Brief Review |
From the Department of Morphology, University Medical Center, Geneva, Switzerland.
Correspondence to Dr M.S. Pepper, Département de Morphologie, Centre Médical Universitaire, 1, rue Michel-Servet, 1211 Genève 4, Switzerland. E-mail: michael.pepper{at}medecine.unige.ch
| Abstract |
|---|
|
|
|---|
Key Words: extracellular matrix endothelium metalloproteinase plasminogen cancer
| Introduction |
|---|
|
|
|---|
Like most other biological processes, angiogenesis is
the result of subtle and often complex interactions between regulatory
and effector molecules. To facilitate its analysis, it is
useful to divide angiogenesis into a phase of activation
(sprouting) and a phase of resolution. The phase of activation
encompasses (1) increased vascular permeability and extravascular
fibrin deposition; (2) vessel wall disassembly; (3) basement membrane
degradation; (4) cell migration and ECM invasion; (5)
endothelial cell proliferation; and (6) capillary lumen
formation. The phase of resolution includes (1) inhibition of
endothelial cell proliferation; (2) cessation of cell
migration; (3) basement membrane reconstitution; (4) junctional complex
maturation; and (5) vessel wall assembly, including recruitment and
differentiation of smooth muscle cells and pericytes. Implicit in the
definition of the resolution phase is the establishment of blood flow
in the newly formed vessel.1
Extracellular proteolysis has been implicated in many of these
processes, including basement membrane degradation, cell migration/ECM
invasion, and capillary lumen formation
(the
Table
).5
|
The classic descriptions by Clark and Clark in 19396 of new blood vessel formation in transparent chambers in the rabbit ear clearly demonstrate the production of fibrinolytic activity by growing capillary sprouts: "As new blood vessels advanced ... the fibrin outside them dissolved, leaving a clear space surrounding the new vessels which was filled at times with fluid but more commonly with a transparent gelatinous substance containing occasional blood cells, macrophages and fibroblasts. Although long, tapering, solid tips might be seen extending directly into the fibrin, which was still present beyond the advancing capillary network, the new capillaries, and even capillaries in which the circulation had not yet started, but which possessed a lumen and a rouleaux of erythrocytes, were all surrounded by a characteristic clear space from which fibrin had disappeared." A similar phenomenon of fibrin dissolution was described in studies of acute inflammation induced by subcutaneous implantation of a fibrin clot in the rat.7 Capillary sprouts were associated with "empty zones" in the fibrin clot, which was indicative of fibrin degradation. Fibrinolytic activity decreased progressively in the later stages of the inflammatory process as the fibrin clot was replaced by connective tissue. These two articles were among the first to describe the presence of extracellular proteolytic activity in the endothelial pericellular environment during angiogenesis. They also highlighted the existence of a temporal sequence of matrix dissolution and assembly, in which fibrin provides a temporary matrix scaffold for migrating endothelial cells, which with time is replaced by a more mature collagenous matrix.
One of the earliest events in the activation (sprouting) phase of angiogenesis is basement membrane degradation, and that this process may be mediated by matrix metalloproteinases (MMPs) was demonstrated almost 20 years ago.8 These authors described the presence of type IV and V collagenolytic activity produced by bovine endothelial cells induced to migrate down a chemotactic gradient. No collagenolytic activity was observed in the absence of the gradient (even in the presence of the chemotactic agent), illustrating that enzyme activity was associated with the migration process.
Considerable progress has since been made in understanding the mechanisms of angiogenesis. Thus, angiogenesis is controlled by the net balance between molecules that have positive and negative regulatory activity.1 With the exception of angiogenesis that occurs in response to tissue injury or in female reproductive organs, endothelial cell turnover in the healthy, adult organism is very low. The maintenance of endothelial quiescence is thought to be due to the presence of endogenous negative regulators, because positive regulators are frequently detected in adult tissues in which there is apparently no angiogenesis. The converse is also true, namely, that positive and negative regulators often coexist in tissues in which endothelial cell turnover is increased. This has led to the notion of the "angiogenic switch," in which endothelial activation status is determined by a balance between positive and negative regulators: in activated (angiogenic) endothelium, positive regulators predominate, whereas endothelial quiescence is achieved and maintained by the dominance of negative regulators. The switch involves either the induction of a positive regulator and/or the loss of a negative regulator.9 10 Used initially in the context of tumor progression to describe the passage from the prevascular to the vascular phase, the notion of the switch can also be applied to developmental, physiological, and pathological angiogenesis.
