Abstract
purpose. Aquaporin-1 (AQP1) water channels are expressed widely in organ and tumor microvascular endothelia. Rapid microvessel proliferation occurs in growing tumors, diabetic and other retinopathies, and prenatal development. The purpose of this study was to investigate the role of AQP1 in retinal vessel proliferation.
methods. Comparative studies were performed on wild-type compared with AQP1 null mice using an established mouse model of oxygen-induced retinopathy. Neonatal mice were maintained in a 75% oxygen atmosphere for 5 days to suppress angiogenesis and then were returned to room air to induce vessel proliferation. AQP1 expression was also studied in extraocular microvessels and in primary endothelial cell cultures from pig retina.
results. Surprisingly, AQP1 immunoreactivity was detected in only a small percentage of newly formed retinal microvessels, whereas AQP1 was strongly expressed in all choroidal and hyaloid vessels and in various extraocular microvessels in neonatal and prenatal mice. Oxygen-induced retinal microvessel proliferation was not significantly impaired in neonatal mice lacking AQP1, as quantified in flat-mounted retinas and thin sections. However, AQP1 was expressed in endothelial cells cultured from retinal microvessels.
conclusions. Microvessel proliferation in oxygen-induced retinopathy is AQP1-independent. Retinal endothelia have the capacity to express AQP1, though intact retinal vessels chronically suppress AQP1 expression.
Aquaporin-1 (AQP1) is a water-transporting protein expressed in cell plasma membranes in many epithelial, endothelial, and other tissues.
1 2 AQP1 is expressed in microvascular endothelia throughout mammalian tissues except in brain and retina,
1 2 3 4 and AQP1 is seen in all tumor microvessels that have been studied.
5 6 In active multiple myeloma, the expression of AQP1 in microvessels appears to correlate with their proliferation.
7 Cell culture studies suggest that brain microvascular endothelial cells have the capacity to express AQP1 but that AQP1 expression in intact brain is suppressed by unknown factor(s).
8 These data suggested the involvement of AQP1 in microvessel proliferation.
In support of a role for AQP1 in microvessel proliferation, we found marked impairment in tumor angiogenesis in mice lacking AQP1 and impairment in growth factor–induced angiogenesis in implanted Matrigel pellets.
9 The density of microvessels was reduced in tumors in AQP1 null mice, producing islands of viable tumor cells surrounded by necrotic tissue. One consequence of altered tumor architecture in AQP1 null mice is reduced diffusion of macromolecules.
10 Analysis of the cellular mechanism of AQP1-dependent angiogenesis using cultured endothelial cells indicated a role for AQP1 in cell migration,
9 which we proposed involves increased water permeability at the leading edge of migrating cells, where lamellipodial extensions form. Involvement of AQPs in cell migration appears to be a general phenomenon seen in different cell types, including brain astroglial cells,
11 12 kidney proximal tubule epithelial cells,
13 neutrophils,
14 corneal epithelial cells,
15 cryptosporidium-infected cholangiocytes,
16 and AQP-expressing tumor cells.
17
In this study, we investigated the involvement of AQP1 in retinal vessel proliferation using an established mouse model of oxygen-induced retinopathy. The motivation for this study is the central role of retinal vessel proliferation in the pathogenesis of diabetic retinopathy, age-related macular degeneration, and retinal vascular occlusion.
18 19 Involvement of AQP1 in retinal vessel proliferation would suggest the possible use of AQP1 inhibitors in these conditions. However, contrary to expectations, we found that rapidly proliferating retinal microvessels did not express AQP1 by standard immunodetection procedures, which was further investigated by comparative functional studies in AQP1 null mice and by analysis of AQP1 expression in prenatal microvessels and endothelial cell cultures from retinal microvessels.
Transgenic mice deficient in AQP1 were generated by targeted gene disruption, as described.
