Human recombinant VEGF-A₁₆₅ was purchased from R&D Systems (Minneapolis, MN). Human recombinant MMP-2 and MMP-9 (purified active forms), PD98059, SP600125, SU1498, and LY294002 were obtained from Calbiochem (Bad Soden, Germany). AG1478 was from Alexis (Grünberg, Germany); U0126 and SB203580 were from Tocris (Ellisville, MO). All other substances were from Sigma-Aldrich (Taufkirchen, Germany), unless stated otherwise. The test substances had no effect on cell viability, as determined by trypan blue exclusion (not shown). Apparently, culturing the cells in the presence of MMP-2 or MMP-9 (50 ng/mL each) for up to 24 hours did not alter the cellular morphology (not shown). 

The use of human material was approved by the Ethics Committee of the University of Leipzig, and the study was performed according to the Declaration of Helsinki. Human RPE cells were obtained from several donors within 48 hours of death and were prepared as described. ³² Cells were suspended in complete Ham F-10 medium containing 10% fetal bovine serum, diagnostic medium (Glutamax II; Invitrogen), and gentamicin and were cultured in tissue culture flasks (Greiner, Nürtingen, Germany) in 95% air/5% CO₂ at 37°C. Cells of passages 3 to 5 were used. The epithelial nature of the RPE cells was routinely identified by immunocytochemistry using the monoclonal antibodies AE1 (recognizing most of the acidic type I keratins) and AE3 (recognizing most of the basic type II keratins), both from Chemicon (Hofheim, Germany). All tissue culture components and solutions were purchased from Gibco BRL (Paisley, UK). 

Total RNA was extracted (RNeasy Mini Kit; Qiagen, Hilden, Germany) and was treated with DNase I (Roche, Mannheim, Germany). cDNA was synthesized from 1 μg total RNA using a cDNA synthesis kit (RevertAid H Minus First Strand cDNA Synthesis Kit; Fermentas, St. Leon-Roth, Germany). Semiquantitative real-time RT-PCR was performed (Single-Color Real-Time PCR Detection System; BioRad, Munich, Germany) with the primer pairs described in Table 1 . The PCR solution contained 1 μL cDNA, specific primer set (0.5 μM each), and 10 μL primer (QuantiTect SYBR Green PCR Kit; Qiagen) in a final volume of 20 μL. PCR parameters were initial denaturation and enzyme activation (one cycle at 95°C for 15 minutes); denaturation, amplification, and quantification for 45 cycles at 95°C for 30 seconds, 60°C for 30 seconds, and 72°C for 1 minute; melting curve, 55°C, with the temperature gradually increased (0.5°C) to 95°C. mRNA expression was normalized to the levels of GAPDH mRNA, and the changes were calculated as described. ³³ For the calculation of the relative expression changes, we used the fluorescence signal of lower cycles during the log phase of the product formation. Real-time PCR efficiency (E) was calculated according to the equation: E = 10^[−1/slope]. Efficiencies of the target genes were found to be similar in the investigated range between 0.5 and 100 ng cDNA (VEGF-A, 2.31; PEDF, 2.21; MMP-2, 2.20; MMP-9, 2.06; GAPDH, 2.30). Amplified PCR products were analyzed routinely by standard agarose gel electrophoresis. 

Cells were cultured at 3 × 10³ cells per well in 96-well plates (100 μL culture medium per well). At a confluence of approximately 80%, the cells were cultured in serum-free medium for 16 hours. Subsequently, the culture medium was changed, and the cells were stimulated by various factors or MMPs in the absence and presence of triamcinolone. Supernatants were collected after 6 hours, and the levels of total MMP-9 (92-kDa proactive and 83-kDa active forms) or VEGF-A₁₆₅ in the cultured media (100 and 200 μL, respectively) were determined by ELISA (R&D Systems). 

Cells were seeded at 3 × 10³ cells per well in 96-well microtiter plates (Greiner) and were allowed to attach for 48 hours. Thereafter, the cells were growth arrested in medium without serum for 5 hours; subsequently, medium containing 0.5% serum with and without test substances was added for another 24 hours. The proliferation rate was determined by measurement of the DNA synthesis rate. The incorporation of bromodeoxyuridine (BrdU) into the genomic DNA was determined with the use of the Cell Proliferation ELISA BrdU Kit (Roche). BrdU (10 μM) was added to the culture medium 5 hours before fixation. 

