or complement C5b-C9 (12) induces conformational changes in the molecule, which expose the heparin binding domain. Moreover, conformational changes are also induced by denaturation with chaotropic agents, heat-treatment and acidification. These changes are accompanied by the spontaneous formation of disulfide-linked multimers of Vn. Using a panel of conformationally sensitive antibodies, we provided evidence that conformational changes within Vn are not limited to the heparin binding domain, but also occur in the N-terminal half of the molecule, including the N-terminal somatomedin B (SMB) domain (13), a site that encompasses the RGD sequence.
Understanding the interrelationship between the binding domains of Vn
has been complicate, because the exposure and function of these sites
depends on the conformational state of this molecule. For example, a
number of ligands interact preferentially with the conformationally
altered and multimeric (i.e. denatured form) of Vn. These
include collagen (14), glycosaminoglycans (15, 16), -endorphin (17),
PAI-1 (18), and urokinase receptor-urokinase complex (19, 20). In
contrast, thrombin-antithrombin III complexes appear to preferentially
interact with native, plasma Vn (21). The function of native and
denatured Vn in promoting cell adhesion is even more complicated.
Native Vn was reported to bind specifically and saturably to
glycoprotein (GP) IIb/IIIa on stimulated platelets (22-24). The
binding of denatured Vn to platelets are conflicting (25). In addition,
Vn has been reported to both block (26) and stimulate (23) platelet
aggregation. The reasons for these discrepancies remain unclear.
Vn is one of many adhesive proteins that bind to cell surface receptors
called integrins (27, 28). Each integrin is a noncovalently associated
heterodimer that spans the plasma membrane. To date, nine
integrin
subunits and 12
subunits have been identified. Since
the ligand binding specificity of each receptor is conferred by its
subunit composition, functional diversity within the family of
integrins is achieved by heterologous pairing between subunits. Three
different integrins have been identified as receptors for Vn,
GPIIb/IIIa (also known as
II
3 (29)),
v
3, expressed by endothelial cells and
implicated in angiogenesis (30), and
v
5
(31). Vn contains a single RGD sequence located adjacent to the SMB
domain (3, 4). Site-directed mutagenesis of the RGD motif shows that
this sequence is required for cell adhesion and can not be compensated
for by other parts of the molecule (32, 33). A number of reports
indicated that the cell adhesion domain is exposed on both native and
denatured Vn (1, 34). Consequently, the adhesive capacity of the
protein is not thought to be regulated by the conformational lability of the molecule. However, this concept is challenged by two
observations. First, only denatured Vn induces tyrosine phosphorylation
in endothelial cells by binding to the
v
3
integrin on the cell (35). Second,
v
5-dependent endocytosis of
Vn by skin fibroblasts was only observed using denatured Vn, the native
form of Vn is not internalized (36, 37).
In the present report, we compare the binding of native and denatured
Vn to purified GPIIb/IIIa and v
3.
Evidence is provided that native Vn is incapable of binding to integrin
adhesion receptors and that its RGD sequence is not exposed.
Conformational changes in Vn, induced by chemical or thermal
denaturation, or by complexation with PAI-1, expose the integrin
binding site on Vn, allowing it to bind tightly to both
v
3 and GPIIb/IIIa.
Denatured Vn was purified by
heparin affinity chromatography in the presence of 8 M urea
(38). Vn was also isolated under nondenaturing conditions
(i.e. native Vn) according to published procedures (39).
Protein concentrations were determined by the bicinchoninic acid method
(Pierce). The final Vn preparations were devoid of fibrinogen and
fibronectin immunoreactivity as judged by immunoblotting using rabbit
anti-human fibrinogen or fibronectin (Calbiochem). Denatured Vn was
biotinylated using NHS-LC-biotin (Pierce). Briefly, 1.5 mg of Vn was
dissolved in 50 mM sodium phosphate buffer, pH 8.5, at 20 µM and incubated with 3 mg of biotinylation reagent
dissolved in 100 µl of Me2SO for 3 h at 37 °C,
followed by extensive dialysis against phosphate-buffered saline (PBS).