Positive regulators of angiogenesis include, but are not limited to, the vascular endothelial growth factor (VEGF) and fibroblast growth factor (FGF) families of cytokines and their tyrosine kinase receptors.11 12 13 14 Interestingly, it has been observed that VEGF and basic FGF (bFGF, or FGF-2) induce a synergistic angiogenic response, both in vitro15 16 and in vivo.17 18 A number of endogenous negative regulators have been described, including inhibitors of extracellular proteinases, thrombospondins 1 and 2, and bioactive fragments of the ECM and other molecules (see below).1 19 20 21
The notion of the "balance" applied to positive and negative regulators of angiogenesis can also be applied to extracellular proteolysis. The resulting notion of the "proteolytic balance,"5 22 which highlights the importance of a precise protease-antiprotease equilibrium, has become one of the cornerstones on which our understanding of the molecular mechanisms of angiogenesis is based. Thus, although it has consistently been observed that endothelial cell migration and invasion are associated with increased proteolysis, protease inhibitors play an important albeit permissive role during angiogenesis by preserving matrix integrity. For example, in wound healing and inflammatory skin lesions, protease inhibitors have been localized to stromal cells surrounding vessels in regions of active vascular proliferation, where they may serve to prevent uncontrolled matrix destruction.23 Excessive proteolysis is incompatible with normal angiogenesis, because it results either in destruction of the matrix scaffold that is required for invasion or in the formation of aberrant vascular structures.24 In the absence of sufficient proteolysis, on the other hand, invasion and lumen formation are inhibited.25
Another important notion is that of spatial restriction of extracellular proteolysis. The objective of spatial restriction is to amplify cell surface activity, while at the same time limiting proteolysis to the immediate pericellular environment. This objective is achieved through the local production of protease inhibitors as outlined above, as well as through the assembly of components of one or more protease systems on the cell surface, which brings the enzymes into close proximity to one another. The result is a highly efficient proteolytic cascade on the cell surface, which increases the efficiency of invasion without destroying the matrix scaffold required for invasion (for example, see Mazzieri et al26 ).
In addition to its requirement for matrix invasion and
capillary morphogenesis, extracellular proteolysis has been implicated
in the regulation of cytokine activity
(the
Table
). Examples include activation of latent
transforming growth factor-ß
(TGF-ß),27 release of
matrix-bound bFGF and
VEGF,27 28
release of membrane-anchored tumor necrosis factor-
(TNF-
),29 and
inactivation of monocyte chemoattractant protein 3, with the
resultant generation of a chemokine
antagonist.30
Finally, extracellular proteolysis has been implicated in
the generation of proteolytic fragments of the ECM and other molecules,
a number of which have been reported to have angioregulatory activity,
either positive or negative
(the
Table
).31
Much attention has been focused on angiostatin and endostatin, which
are negative regulators derived from plasminogen and
collagen XVIII,
respectively.32 33
Fragments of collagen IV, which are also negative regulators, include
arresten, from the
1
chain,34 canstatin, from the
2 chain,35 and tumstatin,
from the
3
chain.36 37 Other
ECM proteins from which fragments with angioregulatory activity have
been derived include thrombospondin (inhibitory),
hyaluronan (stimulatory), SPARC (secreted protein, acidic, rich in
cysteine; stimulatory and inhibitory), and collagen
XV/restin
(inhibitory).38
Non-ECM molecules from which fragments with angioinhibitory
activity have been derived include prolactin, platelet factor 4,
MMP-2 (PEX; see below), calreticulin
(vasostatin),39 and
high-molecular-weight kininogen domain 5
(kininostatin).40 Finally, a
cleaved form of antithrombin has also been shown to have antiangiogenic
activity.41 It should be
pointed out, however, that despite a large amount of information that
has accumulated on their antiangiogenic activity, a degree of caution
is still required when interpreting the effects of many of these
molecules, including the mechanisms and precise molecular nature of
their biological activity.
| Role of the MMP System in Angiogenesis |
|---|
|
|
|---|
At least 5 genetically distinct MT-MMPs have been identified (MMP-14, -15, -16, -17, -21). With the exception of MMP-17 (which is glycosyl-phosphatidyl inositol (GPI)-anchored), these enzymes are single-pass type I membrane proteins with an extracellular N-terminus and a short, cytoplasmic C-terminal domain. MT-MMPs are activated intracellularly in the secretory pathway by furinlike enzymes. By virtue of its localization and activity, MT1-MMP is endowed with 2 characteristics important for cell migration: association with the plasma membrane, which focuses matrix digestion close to the cell surface, and the capacity to catalytically activate the precursor of another extracellular protease, namely MMP-2, which allows for amplification of the degradative process. MT1-MMP also has direct activity against a number of ECM proteins, including gelatin, fibronectin, vitronectin, fibrillar collagens, and aggrecan.44 45
In addition to zymogen activation, MMP activity is subject
to regulation by tissue inhibitors of metalloproteases
(TIMPs), four of which have thus far been described. TIMPs form a 1:1
stoichiometric complex with all activated MMPs, with which they
become covalently linked. TIMP-1 also binds proMMP-9. TIMP-2 and -4
also bind proMMP-2. TIMP-2 and -3 are effective
inhibitors of MT-MMPs, whereas TIMP-3 is a good
inhibitor of TNF-
converting enzyme. TIMP-3 is
deposited in the ECM and is notoriously difficult to extract from
tissues.46
The functional interaction between the two classes of MMPs
has been the subject of intense research over the past few years. For
example, MT1-MMP is a cellular receptor for and activator
of proMMP-2, with which it forms a trimolecular complex on the cell
surface with
TIMP-2.44 45 The
importance of MMP-2 and MT1-MMP during angiogenesis has been
reviewed.47 MMP-2, MT1-MMP,
TIMP-2, and integrin
vß3 have been
colocalized in caveolae in the basolateral compartment of human
endothelial
cells.48 In contrast, MMP-9
and TIMP-1 are uniformly distributed on the cell surface and are
concentrated in the Golgi region. Interestingly, it has been
demonstrated that TIMP-2 and -3 derived from perivascular cells
(pericytes and smooth muscle cells) are capable of inhibiting the
activation of MMP-2 by MT1-MMP in endothelial
cells.49 This interplay may
be important during the resolution phase of angiogenesis and for the
maintenance of endothelial
quiescence.
