Abstract
purpose. Vascular endothelial growth factor (VEGF) is upregulated by
hypoxia and is a major stimulatory factor for retinal
neovascularization in ischemic retinopathies such as diabetic
retinopathy. This study sought to determine if VEGF is a stimulatory
factor in a murine model of choroidal neovascularization (CNV).
methods. Mice with laser-induced ruptures in Bruch’s membrane were treated with
vehicle alone; a drug that inhibits both VEGF and platelet-derived
growth factor (PDGF) receptor kinases; a drug that inhibits PDGF, but
not VEGF receptor kinase; or genistein, a nonspecific kinase inhibitor.
After two weeks, CNV was quantified and compared. results. Blockade of phosphorylation by VEGF and PDGF receptors caused dramatic,
almost complete inhibition of CNV. Genistein also had an inhibitory
effect, but less so than the VEGF/PDGF receptor blocker. Blockade of
phosphorylation by PDGF receptors, but not VEGF receptors, had no
significant effect on cnv. conclusions. These data and our previous study, which demonstrated that a kinase
inhibitor that blocks VEGF and PDGF receptors and several isoforms of
protein kinase C causing dramatic inhibition of CNV, suggest that VEGF
signaling plays a critical role in the development of CNV in this
model. If safety is established, the effect of inhibiting VEGF receptor
kinase activity should be investigated in patients with CNV.
Ocular neovascularization is responsible for the majority of
cases of new blindness in developed countries every year. Retinal and
iris neovascularization occur in the same disease processes, including
diabetic retinopathy, the most common cause of severe loss of vision in
young people.
1 2 Choroidal neovascularization (CNV) occurs
in diseases in which there are abnormalities of Bruch’s membrane and
the retinal pigmented epithelium (RPE). The most common disease of this
type is age-related macular degeneration (AMD), the most prevalent
cause of severe loss of vision in patients more than 60 years of
age.
3
The pathogenesis of retinal neovascularization is much better
understood than the pathogenesis of CNV. Occlusion of retinal vessels
leading to ischemia or hypoxia is a common feature of diseases in which
retinal neovascularization occurs.
4 5 6 Another
well-established fact is that one of the stimulatory factors involved
is vascular endothelial growth factor (VEGF). It is unlikely to be the
only stimulatory factor, because insulin-like growth factor-1 may also
participate,
7 but there is strong evidence indicating that
VEGF plays a central role. It is upregulated by
hypoxia,
8 9 10 and its levels are increased in the retina
and vitreous of patients
11 12 13 14 or laboratory
animals
15 16 with ischemic retinopathies. Increased
expression of VEGF in retinal photoreceptors of transgenic mice
stimulates neovascularization within the retina,
17 18 and
VEGF antagonists partially inhibit retinal or iris neovascularization
in animal models.
19 20 21
It is not known whether VEGF is also a stimulatory factor for CNV.
Unlike retinal neovascularization, it is not clear that hypoxia, a
principal cause of upregulation of VEGF,
10 plays a role in
CNV. There is some suggestion that choroidal blood flow may be altered
in patients with AMD,
22 23 but it is not known whether
this is sufficient to cause hypoxia. Also, it is unlikely that hypoxia
is present in other disease processes, such as ocular histoplasmosis,
in which CNV occurs in young patients. However, increased expression of
VEGF has been demonstrated in CNV removed from
patients
24 25 26 27 and in experimentally induced
CNV.
28 29 But increased expression in association with CNV
is not unique to VEGF, because similar data have been obtained for
fibroblast growth factor (FGF)-2.
24 25 28 29 . Also,
choroidal vessels differ from retinal vessels in several respects, and
it is not clear that they are responsive to VEGF. In transgenic mice
that express VEGF in photoreceptors, new vessel sprouts develop from
retinal vessels, but not from choroidal vessels.
17 Therefore, it is not known whether VEGF is a stimulatory factor for
CNV.
Growth factors such as VEGF act by binding to specific receptors and
inducing them to form receptor dimers. The intracellular portion of
receptors have kinase domains that phosphorylate other proteins, and
each component of a ligand-bound receptor pair phosphorylates its
partner, a process known as receptor autophosphorylation.
Autophosphorylation increases the affinity of receptor kinases for
second messengers that are phosphorylated to propagate the signal that
ultimately results in altered gene expression. Therefore, the kinase
activity of a receptor is absolutely necessary for its signaling
capability and for any biologic process the receptor mediates.
Our laboratory has recently demonstrated striking inhibition of CNV in
a murine model using a selective kinase inhibitor that blocks
phosphorylation by VEGF and platelet-derived growth factor (PDGF)
receptors and several isoforms of protein kinase C
(PKC).