GPIIb/IIIa was obtained from Enzyme Research Laboratories, and
v
3 was isolated from human placenta
extracts by affinity chromatography (40). Human fibrinogen was obtained
from Calbiochem. PAI-1 was purified as described (41), activated with 4 M guanidinium hydrochloride (42), dialyzed against 5 mM sodium phosphate, 1 mM EDTA, 0.345 M NaCl, pH 6.5, to stabilize the biological activity of
PAI-1 (43), and the specific activity was determined by titration against urinary-type plasminogen activator (Calbiochem). Prior to each
experiment, PAI-1 was dialyzed into ice-cold PBS. Latent PAI-1 was
prepared by incubation of activated PAI-1 at 37 °C for 24 h.
After this incubation step, the structural integrity of the resulting
PAI-1 preparation was confirmed by SDS-polyacrylamide gel
electrophoresis followed by staining with silver nitrate. The remaining
PAI-1 activity after this incubation was determined by binding to
immobilized tissue-type plasminogen activator (Calbiochem) (44) and was
less than 0.5% of the starting material (not shown). Monoclonal
antibodies (mAbs) 611, 153, and 1244 were obtained by standard
hybridoma techniques using denatured Vn as immunogen (41). IgG was
produced in mice as ascites fluid and purified by using protein
A-Sepharose. The generation of rabbit anti-human Vn was described
previously (41). Biotin-labeled secondary antibodies, streptavidin
alkaline phosphatase, and substrate were from Zymed. Conformational
changes in the Vn molecule were quantified by competitive enzyme-linked
immunosorbent assay (13). Briefly, microtiter wells were coated with
denatured Vn (1 µg/ml in PBS, 4 °C, 16 h), and nonspecific
binding sites on the plastic dishes were blocked by incubating the
washed wells for 1 h at 37 °C in PBS containing 3% casein and
0.05% Tween 20. The wells were co-incubated for 1 h with a
constant amount of IgG (50 ng/ml final concentration) and the indicated
concentration of Vn in PBS containing 0.1% casein and 0.1% Tween 20. Bound IgG was detected with biotin-conjugated goat anti-mouse IgG,
followed by streptavidin alkaline phosphatase conjugate and the
chromogenic substrate p-nitrophenyl phosphate (Zymed). The
change of absorbance at 405 nm of duplicate wells was determined and
the absorbance of wells not incubated with antibodies but otherwise
treated identically was subtracted.
The human
fibrosarcoma cell line HT 1080 was obtained from ATCC and maintained in
Dulbecco's modified Eagles's medium/F-12 medium containing 10% fetal
bovine serum as described (45). Semiconfluent cultures were labeled for
16 h with [35S]methionine (Amersham Corp; 100 µCi/ml in 80% methionine-deficient medium, 20% complete growth
medium). After washing with PBS, the cells were detached by treatment
with PBS containing 0.2 mM EDTA for 5 min and washed three
times in adhesion buffer (1 × Hanks' balanced salt solution, 50 mM Hepes, 1 mg/ml bovine serum albumin, 1 mM
CaCl2, 1 mM MgCl2, 1 mM
MnCl2, pH 7.4) and resuspended at 2 × 105
cells/ml in adhesion buffer. Flat-bottomed microtiter wells were coated
with 100 ng/well denatured Vn. Following an 18-h coating period at
4 °C, the plates were blocked with PBS containing 10 mg/ml bovine
serum albumin. Antibodies (50 µl) were prebound to the wells at
37 °C for 1 h. Cells (1 × 104) were added to
the antibody containing wells in triplicate and allow to adhere for
1 h. The wells were washed gentle three times in PBS to remove
nonadherent cells, and the bound cells were solubilized in 1% SDS and
quantified by -counting. The radioactivity associated with bovine
serum albumin coated wells was subtracted from that of the Vn-coated
wells. Results are expressed as percentage cell binding in the absence
of antibodies.