With regard to localization studies, basal levels of expression of MMPs and TIMPs are absent or weakly positive in endothelial cells of normal tissues. However, these molecules are upregulated in endothelial cells (and in some instances, in pericytes) in a variety of physiological and pathological settings. Examples in which MMPs and TIMPs have been detected in endothelial cells include embryogenesis (MMP-1, -2), wound healing (MMP-2), endometrial blood vessels during the menstrual cycle (MMP-2, -3, -9; TIMP-1, -2), placental microvessels (MMP-2, -9; TIMP-1), rheumatoid arthritis (MMP-9), and a variety of tumors, including breast (MMP-2), endometrial (MMP-2, -9), hepatocellular (MMP-2, TIMP-2), colorectal (MMP-7), pancreatic (MMP-7), gastric (MMP-7), renal (MMP-7), and lung (MMP-7) carcinomas, as well as various brain tumors (MMP-2, -9, -14).50 51 52 53 54 55 56 57 58 59 60 61
Cultured endothelial cells have been shown
to express MMP-1, -2, -3, -9, and -14 and TIMP-1 and
-2.62 Studies on the
regulation of MMPs and TIMPs in endothelial cells by
angiogenic cytokines have revealed a variable response that
is in part cell-type dependent. Thus, VEGF increased MMP-2 and MT-1
MMP, had no effect on MMP-1, and decreased TIMP-1 and -2 in human
dermal microvascular endothelial
cells.63 64 VEGF
increased MMP-1, MMP-2, and TIMP-1 in human umbilical vein
endothelial cells, and this change was accompanied by
increased activation of
MMP-2.65 66 bFGF
induced MMP-1 in human
omentumderived67 or
dermal68 microvascular
endothelial cells. As a rule, the magnitude of the
increase in MMP gene expression in response to angiogenic
cytokines is significantly less than the increase seen with
components of the PA-plasmin system
(Figure 1A
). Furthermore, expression of naturally occurring
inhibitors of the MMP and PA-plasmin systems are not
necessarily coordinately regulated: in response to TGF-ß1,
plasminogen activator inhibitor 1
(PAI-1) is increased whereas TIMP-1 is decreased
(Figures 1B
and 1C
). In addition to regulation by angiogenic
cytokines, endothelial cell MMP-2 and MT1-MMP
expression is markedly increased when endothelial cells
are cultured in suspension in a 3-dimensional collagen gel compared
with monolayer (2-dimensional)
cultures,69 which in turn
results in a 15-fold increase in MMP-2 activity. The collagen-dependent
increase in MT1-MMP is mediated by the transcription factor
Egr-1,70 whereas the
increase in MMP-2 activation appears to require integrin
2ß1.71
Changes in Arg-Gly-Aspdependent integrin clustering and cell shape
have also been shown to alter MMP-2 activation in
endothelial
cells.72 Finally, MMPs -2
and -9 have been found to colocalize with ß1
integrin in focal
contacts.73
|
That MMPs are required for angiogenesis has been firmly established.62 74 75 76 Support for this notion has come mainly from studies in which angiogenesis, or various components thereof, could be inhibited by naturally occurring or synthetic MMP inhibitors. Most studies have addressed the activation (sprouting) phase of angiogenesis. For example, increased expression of TIMPs -1 and -2 inhibited tumor-associated angiogenesis, which in the case of TIMP-2 was associated with a marked reduction in expression of tumor-derived VEGF.77 78 Only recently has the requirement for inhibition of MMP activity during vessel stabilization in the resolution phase of angiogenesis become apparent (induction of endothelial quiescence, increased adhesion, and basement membrane deposition).79 80 81 The presence of unchecked proteolysis during this phase was associated with regression of newly formed vessels.