20 Litter-matched wild-type and AQP1-deficient mice in a CD1 genetic background were generally used. In some studies, heterozygous mice were used for litter matching because wild-type and AQP1 heterozygous mice have been reported to have a similar phenotype regarding water transport, with similar water permeability and unimpaired urinary-concentrating ability in AQP1 heterozygous mice.
20 21 Investigators were masked to mouse genotype information until the completion of experiments. All experimental methods and animal care procedures adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the University of California, San Francisco Institutional Animal Care and Use Committee (IACUC).
Oxygen-induced proliferative retinopathy was produced as described.
22 Briefly, litter-matched mice on postnatal day (P) 7, along with each dam, were exposed to 75% ± 3% oxygen for 5 days in a custom-made chamber, producing vaso-obliteration in the capillary beds of the central retina. Temperature was maintained at 21°C ± 1°C, and oxygen was measured at least four times a day with an oxygen analyzer (Handi; Maxtec Inc., Salt Lake City, UT). After 5 days, pups were returned to room air for different times (P12-P20), resulting in preretinal neovascularization of the central ischemic retina.
To stain retinal vessels, pups were anesthetized and perfused through the left ventricle with 1 mL PBS containing 50 mg/mL of 2 × 10
6 MWt fluorescein isothiocyanate (FITC)-dextran (Sigma, St. Louis, MO), as described.
22 Left eyes were enucleated and fixed in 4% paraformaldehyde for 1 hour to 4 hours at 4°C Retinas were dissected, and flat-mounts were prepared after four to six radial cuts were made. A coverslip containing a drop of mounting medium (Vectashield; Vector Laboratories, Burlingame, CA) was applied. Some retinas were incubated with
Griffonia simplicifolia isolectin B4 (50 ng/mL; Sigma) overnight at 4°C, followed by a 1-hour incubation at room temperature with streptavidin (1:200; Alexa-488; Molecular Probes, Eugene, OR). Retinas were imaged using an epifluorescence microscope (Leica, Heidelberg, Germany) equipped with a digital camera (Spot, Sterling Heights, MI). To visualize the whole retina, five to six fields per retina at 50× magnification were imaged. Right eyes were enucleated and fixed in 4% paraformaldehyde overnight at 4°C. After fixation, tissue was processed through graded concentrations of ethanol, followed by solvent and clearing agent (Citrisolv; Fisher Scientific, Los Angeles, CA), and was embedded in paraffin. Sixteen serial microtome sections (5-μm thickness; 40 μm between sections) were obtained parallel to the optic nerve in a sagittal plane through the cornea, with eight sections on each side of the optic nerve. Sections were deparaffinized in solvent and clearing agent (Citrisolv; Fisher Scientific), rehydrated in graded ethanols, and stained with periodic acid-Schiff (PAS) reagent and hematoxylin. Retinal sections were imaged using bright-field microscopy.
Fluorescein Staining of Whole-Mounts.
Whole-mounted retinas stained with FITC-dextran or with isolectin B4 (both from Sigma) were analyzed using image analysis software (Spot) according to established methods.
23 Qualitative assessment of the retinal vasculature was performed on hyperoxia-exposed eyes at P12, P15, P17, and P20. Quantitative analysis of the neovascular response was performed in hyperoxia-exposed retinas from P17 mice, at which time a strong neovascular response was seen. Briefly, vessel-obliterated and preretinal neovascularization areas were outlined and quantified in each quadrant of the retina as the percentage of total area of retina analyzed. Statistical analysis was performed using the two tailed-Student’s
t-test.
P = 0.05 was considered significant.
Quantification of Neovascular Nuclei.
Retinal vessel endothelial cell nuclei anterior to the inner limiting membrane were counted in PAS-stained retinal sections of hyperoxia-exposed eyes at P17. The average number of neovascular nuclei per section was calculated as the mean of 16 sections adjacent to the optic nerve (eight on each side of the optic nerve) counted per eye (four eyes per genotype). Statistical analysis was performed as described.