The binding of purified Vn
preparations to immobilized integrins was analyzed with the methods
essential as described previously (46). Briefly, integrins were
immobilized at 50 ng/well (v
3) or 250 ng/well (GPIIb/IIIa) for 18 h at 4 °C. Subsequently,
nonspecific protein binding sites on the plates were blocked with 50 mM Tris, pH 7.4, containing 150 mM NaCl, 1 mM MgCl2, 1 mM CaCl2, 1 mM MnCl2 (binding buffer), and 35 mg/ml bovine
serum albumin. After washing with binding buffer, unlabeled Vn was
added for 1 h at 37 °C in binding buffer containing 0.1%
bovine serum albumin. Bound Vn was detected using rabbit anti-human Vn
IgG (10 µg/ml), followed by biotin-labeled anti-rabbit IgG,
streptavidin alkaline phosphatase, and substrate. The change of
absorbance (405 nm) of duplicate wells was determined. Specific binding
was determined by subtracting the binding to wells incubated with the
same concentration of Vn in the presence of 10 mM EDTA. For
competitive ligand binding studies, biotin-labeled denatured Vn (1 µg/ml; a concentration required to obtain half- maximal binding) was
co-incubated (3 h, 37 °C) with unlabeled native Vn, denatured Vn, or
fibrinogen on integrin-coated plates (see above). After washing, bound
Vn was detected using streptavidin alkaline phosphatase and substrate. Specific binding was determined by subtracting the binding of biotin-labeled wells incubated in the presence of 10 mM
EDTA from that incubated in the presence of divalent cations. Results
are expressed as percentage binding in the absence of soluble
competitor.
The
effects of three different anti-Vn mAbs on cell adhesion to immobilized
Vn were determined (Fig. 1A). Microtiter
wells were coated with denatured Vn and incubated with the indicated concentration of antibody and metabolically labeled HT 1080 cells. Cells were allowed to attach for 1 h. Bound cells were quantified by -counting. mAb 153 and 611 inhibited cell adhesion to Vn in a
dose-dependent manner, whereas mAb 1244 was significantly
(approximately 20-fold) less effective. The observed difference was
apparently not due to a lower affinity of mAb 1244 for denatured Vn,
since this antibody and clone 153 differ in the affinity for denatured Vn only by factor 2 (41). In contrast, no inhibition was observed using
normal mouse IgG (data not shown). The mAbs 153 and 611 bind to Vn
polypeptides encompassing amino acids 1-51, a fragment of Vn
containing the RGD sequence, whereas the control antibody interacts
with amino acids 52-239 (41) (data not shown).
We previously provided evidence that an immunoepitope (mAb 153) located in the SMB domain of Vn (amino acids 1-51) is cryptic in plasma Vn (13). Experiments were performed to understand whether lack of reactivity with native plasma Vn is a general feature of antibodies derived to this region of the Vn molecule. Competitive enzyme-linked immunosorbent assays were employed to determine the reactivity of mAB 611 with purified preparations of native and denatured Vn and Vn present in unfractionated plasma (Fig. 1B). Results were compared with mAb 153. Microtiter wells coated with denatured Vn were incubated with a constant amount of anti-Vn IgG and increasing concentrations of different types of Vn (Fig. 1B). Denatured Vn competes with the binding of both antibodies to immobilized denatured Vn in a dose-dependent manner, whereas neither antibody reacted well with Vn present in unfractionated plasma. The reactivity with purified native Vn with these two conformationally sensitive antibodies was at least 50-fold lower than that of denatured Vn.