Although a large number of studies have demonstrated the antiangiogenic effect of synthetic MMP inhibitors,62 virtually all of these inhibitors lack specificity for a single MMP. For example, decreased vessel density and increased tumor cell apoptosis were observed in primary tumors and metastases in mice treated with KB-R7785, which inhibits MMPs -1, -3, and -9.82 Similar effects were observed with N-biphenylsulfonylphenylalanine hydroxamic acid, which potently inhibits MMPs -2, -9, and -14 but not MMPs -1, -3, or -7.83
MMPs have also been implicated in the generation of protein
fragments that have angioinhibitory activity. These
include, but are not limited to, angiostatin and endostatin. Thus,
certain MMPs, including metalloelastase (MMP-12), gelatinase B
(MMP-9), stromelysin-1 (MMP-3), and matrilysin (MMP-7), are capable of
converting plasminogen into the angiogenesis-inhibiting
angiostatin.84 85 86 87
Interestingly, it has been reported that tissue levels of MMPs -7 and
-9 are increased in
1-integrinnull mice and
that these mice have increased circulating levels of
angiostatin.88 Tumor growth
and vascularization were decreased in these mice, and this finding was
attributed to the increase in circulating angiostatin levels.
Endostatin is generated from the NC1 domain of collagen XVIII by
several different proteinases (principally elastase and cathepsin
L, but also several MMPs and other
cathepsins).89 90 91
| Requirement for Gelatinases and MT1-MMP During Angiogenesis |
|---|
|
|
|---|
The C-terminal
hemopexin-like domain of MMP-2 (termed PEX) has been shown to inhibit
bFGF-induced angiogenesis in the chicken chorioallantoic membrane (CAM)
and to disrupt tumor
growth.96 The antiangiogenic
activity of PEX has been demonstrated with both a recombinant protein
and a lentiviral delivery
approach.96 97
Physiologically active levels of PEX can
purportedly be recovered from sites of active neovascularization in
vivo. Not surprisingly, PEX prevents cell surface collagenolytic
activity by blocking MMP-2 activation and enzyme
activity.96 It has also been
reported that PEX prevents binding of MMP-2 to the integrin
vß3.96
An organic compound (TSR1359) selected for its ability to inhibit MMP-2
binding to
vß3 has
also been reported to have potent antiangiogenic activity in the CAM
assay.98
Additional evidence for a requirement for gelatinases during angiogenesis has come from genetic studies in mice. With regard to MMP knockouts, although mice deficient in MMP-299 and MMP-9100 are viable, both have a clear angiogenic phenotype. Thus, in MMP-2deficient mice, tumor-induced angiogenesis was markedly reduced in a dorsal air sac assay with B16-BL6 melanoma cells.101 This finding indicates that host-derived MMP-2 plays an important role in angiogenesis and tumor progression, with an important contribution from stromal cells. In MMP-9deficient mice, bone growth plate angiogenesis was reduced, probably owing to delayed release of a positive angiogenesis regulator by hypertrophic cartilage.100
It has recently been reported that tumor angiogenesis is VEGF dependent in a transgenic mouse model of pancreatic beta-cell carcinogenesis (Rip1Tag2 mice). However, in these mice, there is no net increase in the level of expression of VEGF or its tyrosine kinase receptors. Instead, extracellular VEGF is mobilized from matrix stores, making it more available to VEGF receptors. This mobilization is dependent on MMP-9 activity (probably macrophage derived), which is upregulated during tumor progression, more specifically, during the switch from the prevascular to the vascular phase.102 Tumor progression was inhibited in Rip1Tag2xMMP-9null double-transgenic mice. In contrast, although MMP-2 activity was increased in parallel with MMP-9 in Rip1Tag2 mice, tumor progression was unaffected in Rip1Tag2xMMP-2null double transgenics. With regard to the PAs (urokinase-type PA [uPA] and tissue-type PA [tPA]), no reproducible changes in activity were noted during tumor progression, and tumor progression was unaffected in Rip1Tag2xuPAnull double-transgenic mice. Taken together, these findings reveal a specific requirement for MMP-9 in the angiogenic switch in this murine model of pancreatic beta-cell carcinogenesis.
MT1-MMPdeficient mice display severe skeletal defects that lead to death by 3 to 16 weeks of age.103 104 Other tissues in these null mice develop relatively normally. The skeletal defects are due in part to decreased vascular invasion of calcified cartilage. In the corneal angiogenesis assay, MT1-MMPdeficient mice also failed to respond to implanted FGF-2. Reduced activation of latent MMP-2 (but not MMP-9) was observed in various tissues of MT1-MMPnull mice.
| MMP Requirement for Fibrinolysis During Angiogenesis |
|---|
|
|
|---|
These unexpected and highly novel observations suggest that when endothelial cells invade a provisional fibrin ECM scaffold, this event occurs in a plasmin-independent manner and may require the activity of MMPs such as MT1-MMP and possibly MMP-9. These observations have potentially far-reaching mechanistic and therapeutic implications.
| Role of the PA-Plasmin System in Angiogenesis |
|---|
|
|
|---|
2-antiplasmin) and PAs (ie, PAI-1 and PAI-2)
provides additional points of regulation along this protease cascade.