Eye sections from hyperoxia-exposed neonatal mice at different time points were deparaffinized in solvent and clearing agent (Citrisolv; Fisher Scientific) and were rehydrated in graded ethanols. After epitope retrieval with citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6, 30 minutes, 95°C–100°C), sections were blocked with goat serum and incubated with rabbit anti-AQP1 (1:500; Chemicon, Temecula, CA) and isolectin B4 (Sigma). Sections of extraocular tissues (skeletal muscle and kidney) from embryonic day (E) 16.5 mice were processed identically and incubated with rabbit anti-AQP1 and rat anti-mouse CD31/PECAM (PharMingen, San Diego, CA). Primary antibodies were detected with Texas Red goat anti-rabbit and Alexa-488 goat anti-rat secondary antibodies, and isolectin B4 was detected with Alexa-488 streptavidin (all at 1:200; Molecular Probes). For immunostaining of retinal endothelial cell cultures, cells were fixed for 15 minutes with paraformaldehyde at room temperature, blocked with bovine serum, and incubated with rabbit anti-AQP1, rat anti-mouse CD31/PECAM, or rabbit anti-human von Willebrand factor (VWF; 1:50; Dako, Carpinteria, CA) and then incubated with secondary antibodies as described. Colocalization studies of AQP1 and VWF were performed using labeling reagents (Zenon Technology; Molecular Probes). Antibody specificity was confirmed by excluding primary or secondary antibodies.
Retinal endothelial cells were scraped from coverslips and homogenized in 250 mM sucrose, 1 mM EDTA, 2 μg/mL aprotinin, 2 μg/mL pepstatin A, 2 μg/mL leupeptin, and 100 μg/mL serine protease inhibitor (Sigma). Homogenates were centrifuged at 1500g for 10 minutes, and the supernatant was loaded onto a 4% to 12% sodium dodecyl sulfate-polyacrylamide gel (20 μg protein/lane). Proteins were transferred to a polyvinylidene difluoride membrane and incubated with rabbit anti-AQP1 antibody (1:1000) followed by horseradish peroxidase-linked anti-rabbit IgG (1:10.000; GE Healthcare, Slough, Berkshire, UK), and visualized by enhanced chemiluminescence (Roche Diagnostics, Indianapolis, IN).
Primary cultures of retinal endothelial cells were generated from fresh pig eyes, as described previously for bovine retinal endothelial cells, with modifications.
24 25 Under a dissecting microscope (SMZ1500; Nikon, Tokyo, Japan), the largest retinal vessels from four retinas were microdissected with the use of two microforceps and were suspended in PBS containing collagenase (500 μg/mL) and pronase (100 μg/mL; both from Sigma). Isolated fragments of vessels were incubated with shaking (140 rpm) at 37°C for 30 minutes. Digested vessels were then sieved through 100-μm nylon mesh, and the remaining retinal capillary fragments were cultured on fibronectin-coated coverslips (1 μg/cm
2; Sigma) in media (MCDB-131; Sigma) supplemented with 10% FBS (Hyclone, Logan, UT), 10 ng/mL epidermal growth factor (Sigma), 0.2 mg/mL endothelial cell growth medium additive (EndoGro; Vec Technologies, Rensselaer, NY), 0.09 mg/mL heparin (Fisher Scientific), and 0.01 mg/mL antibiotic-antimycotic (Invitrogen-Life Technologies, Rockville, MD). Cell purity was assessed by immunoreactivity against endothelial markers CD31 and VWF.
AQP1 expression was studied in retinal microvasculature in a neonatal mouse model of oxygen-induced retinopathy in which mice were maintained in a 75% oxygen atmosphere for 5 days (P7–P12) to suppress angiogenesis and then were returned to room air to induce vessel proliferation. AQP1 was detected by immunostaining, with vascular endothelia identified by isolectin B4.