Binding of Native and Denatured Vn to GPIIb/IIIa andBased on the strong reactivity
of the inhibitory antibodies with denatured Vn, and the inability of
these antibodies to bind native Vn, we hypothesized that the RGD
sequence in native Vn is cryptic. This hypothesis was tested by
measuring the binding of native and denatured Vn to purified integrins
(Figs. 2 and 3). To measure binding,
microtiter wells were coated with purified integrin and then incubated
with a range of either native or denatured Vn. Bound Vn was detected
with an polyclonal antibody against Vn. Denatured Vn bound to
GPIIb/IIIa in a dose-dependent manner and approached
saturation at 5 µg/ml. Specific binding did not further increase
using 10 µg/ml denatured Vn. In the concentration range presented in
Fig. 2A, specific binding was more than 85% of total
binding. In contrast, little binding of native Vn was observed under
the same conditions.
To exclude the possibility that the lack of binding of native Vn to GPIIb/IIIa was due to reduced reactivity of the rabbit anti-Vn IgG with native Vn, competitive binding assays were employed to confirm the results of the direct binding studies (Fig. 2B). In this assay, immobilized GPIIb/IIIa was co-incubated with a constant amount of biotinylated denatured Vn and a range of competing native or denatured Vn. Fibrinogen was also used as a competitor, because it is the physiological ligand for GPIIb/IIIa. Fibrinogen was 10-50-fold more efficient in competing with the binding of labeled Vn for GPIIb/IIIa than denatured Vn. More importantly, no competition was observed using native Vn, confirming the inability of the native form of Vn to bind to GPIIb/IIIa.
GPIIb/IIIa is only one of several integrins that are reported to bind
Vn. The v
3 integrin is also considered a
receptor for Vn and was originally named the "vitronectin
receptor." To determine whether the
v
3
integrin also exhibited preferential binding to the denatured form of
Vn, similar binding studies were performed using immobilized
v
3 (Fig. 3A). Denatured Vn
bound to
v
3 in a
dose-dependent manner and again approached saturation at 5 µg/ml Vn. In contrast, native Vn showed little binding. Again, results were confirmed in competitive binding studies and compared with
fibrinogen. Native Vn failed to compete with the binding of denatured
Vn for binding to
v
3. Both denatured Vn
and fibrinogen competed for the binding of biotinylated Vn to
v
3. Little difference was observed in the
extent of competition between the latter two ligands.
Only about
7% of the total Vn present in plasma is recovered using a purification
procedure that maintains the protein in its native conformation (39).
This raised the possibility that a nonintegrin-binding subpopulation of
native Vn could have been purified and employed in the binding studies.
To address this issue, a heat denaturation procedure was employed to
convert native Vn into conformationally altered, multimeric Vn (Fig.
4). Heat treatment of native Vn resulted in the exposure
of the mAb 611 epitope (Fig. 4A) and mAb 153 epitope (not
shown) to a similar extent as in denatured Vn. In addition, the
electrophoretic mobility of heat-treated native Vn on native
polyacrylamide gel electrophoresis was indistinguishable from denatured
Vn (i.e. heat treatment resulted in the formation of high
molecular weight multimers that failed to enter the separating gel)
(not shown). The interaction of heat-treated native Vn with
v
3 was compared with that of purified
denatured Vn (Fig. 4B). Heat-treated Vn binds to immobilized
v
3 in a dose-dependent manner
similar to denatured Vn (Fig. 4B). In addition, heat-treated native Vn competes with labeled denatured Vn for binding to this integrin to an extent comparable with that of denatured Vn (Fig. 4C). Similar results were obtained when binding to
GPIIb/IIIa was examined (not shown). In addition, an alternative
denaturation procedure (i.e. treatment of native Vn with 8 M urea) also induced integrin binding (not shown). These
observations indicate that denaturation of native Vn results in the
exposure of the cryptic cell binding domain and that the inability of
native Vn to bind to
v
3 or GPIIb/IIIa was
not due to the purification of a nonintegrin-binding subpopulation of
Vn from plasma.