30 In this study, we have used other kinase
inhibitors with overlapping, but different activities to explore the
signaling pathways involved in the development of CNV.
C57BL/6J mice were treated in accordance with the ARVO Statement
for the Use of Animals in Ophthalmic and Vision Research. CNV was
generated by laser-induced rupture of Bruch’s membrane, as previously
described.
31 Briefly, 4- to 5-week-old male C57BL/6J mice
were anesthetized with ketamine hydrochloride (100 mg/kg body weight)
and the pupils were dilated with 1% tropicamide. One burn of krypton
laser photocoagulation (100-μm spot size, 0.1-second duration, 150
mW) was delivered to the 9, 12, and 3 o’clock positions of the
posterior pole of the retina in each eye using the slit lamp delivery
system of a photocoagulator (model 920; Coherent, Palo Alto, CA) and a
handheld cover slide as a contact lens. Production of a bubble at the
time of laser exposure, which indicates rupture of Bruch’s membrane,
is an important factor in inducing CNV,
31 and therefore
only mice in which a bubble was produced for all six burns were
included in the study.
Forty mice (10 in each group) were randomly assigned to treatment
by gavage with vehicle alone or with one of three kinase inhibitors.
The first group was treated with vehicle, and the second group was
treated with 50 mg/kg once a day of PTK787/ZK 222584 (PTK787), a
selective inhibitor of VEGF receptor kinases and other class III
kinases, but not kinases from other classes.
32 33 A third
group was treated once a day with the molar equivalent (120
micromoles/kg; 46 mg/kg) of a selective inhibitor of PDGF receptor
kinases.
34 35 The fourth group was treated once a day with
120 micromoles/kg (32 mg/kg) of genistein, a nonselective tyrosine
kinase inhibitor, that also has some effects unrelated to tyrosine
phosphorylation.
36 Mice were treated by gavage on the day
of laser and then once a day for 2 weeks when they were killed and
evaluated.
After 14 days of treatment, the mice were killed with an overdose
of sodium pentobarbital, and their eyes were rapidly removed and frozen
in optimal cutting temperature embedding compound (OCT; Miles, Elkhart,
IN). Frozen serial sections (10 μm) were cut through the entire
extent of each of the three CNV lesions associated with laser sites in
each eye. Sections were histochemically stained as previously
described
37 with biotinylated
Griffonia
simplicifolia lectin B4 (GSA; Vector, Burlingame, CA), which
selectively binds to vascular cells. Slides were incubated in 4%
paraformaldehyde for 30 minutes, washed with 0.05 M Tris buffer, (TB;
pH 7.4), incubated in
methanol-H
2O
2 for 10
minutes at 4°C, washed with 0.05 M TB, and incubated for 30 minutes
in 10% normal swine serum. Slides were rinsed with 0.05 M TB and
incubated 2 hours at 37°C with 1:20 biotinylated lectin. After they
were rinsed with 0.05 M TB, slides were incubated with undiluted
streptavidin-phosphatase (Kirkegaard & Perry, Cabin John, MD) for 30
minutes at room temperature. After a 10-minute wash in 0.05 M TB (pH
7.6) slides were developed with staining (HistoMark Red
38 ;
Kirkegaard & Perry). Some slides were counterstained (Contrast Blue;
Kirkegaard & Perry).
To perform quantitative assessments, GSA-stained sections were
examined at ×400 with an (Axioskop; Carl Zeiss, Thornwood, NY)
microscope with the observer masked to treatment group. Images were
digitized using a three CCD color video camera and a frame grabber.
GSA-stained blood vessels were delineated and the area in the
subretinal space measured by computer with image analysis software
(Image-Pro Plus software; Media Cybernetics, Silver Spring, MD). For
each CNV lesion, area measurements were made for all sections on which
some of the lesion appeared and added together to give the integrated
area measurement. Only CNV lesions in which good sections were obtained
through the entire extent of CNV, so that a valid area measurement
could be made on each, were included in the analysis. Lesions were
excluded based solely on the inability to obtain an accurate
measurement because of poor quality of some sections from that
particular CNV lesion with the observer masked with regard to treatment
group, so that there was no possibility of a bias with regard to lesion
exclusion. After unmasking, it was found that precise measurements had
been obtained on the following number of lesions in each treatment
group: 52 controls, 56 PTK787, 52 CGP 53716, and 52 genistein. There
were 60 CNV lesions in each group, and therefore there were 4 that were
unmeasurable in the PTK787 group and 8 in each of the other three
groups.