Figure 1shows representative fluorescence micrographs of eyes from control mice at 0 to 8 days after return to room air (P12–P20). Unexpectedly, AQP1 expression was not detected in most retinal vessels. Retinal vessel proliferation was not seen in the first 3 days after mice were returned to normoxia (P12–P14), with vessels restricted to the inner layers of the retina, including the ganglion cell layer
(Figs 1A 1B 1C) . A strong neovascular response was seen at P15, with new vessels reaching the outer plexiform layer. Few neovascular tufts were seen beyond the inner limiting membrane
(Fig. 1D) . The strongest neovascular proliferative response occurred from P16 to P18,with new vessels seen throughout most retinal layers (up to the outer plexiform layer) and numerous neovascular tufts seen beyond the inner limiting membrane invading the vitreous body
(Figs. 1E 1F 1G) . Neovascular proliferation regressed by day P20. Apart from the presence of sparse microvessels in some layers above the ganglion cell layer, few neovascular tufts were seen on the vitreous side of the retina
(Fig. 1H) .
We found that most retinal microvessels did not express AQP1, even when neovascular proliferation was active, as at P17
(Figs. 1E 1F 1G) . In contrast, AQP1 was strongly expressed in the nonpigmented layers of the ciliary epithelium and in the apical and basolateral membranes of lens epithelial cells
(Fig. 1A) . We also found the expression of AQP1 in hyaloid vessels, which form transiently to nourish the retina and lens during early development
(Fig. 1D) . Weak expression of AQP1 was also detected in the outer retinal layers
(Figs. 1F 1H) .
Figure 2Ashows the absence of AQP1 immunoreactivity in the retina of a P17 oxygen-exposed mouse.
Figures 2B 2C 2Dshow a small number of vessels in some of the neovascular tufts that expressed AQP1, representing approximately 10% of the vessels in the tufts, seen in less than 1% of neovascular tufts analyzed. As expected, AQP1 immunoreactivity was not seen in any part of the eye from AQP1 null mice
(Fig. 1I) .
AQP1 expression was studied in mice not exposed to hyperoxia, at day E16.5 during embryonic retinal development, and in adult mice. AQP1 was not detected in retinal microvessels in non-hyperoxia exposed mice, but could be seen in erythrocytes within some vessels
(Figs. 3A 3C) . AQP1 was strongly expressed in nonpigmented layers of the ciliary epithelium, cornea, and at E16.5, in hyaloid vessels. As expected, AQP1 immunoreactivity was absent in the eyes of AQP1 null mice
(Figs. 3B 3D) .
To determine whether AQP1 was expressed in extraocular microvascular endothelia and whether its deletion in mice was associated with altered microvascular architecture, we analyzed microvessels in skeletal muscle and kidney medulla of nonoxygen-exposed wild-type and AQP1-null mouse embryos at E16.5. AQP1 was expressed in the microvascular endothelia of skeletal muscle and kidney medulla, with similar microvessel morphology in wild-type and AQP1-null mice at the light microscopic level
(Fig. 4) .
The absence of AQP1 immunofluorescence in retinal microvessels suggested that AQP1 is not involved in vessel proliferation in oxygen-induced retinopathy. However, the absence of AQP1 immunoreactivity cannot rule out possible contributions from small amounts or immunosilent AQP1 not detected by the antibody used here. Indeed, there is evidence that AQP4 immunoreactivity in vivo depends on its conformational state.