Effects of PAI-1 on the Adhesive Functions of Native and Denatured Vn
Vn binds to a number of plasma proteins, and these
interactions alter the conformation of this molecule (10-12).
Experiments were performed to test whether physiological ligands of
native Vn can induce the exposure of the cell adhesion domain. The
binding of native Vn to active PAI-1 results in the formation of Vn
multimers that express epitopes for conformationally sensitive
antibodies like mAbs 153 (10) and 611 (data not shown). Interestingly, PAI-1·Vn complexes are relatively labile and PAI-1 readily
dissociates from Vn, but Vn remains multimeric and conformationally
altered (10). We reasoned that the binding of PAI-1 to native Vn is likely to expose the integrin binding site on Vn. Native Vn was incubated with a 2-fold molar excess of active PAI-1 for 16 h at
37 °C. After this incubation period, more than 99% of the PAI-1 was
converted into the latent conformation as based by binding to
immobilized tissue-type plasminogen activator (not shown). Little
competition of native Vn (Fig. 5A, left closed
bar) for mAB 153 binding to immobilized denatured Vn was observed.
Incubation of native Vn with active PAI-1 (Fig. 5A, left open
bar) resulted in the exposure of the conformationally sensitive
mAb 153 epitope. In contrast, no effects of latent PAI-1 on the
exposure of this epitope was observed (Fig. 5A, left hatched
bar). The mAb 153 does not detect Vn in complex with PAI-1 (41),
confirming that the majority of the Vn was present in an uncomplexed,
conformationally altered form. In parallel experiments, the effects of
pretreatment of native Vn with PAI-1 on its ability to bind to
v
3 (Fig. 5A, right bars) was
analyzed. Again, while little binding of native Vn (Fig. 5A,
right closed bar) to
v
3 was
observed, PAI-1 induced Vn multimers (A, right open bar)
demonstrated increased binding to
v
3. In
contrast, latent PAI-1 (Fig. 5A, right hatched bar) had
failed to induce
v
3 binding of native
Vn.
The PAI-1 and integrin binding sites in denatured Vn are structurally
distinct but located in close proximity, raising the possibility that
PAI-1 may modulate the adhesive functions of denatured Vn. To test this
hypothesis, microtiter wells were co-incubated with a constant amount
of denatured Vn (which binds to v
3) and increasing concentrations of either active or latent PAI-1 in this
binding assay (Fig. 5B). Active PAI-1 blocked the binding of
denatured Vn to immobilized
v
3 in a
dose-dependent manner. At the highest concentration of
PAI-1, approximately 80% inhibition was observed. The competition
reached a plateau at a 10-fold molar excess of active PAI-1 over
denatured Vn. In contrast, latent PAI-1 has little effect on the
binding of Vn to
v
3. These findings are
consistent with the observation that latent PAI-1 lacks high-affinity binding to denatured Vn (41). However, when
Vn·
v
3 were preformed on microtiter
wells, active PAI-1 failed to dissociated these complexes (data not
shown).