Adult C57BL/6J mice (eight mice in each group) were treated by
gavage with vehicle or vehicle containing 120 micromoles/kg per day of
one of the kinase inhibitors. After 2 weeks, mice were killed with an
overdose of pentobarbital sodium, and their eyes were rapidly removed
and frozen in OCT. Serial sections (10 μm) were cut entirely through
each eye and stained with GSA. The retinal vessel area was determined
by image analysis on every 10th section, and the mean was calculated
for each eye as previously described.
37
Because more than one CNV lesion was created in each mouse,
repeated measures analysis of variance was used to compare differences
in the area of CNV lesions among mice, and after adjusting for
differences among mice, comparison of differences in the area of CNV
lesions among treatment groups was performed. To provide a graphic
demonstration that portrays the differences among mice within treatment
groups as well as the differences among treatment groups, box plots
were made for each group showing the median log (CNV area), the 25% to
75% range, and the highest and lowest value. The same analysis was
performed to make statistical comparisons of retinal vascular area for
the toxicity study.
Two weeks after laser photocoagulation, histopathologic evaluation
showed a discontinuity in Bruch’s membrane in the area of each laser
burn in all mice. Control mice that were treated with aqueous buffer
(phosphate-buffered saline [PBS]) or buffer containing 1% dimethyl
sulfoxide, showed large areas of CNV at each site of laser-induced
rupture of Bruch’s membrane (
Fig. 1A ,
1E) . Previous characterization of this model has demonstrated the
origin of the neovascularization to be from choroidal
vessels.
31 There was proliferation of RPE cells along the
margin of the new vessels. The overlying lectin-stained retinal blood
vessels were normal.
Selective kinase inhibitors were used to explore the signaling
pathways involved in the development of CNV. PTK787 is a good inhibitor
of phosphorylation by human VEGF receptor 2 and its mouse homologue,
Flk-1 (50% inhibitory concentrations [IC
50s]:
0.1 and 0.27 μM, respectively).
32 33 It is a less potent
inhibitor of VEGF receptor 1 (IC
50: 0.49 μM)
and also blocks the related tyrosine kinases PDGF β-receptor, c-Kit
(the receptor for stem cell factor), and cfms (the receptor for
macrophage colony stimulating factor-1; IC
50s:
0.2, 0.38, and 1.2 μM, respectively). In concentrations up to 10μ
M, PTK787 does not inhibit kinases from other families including
FGFR1, c-Met, Tie2, epithelial growth factor receptor (EGFR), c-Src,
v-Abl, or PKC-α.
Mice treated with the dihydrochloride salt of PTK787 dissolved in PBS,
showed a dramatic decrease in the size of CNV associated with
laser-induced rupture of Bruch’s membrane
(Figs. 1B 1F) . In many
instances, there was no identifiable lectin-stained neovascular tissue
throughout the entire burn, but some burns contained regions in which
there were thin disks of lectin-stained tissue. There was mild
proliferation of RPE cells. Despite the marked decrease in CNV in the
eyes of treated mice, the overlying retinal vessels appeared normal.
This is best seen in sections with no counterstain
(Fig. 1B) .
Measurement by image analysis of the integrated area of lectin staining
per lesion showed a dramatic decrease in mice treated with PTK787
(0.0134638 ± 0.001799 mm
2) compared with
lesions in mice treated with vehicle (PBS) alone (0.0627643 ±
0.008704 mm
2). This difference was highly
statistically significant (
P < 0.0001 by repeated
measures analysis of variance [ANOVA];
Fig. 2 ).
CGP 53716 is a selective inhibitor of the kinase activity of PDGFβ
-receptors (IC
50: 0.1 μM) and v-Abl
(IC
50: 0.6 μM) and does not inhibit
phosphorylation by other kinases, including VEGF receptors or PKCs in
concentrations greater than 20 μM.
34 35 Mice treated
with CGP 53716, showed large areas of CNV similar to those seen in
control mice
(Figs. 1C 1G) . Measurement of CNV lesion size by image
analysis demonstrated that there was no significant difference in the
integrated area of lectin staining per lesion in CGP 53716–treated
mice compared with vehicle-treated mice (
P = 0.56 by
repeated measures ANOVA).
Mice treated with genistein
(Figs. 1D 1H) , a nonspecific kinase
inhibitor, showed areas of CNV that appeared intermediate in size
between control mice and those treated with PTK787. Image analysis
confirmed that this was the case. Mice treated with genistein had
lesions with an integrated area of lectin staining that was
significantly smaller than that in control mice (
P <
0.0001) but greater than that in PTK787-treated mice (
P < 0.0001;
Fig. 2 ). Expressed as a mathematical relationship, the area
of CNV in the four treatment groups is vehicle control = CGP
53716 > genistein > PTK787.