The specificity, activity, and localization of serine protease
inhibitors can be regulated by the ECM and other
glycoproteins.3 105 With regard to localization studies, the overall picture is that uPA, uPAR, and PAI-1 are not expressed by quiescent endothelium. tPA, on the other hand, has been detected in the quiescent endothelium of a small percentage of arterioles, venules, and vasa vasorum of normal human tissues. In contrast, uPA, uPAR, and PAI-1 are all expressed during angiogenesis in vivo in a variety of settings. uPA and uPAR appear to be expressed by endothelial cells, and, depending on the situation, PAI-1 is expressed either by endothelial cells themselves or by nonendothelial (stromal and epithelial) cells, where it is thought to play a role in the preservation of matrix integrity.5 109 Little is known about tPA expression during angiogenesis.
These in vivo observations are amply supported by results
obtained in vitro. That cultured endothelial cells
(which are closer to activated than quiescent
endothelium) produce uPA, uPAR, tPA, and PAI-1 has been
known for some time, and their regulation by bFGF and VEGF has been
described
(Figure 2
).2 5
It should be pointed out, however, that endothelial
cells of human origin respond less well than do cells of bovine origin
with regard to induction of PAs by bFGF and VEGF. The reason for the
lack of response may be related to the chronic stimulation of these
cells by the components of endothelial growth
supplements (FGFs and probably VEGF) used to culture the cells. The
response to VEGF also appears to be dependent on the vascular bed from
which the cells were isolated. Thus, VEGF induces uPA and tPA in
endothelial cells derived from the microvasculature but
not in cells derived from the aorta
(Figure 3
).110 111
This distinction may have important implications for the regulation of
angiogenesis, which traditionally has been considered to occur from
preexisting vessels of the microvasculature. Hypoxia, a major
stimulus for angiogenesis, has been reported to increase
uPAR112 and PAI-1
(Figure 4
)113 in
endothelial cells. We have previously demonstrated that
uPA, uPAR, and PAI-1 are increased in endothelial cells
in a 2-dimensional wound-induced migration assay and that both
migration and protease/inhibitor induction are dependent on
endogenous
bFGF.114 115 116
In a coculture system of endothelial cells and
fibroblasts, PAI-1 mRNA and promoter activity were induced only in
those fibroblasts apposed to sprouting but not resting
endothelium.117
The authors reasoned that this may be of importance during sprouting,
which constitutes the only period during which
endothelial cells establish direct contact with
fibroblasts. They suggested that the paracrine induction of PAI-1
provides a mechanism to negate excessive pericellular
proteolysis.
|
|
|
Angiogenesis results in the formation of new vessels in the blood vascular system. A comparable process, called lymphangiogenesis, is responsible for the formation of new vessels in the lymphatic system. Lymphangiogenesis has traditionally been overshadowed by the greater emphasis placed on angiogenesis. This is due in part to the lack of identification of lymphangiogenic factors as well as suitable markers with which to distinguish blood from lymphatic vascular endothelium. This scenario is changing rapidly after discovery of the first lymphangiogenic factors (VEGF-C and -D) as well as a limited number of lymphatic endothelium-specific markers.118 In adult tissues, new lymphatic capillaries grow by sprouting as extensions from preexisting lymphatics, much in the same way as new blood capillaries arise by sprouting from the preexisting microvasculature during angiogenesis. Although very little is known about the mechanisms of lymphangiogenesis, it has been demonstrated that VEGF, VEGF-C, and bFGF are capable of stimulating uPA, uPAR, tPA, and PAI-1 in bovine large-vessel lymphatic endothelial cells.119 120 121 122 The relevance of these in vitro observations to lymphangiogenesis in vivo remains to be determined.
The generation of mice deficient in components of the PA-plasmin system109 provided an opportunity to assess the role of individual components of this system during angiogenesis in vivo. Previous studies, both in vivo and in vitro, had pointed to a crucial role for the PA-plasmin system during angiogenesis. Unexpectedly, however, it was found that embryonic and postnatal development was unperturbed in these mice, at least from the perspective of an angiogenesis-associated phenotype.5 A seminal study on wound healing in plasminogen-deficient mice clearly defined a role for the PA-plasmin system during keratinocyte migration in vivo.123 However, although keratinocyte migration was severely impaired, there were no overt qualitative differences in the granulation tissue (inflammatory cell infiltration, fibroblast migration, and neovascularization) between wild-type and plasminogen-/- mice, suggesting that in this setting, angiogenesis is not PA-plasmin dependent. In addition, expression of 6 MMPs, including MMPs -1, -3, and -9 by keratinocytes and MMPs -2 and -11 and MT1-MMP by cells in granulation tissue, was unaffected in plasminogen-/- mice. These findings suggest that although keratinocyte migration is plasmin dependent, implementation of the genetic program that controls keratinocyte migration is plasmin independent.