The observations in this report demonstrate that native Vn,
present in blood, is not an adhesive glycoprotein. This conclusion is
based on the findings that antibodies against Vn that block its
adhesive function fail to bind to the native form of Vn. These blocking
antibodies are capable of binding to denatured Vn or PAI-1-induced Vn
multimers. In addition, direct binding studies prove that denatured Vn
binds to both GPIIb/IIIa and v
3, but that
native Vn is unable to bind to these integrins. Moreover, native Vn,
which is not adhesive, could be converted into an adhesive glycoprotein
by chemical and thermal denaturation. It should be noted that at high
concentration, binding of native Vn to integrins was observed. This
finding is probably related to the presence of some denatured Vn in the
native Vn preparation. This conclusion is supported by the observation
that unfractionated plasma was unreactive with conformationally
sensitive antibodies, whereas the native Vn employed in this study
showed exposure, although limited in comparison with denatured Vn, of
conformationally sensitive epitopes. Based on the observations
presented here, the majority of Vn present in vivo
(i.e. in plasma) is not expected to bind to integrins. A
number of consideration suggest that the adhesive functions of Vn are
mediated through the N-terminal SMB domain rather then the C-terminal
glycosaminoglycan binding domain. First, site-directed mutagenesis
studies revealed that the RGD motif located between amino acids 45 and
47 is required for cell adhesion to Vn (32, 33). Second, heparin does
not effect the adhesion of HT 1080 or a number of other cell lines to
immobilized Vn.2 Similarly, heparin was
without effect in the purified receptor binding assay. Third, recent
studies from our laboratory indicate that native Vn binds heparin,
implying that the glycosaminoglycan binding domain is exposed in the
native conformation.3
Interestingly, PAI-1 can have opposing effects on the adhesive
properties of Vn. First, active PAI-1 blocked the binding of denatured
Vn to v
3 in a dose-dependent
manner. The binding site for active PAI-1 has been localized to amino
acids 1 to 40 (41), a region in close proximity to the RGD sequence at
residues 45-47 of Vn (3, 4). RGD-containing peptide has no effect on
the binding of PAI-1 to Vn (41), suggesting that PAI-1 is not binding directly to the cell adhesion domain in Vn. The close proximity of both
sides makes it likely that the competition of PAI-1 is due to steric
hinderance. In contrast, latent PAI-1, which does not bind to the
N-terminal high affinity PAI-1 binding site in Vn (41), failed to
compete with integrin binding of Vn. In addition, PAI-1 failed to
dissociate pre-existing Vn·
v
3
complexes. Thus, PAI-1 is not expected to detach anchored cells, but
may modulate biological processes dependent on de novo
formation of adhesive contacts between Vn and
v
3, including cell migration. These observations have been confirmed in a recent publication (47).
Second, active PAI-1 had an additional affect on Vn binding to
integrins. PAI-1 induced the formation of conformationally altered Vn
multimers, and these multimers present, after dissociation from PAI-1,
an exposed cell adhesion domain. The Vn concentration in plasma is
between 2.5 and 5 µM (1, 2), whereas normal PAI-1 levels
are approximately 0.4 nM (48). Plasma levels of PAI-1 are
regulated under pathophysiological conditions, and for example
endotoxinemia results in a dramatic up-regulation of plasma PAI-1
levels to 20 nM (48). However, these concentrations are still 100-fold lower than that of Vn. Using this ratio of PAI-1 and Vn,
we were unable to detect PAI-1 induced Vn multimers (10). Thus, PAI-1
induced Vn multimerization in plasma is expected to have little affects
on the adhesive properties of Vn. The scenario may be quite different
in tissues and in platelets. PAI-1 and Vn are both contained within
platelet -granules (49, 50), and platelets contain 26 nM
Vn and 12 nM PAI-1 per 109 platelets (49, 51,
52). Interestingly, platelet Vn is present in high molecular weight
multimers that express the mAB 153 epitope (50). It should be noted
that the expression of the cryptic mAB 153 epitope appears to correlate
with the adhesive properties of Vn. These observation suggest that
platelet Vn is adhesion competent. Platelet-derived proteins
concentrate up to 200-fold in areas of thrombosis, suggesting that
platelet-derived adhesion competent Vn will accumulate in areas of
tissue injury and thrombosis. Similarly, increased tissue-associated
levels of PAI-1 have been reported in a number of pathophysiological
conditions, including endotoxinemia (48), raising the possibility that
PAI-1 will regulate the adhesive properties of Vn in inflamed tissues.
Taken together, these observation support the concept that plasma Vn is
in a latent form. Interactions with either free proteases inhibitors or
inhibitors in complex with the respective target protease results in
the exposure of the cryptic RGD motif.