To investigate for toxic effects on retinal vessels, mice that did
not receive any laser photocoagulation were treated with PTK787, CGP
53716, or genistein for 2 weeks. None of the drugs caused changes in
morphology of retinal vessels (notshown), and there were no
significant differences in retinal vascular area per section compared
with control mice
(Fig. 3) .
Identification of stimulatory factors for a pathologic process
makes it possible to design drug treatments that are likely to inhibit
the process with minimal unwanted side effects. There are several lines
of evidence suggesting that VEGF plays a major role in the stimulation
of retinal neovascularization and therefore many strategies for
inhibiting VEGF signaling are under study as potential
treatments.
19 20 21 Less is known concerning possible
stimulatory factors for CNV, but it has been noted that expression of
FGF2 and VEGF is increased in association with the development of
CNV.
24 25 26 27 28 29 We have recently demonstrated that FGF2 is not
necessary for the development of CNV, suggesting that it does not play
a central role.
31 In this study and another recent
study,
30 we used kinase inhibitors with overlapping but
different activities, to demonstrate that inhibition of VEGF receptor
tyrosine kinase activity dramatically inhibits CNV. This indicates that
VEGF is a stimulator for CNV in this model.
Our experimental paradigm involved initiation of drug treatment right
after laser-induced rupture of Bruch’s membrane. Therefore, it is
relevant to prevention of CNV, but not necessarily regression of
already established CNV. Also, although the murine laser-induced model
does not exactly mimic all aspects of CNV that occurs in association
with AMD, it shares two critical features: The new vessels originate
from choroidal vessels, and there are abnormalities in Bruch’s
membrane. Laser-induced disruption of Bruch’s membrane in humans is
also associated with CNV.
39 40 41 Therefore, our data
provide suggestive, although not definitive, evidence that VEGF is a
stimulatory factor for CNV in humans.
What is the source of VEGF in the setting of CNV? Studies of human
pathologic specimens
25 26 27 as well as a study in a
laser-induced experimental model,
29 suggest that RPE cells
are the source of VEGF, and it is well-established that cultured RPE
cells produce VEGF.
42 But there is no evidence suggesting
that RPE cells are hypoxic in patients in whom CNV develops. Hypoxia of
the RPE seems particularly unlikely in patients with ocular
histoplasmosis, angioid streaks, pathologic myopia, or choroidal
rupture, all of which are conditions in which there is a high incidence
of CNV. Features that these conditions share with each other and with
AMD, are disruption of Bruch’s membrane and/or other abnormalities of
the extracellular matrix of the RPE and Bruch’s membrane. We have
recently demonstrated that exposure of cultured RPE cells to certain
extracellular matrix proteins that are not normally part of the
microenvironment of RPE cells, particularly thrombospondin 1, can
increase secretion of VEGF.
43 Therefore, alterations in
Bruch’s membrane accompanied by changes in the extracellular matrix of
RPE cells may contribute to increased production of VEGF in patients
with CNV.
Transgenic mice that express high levels of VEGF in photoreceptors,
show neovascularization originating from retinal vessels, but not
choroidal vessels.
17 Perhaps normal RPE cells prevent
retina-derived VEGF from getting to choroidal vessels. Another
possibility is that normal RPE cells and/or Bruch’s membrane provide a
physical or biochemical barrier to vascular invasion from the choroid.
Mice or rats with photoreceptor degeneration frequently have
neovascularization that extends from retinal or choroidal vessels into
the RPE,
44 and we have demonstrated that photoreceptor
degeneration results in increased expression of VEGF in RPE
cells.
45 Taken together, these various pieces of evidence
suggest that exposure of RPE cells to certain extracellular matrix
components results in increased production of VEGF and that unlike
retina-derived VEGF, RPE-derived VEGF gains access to choroidal vessels
and can stimulate CNV.
The results of the present study suggest that VEGF plays an important
role in the stimulation of CNV, but they do not rule out the
participation of other stimulatory factors. However, blockade of VEGF
signaling dramatically, and in many mice completely, inhibits CNV.
Therefore, even if there are other contributing stimulatory factors, it
may be possible to effectively prevent CNV in patients by concentrating
solely on inhibition of VEGF signaling. This also appears to be an
effective strategy for retinal neovascularization, because blockage of
VEGF signaling completely inhibits retinal neovascularization in murine
oxygen-induced ischemic retinopathy.
30 46 These data
suggest that if safety is established, these drugs should be tested for
their ability to inhibit ocular neovascularization in patients.
The authors thank Michele Melia of the Wilmer Institute Core
statistics center for performing the statistical analysis.