Although developmental and wound healingassociated angiogenesis appears to be unaffected in uPA-, uPAR-, tPA-, PAI-1, and plasminogen-null mice, several in vivo studies have demonstrated a requirement for this system during angiogenesis in other settings. For example, tumor-associated angiogenesis (vascular density) was inhibited in an athymic mouse model utilizing a human prostate carcinoma cell line, PC-3, that was stably transfected with human PAI-1.124 Although this study pointed to a role for PAs in tumor angiogenesis, it provided no mechanistic information, including the cell type specificity of the effect. More precise information has been provided in a murine model of myocardial infarction, in which postinfarction myocardial revascularization (angiogenesis) was severely impaired in uPA-/- mice.125 Because this phenotype could not be rescued by exogenously added VEGF, and because endothelial and smooth muscle cells expressed uPA during revascularization, it was concluded that the defect in angiogenesis was due to impaired endothelial cell invasion. Interestingly, when assayed in the rabbit cornea or on the chicken CAM, uPA itself had a proangiogenic effect.126 127 128 In the CAM, this event could be ascribed to the release of endogenous bFGF.128
uPA-uPAR Interactions Are Required for
Angiogenesis
The importance of the interaction between uPA and uPAR
during angiogenesis has been demonstrated in a number of in vivo
systems. Thus, bFGF-induced angiogenesis in subcutaneously injected
Matrigel was inhibited by a fusion protein consisting of the
receptor-binding amino-terminal fragment of uPA and the Fc portion of
human IgG, which functions as a high-affinity uPAR
antagonist.129
Similarly, in syngeneic and xenograft murine tumor models, adenovirally
delivered amino-terminal fragment specifically inhibited tumor
angiogenesis.130 In a
homologous immunocompetent rat model of tumor cells transfected with a
mutant murine uPA that retains receptor binding but not proteolytic
activity, microvessel density was markedly
reduced.131 Tumor
angiogenesis in several murine models was also inhibited by an octamer
peptide derived from the nonreceptor-binding region of
uPA.132 133
These studies defined a novel epitope in uPA involved in the uPA-uPAR
interaction.
PAI-1 Is Required for Angiogenesis
Localization studies have demonstrated that PAI-1 is
strongly expressed in endothelial cells in a number of
tumor types.5 Studies in
PAI-1null mice have revealed an absolute requirement for PAI-1 in
tumor-induced angiogenesis. In one model, tumor and stromal cells are
separated by the interposition of a type I collagen gel, which allows
independent assessment of exogenously added tumor cell as well as of
host stromal cell
invasion.134 135
The absence of PAI-1 (but not of uPA, uPAR, or tPA) from the host
markedly impaired tumor cell invasion and host-derived
neovascularization. Reversal of this phenotype was achieved by
adenovirally delivered human PAI-1 to PAI-1null mice. In wild-type
mice, PAI-1 expression in the stroma could be localized to
endothelial and nonendothelial cells.
Interestingly, although uPA was localized to newly formed blood vessels
in wild-type mice, uPA was replaced by tPA in mice deficient in uPA,
demonstrating the capacity for compensation between the PAs. The
requirement for host PAI-1 (but not uPA) during tumor angiogenesis has
also been demonstrated in a fibrosarcoma model in gene inactivation
mice.136 bFGF-induced
corneal angiogenesis was similarly impeded in PAI-1 but not in
uPA-deficient mice.
Two potential mechanisms may explain why PAI-1 is essential for angiogenesis. First, by protecting the ECM against excessive degradation, PAI-1 may serve to maintain the matrix scaffold required for endothelial cell migration and tube formation.22 Second, a complex series of interactions have been described between PAI-1, uPAR, integrins, and the ECM component vitronectin.137 138 139 Alterations in PAI-1 expression and activity would therefore be expected to alter the adhesive, migratory, and growth properties of endothelial cells, which in turn would regulate angiogenesis. To address this question directly, adenovirus was used to deliver two mutant forms of PAI-1 to PAI-1deficient mice in the collagen gel interposition model described above.135 The PAI-1 mutants either (1) bound vitronectin normally but failed to inhibit uPA or tPA or (2) inhibited PAs normally but had negligible binding to vitronectin. Angiogenesis was restored by the PAI-1 mutant that retained PA inhibitory activity but not by the mutant that bound vitronectin but failed to inhibit the PAs, demonstrating that the requirement for PAI-1 in this model is due to inhibition of excessive proteolysis rather than inhibition of cellular adhesion to vitronectin. In a separate study, exogenously added PAI-1 ("at therapeutic concentrations") was shown to inhibit bFGF-induced angiogenesis in the CAM.140 With use of a variety of PAI-1 mutants, this activity could be ascribed both to PAI-1s antiprotease activity as well as its vitronectin-binding capacity.
Does tPA Play a Role in Angiogenesis?
Little is known about the role of tPA in angiogenesis.