26 Therefore, we carried out comparative functional studies in wild-type and AQP1 null mice.
Quantitative analysis of areas of vaso-obliteration and neovascular response was performed in retinal mounts in which vessels were fluorescently stained by perfusion with FITC-dextran or isolectin B4; the latter provided better visualization of neovascular tuft morphology. Relative areas of vaso-obliteration and neovascular response were compared in wild-type and AQP1-null mice at days P12, P15, P17, and P20
(Fig. 5) . After 5 days of hyperoxia (P12), the central retina showed maximum vessel regression, with no significant difference in the areas of vaso-obliteration in wild-type and AQP1 null mice
(Figs. 5A 5B 5C 5D) . From P12 to P20, the neovascular response progressed from the perfused retinal edge to the optic nerve and the ischemic central retina, producing a progressive reduction in the obliterated vascular area, which was qualitatively similar in wild-type and AQP1 mice
(Figs. 5A 5B) . Neovascularization was complete at approximately P20 in both cases, with retina morphology at that time resembling that in control mice not subjected to hyperoxia
(Fig. 5D) . Quantitative analysis of retinal vessel proliferation was performed at P17
(Fig. 5C) . After 5 days on room air (P17), there was no significant difference in the relative areas of neovascularization between wild-type and AQP1 null mice (each approximately 7%;
Fig. 5E ). The initial percentage area of vaso-obliteration of the central retina in wild-type and AQP1 null mice (approximately 28%) also did not differ significantly.
As an independent assessment of neovascularization, we counted the number of nuclei in paraffin sections of day P17 retina that extended beyond the inner limiting membrane into the vitreous. The appearance and number of nuclei extending beyond the inner limiting membrane were similar in wild-type and AQP1 null mice
(Fig. 6) , supporting the conclusion that
AQP1 gene deletion in mice does not affect the retinal neovascular response after hyperoxia.
We tested the hypothesis that retinal microvascular endothelial cells have the capacity to express AQP1 in cell culture when removed from their native environment. For these studies, we cultured microvascular endothelia from pig retina, as described in Materials and Methods, yielding highly purified retinal endothelial cells. AQP1 immunoreactivity was not seen in retinal microvessels in sections of pig eye but was seen in other structures, such as the ciliary epithelium (data not shown).
Figure 7Ashows representative fluorescence micrographs of the primary cultured cells. Most cells were positive for AQP1, which colocalized with the endothelial cell markers VWF or CD31. The few cells negative for VWF or CD31 (arrows) were also negative for AQP1.
Figure 7Bshows immunoblot analysis with AQP1 antibody. A band at approximately 28 kDa was seen, corresponding to nonglycosylated AQP1. A band at the same molecular size was seen for the positive control (red blood cell membranes), as was a band at approximately 35 kDa corresponding to glycosylated AQP1.
We report AQP1-independent retinal microvessel proliferation in a well-established model of oxygen-induced retinopathy. Expression of AQP1 in the retinal vasculature was largely absent during the phase of active neovascular proliferation, which was supported by functional measurements showing no significant differences in the vaso-obliterative or neovascular responses in wild-type and AQP1 null mice. Relative areas of vaso-obliteration and the kinetics of neovascularization reported here are comparable to those reported previously in oxygen-induced retinopathy in mice.
23 27 28
The absence of significant AQP1 expression in rapidly proliferating retinal microvessels was an unexpected result because AQP1 is expressed strongly in rapidly proliferating microvessels in tumors and other tissues and because AQP1 deletion in mice impairs angiogenesis in tumor and nontumor models.
9 We proposed that defective angiogenesis in AQP1 deficiency is caused by impaired endothelial cell migration as a consequence of reduced water permeability in lamellipodia at the leading edge of migrating endothelia. A potential concern in the in vivo experiments in that study was possible effects of physiological differences between wild-type and AQP1 null mice, such as mildly reduced blood pressure.
29 The comparable retinal angiogenesis between wild-type and AQP1 null mice found here provides a good control that excludes secondary physiological effects as impairing angiogenesis in in vivo tumor models.
It is not known why AQP1 is absent in retinal vasculature. In the central nervous system, including brain and spinal cord, AQP1 is absent in vascular endothelia but has been found in all other vascular endothelia in which it has been sought.
1 4 One study even reported low levels of AQP4 in brain endothelia, as revealed by immunogold staining.
30 AQP1 is also expressed in many nonvascular “endothelia” such as corneal endothelia, lymphatics, and endocardium.
1 31 32 Although no aquaporin has been identified in brain endothelia, AQP4 is expressed in astrocyte foot processes that are closely apposed to blood vessels.