In vivo, tPA antigen is localized to the endothelium of
the microvasculature,141
which, as morphological studies have revealed, is the region of the
vascular bed from which new blood vessels are formed during
angiogenesis. The presence of tPA in angiogenic
endothelium has yet to be described. In vitro, tPA has
been implicated in capillary-like tube formation on matrices
constituted of type I collagen, Matrigel, and the explanted
amnion.142 143 144 145 146
The requirement for tPA in endothelial cell migration
in vitro has been demonstrated in some
studies147 but not in
others.148 Interestingly,
the combination of bFGF and VEGF induced tPA mRNA and enzyme activity
in a synergistic manner in cultured endothelial
cells.111 In this study,
immunofluorescence analysis revealed that
cell-associated tPA was colocalized with von Willebrand factor,
a marker for Weibel-Palade bodies, which are regulated,
endothelial cellspecific storage granules. A
combination of bFGF and VEGF increased the number of tPA-positive cells
as well as the number of tPA-positive granules per cell. Angiostatin,
an endogenous inhibitor of angiogenesis, has
been shown to inhibit tPA-catalyzed plasminogen activation
in endothelial
cells.149
At present, one can only speculate on the function of tPA in Weibel-Palade bodies in the context of angiogenesis. An attractive role would be to protect newly forming blood vessels from thrombotic occlusion. Indeed, stagnant blood in the forming dead end of new blood vessels would be particularly prone to coagulation, first because of the potential exposure of blood to the ECM (and in particular, to collagen) in the hyperpermeable and rapidly remodeling migrating sprout and second, because activated blood coagulation would be less easily washed away. To overcome this problem, locally generated thrombin would result in the acute release of tPA from Weibel-Palade bodies,150 151 which in turn would aid in the removal of fibrin deposits.
It is of interest to note that secretion of MMP-1 and MMP-2, but not of TIMP-2, can occur through a posttranscriptional mechanism in endothelial cells.152 MMPs are known to be regulated at different levels; this study indicates that, in endothelial cells, the stimulation of MMPs can also occur at the level of secretion, providing a mechanism for their rapid mobilization in the early phases of angiogenesis. Whether these enzymes, like tPA, are localized to Weibel-Palade bodies (regulated, endothelial cellspecific storage granules) remains to be determined. The intracellular accumulation of MMP-9 in secretory vesicles in response to phorbol esters has been reported.71
| Requirement for the MMP and PA-Plasmin Systems in Polyomavirus Middle T OncogeneInduced Vascular Tumor Formation |
|---|
|
|
|---|
PymT-transformed endothelial cells (End. cells), which can be isolated with relative ease, have the potential to grow indefinitely, and they retain many important features of differentiated endothelium. Endotheliomas can be induced not only by PymT itself but also by subcutaneous or intraperitoneal injection of End. cells. The following is a working hypothesis for endothelioma formation after subcutaneous injection.153 End. cells invade and integrate into host blood vessels, leading to their rupture and subsequent hemorrhage. This event in turn is followed by rapid recruitment of host endothelial and nonendothelial cells. Recruited endothelial cells form new capillary blood vessels, which feed the growing cystic tumor. Recruitment is dependent on the continuous presence of End. cells and mainly involves host cell migration with low levels of proliferation. The signal for host endothelial cell recruitment is not known, but it probably originates from the injected End. cells as well as from the intense inflammatory reaction observed at the periphery of the hemorrhagic cysts. It is likely that in addition to host cell recruitment, proliferating End. cells themselves contribute significantly to tumor expansion during the later phases of tumor growth. Taraboletti et al154 and Vergani et al155 have partially characterized an End. cellderived motility factor that may be responsible for recruitment of host endothelial cells.
An in vitro model has been developed to study the
autonomous, morphogenetic behavior of End.
cells.24 This consists of
growing End. cells in suspension in 3-dimensional fibrin gels. In these
fibrin gels, End. cells proliferate rapidly, and after 10 to 14 days
form large cystic structures lined by a monolayer of
endothelial cells
(Figure 5
). These cystlike structures are strikingly
reminiscent of the cavernous endotheliomas that develop in vivo. Under
the same experimental conditions, nontransformed
endothelial cells from a variety of species and organs
form elongated, branching, tubelike structures
(Figure 5
); in no instance are cystlike structures seen. The
fibrin gel model, in which normal vascular morphogenesis is disrupted,
ie, formation of endothelium-lined cystlike structures
rather than tubelike structures, provides a powerful in vitro system
for investigating the mechanisms of protease-dependent tumor
formation.
|
With regard to MMPs and endothelioma formation in vivo, it has been demonstrated that End. cellinduced endothelioma growth can be inhibited by the broad-spectrum hydroxamic acidbased peptide derivative Batimastat (also known as BB94).156 Daily local injection of the inhibitor at the site of End. cell injection increased tumor doubling time in a dose-dependent manner. Inhibition of tumor growth lasted as long as Batimastat was given, and when treatment was stopped, the tumors resumed a pattern of growth similar to that seen in controls. Histologically, treated tumors consisted of solid, cellular aggregates that circumscribed small- to medium-size blood-filled spaces. Subcutaneous hemorrhage was rarely observed in treated animals. In a subsequent study, these authors demonstrated that tumor growth could be efficiently inhibited by overexpressing TIMP-2 in End. cells.155 When compared with the cavernous tumors induced by wild-type cells, tumors induced by TIMP-2 overproducers had small-to-medium blood-filled spaces and multiple apoptotic foci.
With regard to the PA-plasmin system, End. cells display increased PA activity when compared with nontransformed endothelial cells, which can be accounted for by an increase in uPA and, in some cell lines, by a concomitant decrease in PAI-1.24 157 158 159 The mechanism for this increase is not known, although it has been shown that PymT increases uPA gene activation without influencing uPA mRNA stability.158 Taken together with the observation that transformed endothelial cells form cystlike structures when embedded in fibrin gels (see above), these findings demonstrate that excessive proteolysis is associated with aberrant vascular morphogenesis. Reduction of this activity by inhibiting plasmin restores normal morphogenetic properties to these cells, ie, the formation of capillary-like tubes. These findings clearly implicate increased PA-plasminmediated proteolysis in aberrant vascular morphogenesis in vitro.24
Based on these in vitro observations, it was hypothesized that the PA-plasmin system would also be required for endothelioma formation in vivo. This hypothesis was tested directly in uPA-, tPA-, PAI-1, and plasminogen-deficient mice. Essentially, tumor formation after viral infection/transduction was unaffected in the different genotypes, suggesting that the primary transformation process is PA-plasmin independent. However, compared to tumors induced by injection of wild-type End. cells, tumor incidence and latency were increased in uPA-/- and double-transgenic uPA:tPA-/- mice but not in tPA-/- or PAI-1-/- mice. Tumor incidence and latency were also increased when uPA-/- and uPA:tPA-/- End. cells were injected into wild-type mice.160 These genetic studies demonstrated that components of the PA-plasmin system, in particular uPA, are critical for tumor growth in vivo. With respect to the in vitro morphogenetic behavior of End. cells derived from protease-knockout mice, although PA activity was necessary, it did not appear to be sufficient for cyst formation.160
The plasminogen-dependent and -independent proteolytic activity of wild-type and mutant End. cells has been determined.161 Thus, plasminogen activation with wild-type and tPA-/- cells was comparable but was reduced 4-fold with uPA-/- cells and increased 3-fold with PAI-1-/- cells. Lysis of an 125I-labeled matrix in the presence of plasminogen was comparable in wild-type, tPA-/-, and PAI-1-/- cells but was significantly reduced in uPA-/- cells. Lysis of a human umbilical vein endothelial cellderived, [3H]proline-labeled ECM in the presence of plasminogen with wild-type, tPA-/-, and PAI-1-/- cells was comparable but was virtually abolished in uPA-/- cells. These findings provide a mechanism for the reduction in tumor growth seen in the absence of uPA (see above).
Finally, using the in vitro fibrin gel model, we have infrequently observed the formation of cystlike structures by End. cells deficient in both uPA and tPA (Pepper et al, unpublished observations, 2000). Clonal populations of these cells have been obtained and in which cyst formation can be completely inhibited by synthetic MMP inhibitors (Pepper et al, unpublished observations, 2000). This result implicates another protease system(s) (possibly MMPs) in the fibrinolysis required for cyst formation and is in accord with the emerging notion of MMP-mediated fibrinolysis in endothelial cells.
| Conclusions and Perspectives |
|---|
|
|
|---|
With respect to MMPs, potential targets for inhibiting angiogenesis include the gelatinases and MT-MMPs. The additional role of MT1-MMP in fibrinolysis during angiogenesis makes this a particularly attractive target. Several novel synthetic inhibitors are currently being investigated in clinical trials.163 164 However, despite the abundance of promising preclinical data, the clinical application of MMP inhibitors has to date met with limited success, and a number of phase III trials have been discontinued.165 Because MMPs have been implicated in the generation of molecules with antiangiogenic activity (eg, endostatin), MMP inhibition may result in the stimulation rather than inhibition of angiogenesis. Furthermore, one hydroxamate-based MMP inhibitor has been reported to promote the formation of liver metastases in a murine breast cancer xenotransplantation model,166 and systemic administration of human TIMP-4 stimulates the growth of human breast cancer cells in a murine xenotransplantation model.167 At present, therefore, a degree of uncertainty surrounds the use of MMP inhibitors in cancer, and it appears that under certain circumstances, MMP inhibition may even be detrimental to cancer patients.
The studies on the PA-plasmin system cited in this review provide ample evidence for the requirement for this system during angiogenesis in vivo. In particular, a role for uPAR and PAI-1 has been clearly defined, with uPA playing a more ancillary role. The counterintuitive possibility that inhibition of PAI-1 might serve to inhibit angiogenesis is one that merits further investigation. To date, little is known about the role of tPA in angiogenesis.
The tumor vasculature has become a key target in the war against cancer.168 Increased understanding of the complex roles of the MMP and PA-plasmin systems in the regulation of angiogenesis will provide a rational basis for the development of antiangiogenesis-based therapeutic strategies for cancer and other angiogenesis-associated diseases.
| Acknowledgments |
|---|
Received April 6, 2001; accepted May 25, 2001.
| References |
|---|
|
|
|---|