33 34 The absence of AQP1 in microvascular endothelia in the retina, which comprises the blood-retinal barrier, appears to parallel its absence at the blood-brain barrier. In brain, AQP4 (but not AQP1) expression has been reported in chick embryos at the time of development of the blood-brain barrier.
35 An analogous situation is found in the eye, in which AQP1 is expressed in choroidal vessels (principal blood supply for photoreceptors) and hyaloid vessels (transient intraocular circulatory system that supplies retina and lens during early eye development) but not in the retinal vessels, where the blood-retinal barrier is formed. The expression of AQP1 in a small percentage of proliferating vessels in some retinal neovascular tufts may be related to sites of disruption of the blood-retinal barrier that occur during rapid neovascular proliferation in the oxygen-induced retinopathy model.
36 It has been proposed that proliferating vessels in human retinopathy of prematurity have a disrupted barrier system and are more permeable than normal retinal vessels.
37
We found that retinal endothelial cells have the capacity to express AQP1 after removal from their native environment and growth in culture. For these studies, we cultured retinal microvascular endothelia from pig. We used pig rather than mouse for these studies because of its substantially larger size and because of many similarities with human retina.
38 Retinal endothelial cells were isolated according to a modification of methods described previously for bovine retina.
24 25 The main procedural difference was that digestion occurred with isolated vessels (rather than the whole retina), which greatly enriched the endothelial cell population, yielding cultures with greater than 90% endothelial cells. AQP1 expression in primary cultured endothelial cells, but not in retina in vivo, may be related to cell dedifferentiation in culture. Downregulation of AQP1 expression has been reported when brain endothelial cells are cocultured with astrocytes,
8 where it was proposed that AQP1 expression is suppressed in the specialized endothelium of the blood-brain barrier by factors released from astroglial cells. The apparent lack of AQP1 expression in brain/retinal endothelial cells represents an important phenotypic difference between these and most other endothelia, which might be related to the specialized blood-brain/retinal barrier function.
Aberrant growth of blood vessels in the eye is a major cause of visual impairment. It occurs as a complication of several diseases, including diabetic retinopathy, age-related macular degeneration, and retinal vascular occlusions.
39 During normal retinal vascular development, primary vascular remodeling involves capillary constriction, retraction, and atrophy, resulting in the elimination of redundant channels.
37 In diabetic retinopathy, retinal neovascularization follows from microangiopathic retinal damage induced by hyperglycemia and in response to a variety of factors, including growth hormone, insulinlike growth factor, basic fibroblast growth factor, and vascular endothelial growth factor (VEGF). VEGF is thought to be the principal effector of retinal neovascularization in proliferative retinopathies. It is thought that ischemia of the inner retinal layers secondary to capillary bed closure induces vascular growth factors, which induce new blood vessel growth into the vitreoretinal interface.
40
Current therapies to treat retinal oculopathies, such as age-related macular degeneration and diabetic retinopathy, target VEGF or its receptors. Although promising, a concern of VEGF-related strategies is that their effects are not restricted to active angiogenesis but are also required for the maintenance and differentiation of mature blood vessels.
19 A possible implication of AQP1 involvement in endothelial cell migration would have been the use of specific inhibitors of AQP1 to slow the proliferation of microvessels in vascular-related retinopathies, with predicted minor effect on normal, nonproliferating vessels. Unfortunately, we found that retinal vessel proliferation was AQP1 independent, though it remains to be determined whether AQP1 is present in diabetic retinopathy in humans.
In summary, we found that most rapidly proliferating retinal microvessels in a mouse model of oxygen-induced retinopathy do not express AQP1, yet retinal microvascular endothelial cells have the capacity to express AQP1 when cultured. Consequently, we found that retinal vessel proliferation was not affected by AQP1 deletion. This study supports the use of AQP1 inhibitors (when available) to treat ocular disorders such as glaucoma because these inhibitors should not affect the retinal vasculature.
The authors thank Liman Qian and Louise Han for mouse breeding and genotype analysis.