Chemotaxis was determined with a modified Boyden chamber assay. Suspensions of RPE cells (100 μL; 5 × 10⁵ cells/mL serum-free medium) were seeded onto polyethylene terephthalate filters (diameter, 6.4 mm; pore size, 8 μm; Becton Dickinson, Heidelberg, Germany) coated with fibronectin (50 μg/mL) and gelatin (0.5 mg/mL). Within 16 hours of seeding, the cells attached to the filter and formed a semiconfluent monolayer. Cells were pretreated with blocking substances for 30 minutes, and thereafter the medium was changed to medium without additives in the upper well and medium containing MMP-2 or MMP-9 and the appropriate blockers in the lower well. After incubation for 6 hours, the inserts were washed with buffered saline, fixed with Karnofsky reagent, and stained with hematoxylin. Nonmigrated cells were removed from the filters by gentle scrubbing with a cotton swab. Migrated cells were counted, and the results were expressed relative to the cell migration rate in the absence of test substances. 

Rates of BrdU incorporation and migration are expressed as percentages of untreated control (100%). ELISA data were normalized to the total cellular protein content and were presented as picogram of secreted protein per microgram of cellular protein. For each test, at least three independent experiments were performed in triplicate. Data are expressed as mean ± SEM. Statistical significance (Kruskal-Wallis test followed by Dunn comparison for multiple groups) was accepted at P < 0.05. 

Human RPE cells expressed mRNAs for MMP-2 and MMP-9 (Fig. 1A) . With real-time PCR, we determined the gene expression levels for both MMPs. Under unstimulated control conditions, cultured RPE cells expressed a significantly (P < 0.001) higher amount of mRNA for MMP-2 than for MMP-9, as indicated by the lower cycle threshold number necessary to detect the MMP-2 transcript compared with the MMP-9 transcript (Fig. 1B) . To determine whether the gene expression of both MMPs is regulated by pathogenic factors, we tested chemical hypoxia induced by CoCl₂ (150 μM), oxidative stress evoked by H₂O₂ (20 μM), high-glucose conditions (with 25 mM glucose in the culture medium), and stimulation of the cultures with VEGF. Chemical hypoxia caused a transient upregulation of the gene expression of MMP-2 and MMP-9 that was apparent after two hours of hypoxia (Figs. 1C 1D)but not after 6 and 24 hours (not shown). Oxidative stress slightly decreased the gene expression of both MMPs (not shown), whereas VEGF induced a transient upregulation of the gene expression of MMP-9 after 2 hours of stimulation (Fig. 1D)and had no effect on the level of MMP-2 mRNA (Fig. 1C) . High-glucose conditions did not alter the gene expression of MMP-2 and MMP-9 (not shown). The data indicated that the gene expression of MMP-2 and MMP-9 in RPE cells was upregulated under hypoxic conditions and that VEGF increased the gene expression of MMP-9. 

We found that chemical hypoxia and VEGF increase the gene expression of MMP-9 in RPE cells. To determine whether this increase is associated with the secretion of MMP-9, the content of total MMP-9 protein in the cultured media was measured by ELISA. Similar to the effects on gene expression, chemical hypoxia caused a significant increase in the secretion of MMP-9, whereas conditions involving high glucose or oxidative stress did not alter secretion (Fig. 2A) . VEGF induced a dose-dependent increase in the secretion of MMP-9 (Fig. 2B) . The data indicate that the secretion of MMP-9 by RPE cells is stimulated by VEGF and under hypoxic conditions. Because hypoxia is the primary condition that induces upregulation of VEGF in retinal cells, ⁴ we tested whether the hypoxic stimulation of MMP-9 secretion is mediated by autocrine VEGF signaling. As shown in Figure 2C , the selective inhibitor of the kinase insert domain-containing (KDR)/flk-1 receptor, SU1498, prevented the increase in MMP-9 secretion induced by chemical hypoxia, suggesting a crucial role of autocrine VEGF signaling in the hypoxic stimulation of MMP-9 secretion. 

With real-time PCR and ELISA, we investigated whether the stimulation of RPE cells by MMP-2 and MMP-9 induces gene expression and secretion of VEGF. As shown in Figure 3 , exogenous MMP-9, but not MMP-2, caused a transient increase in the gene expression of VEGF-A. MMP-9 did not alter the gene expression of various other members of the VEGF family or for their receptors, including VEGF-B, VEGF -C, VEGF -D, flt-1, and KDR (Fig. 3) . Furthermore, MMP-2 and MMP-9 had no effects on the gene expression of the pigment epithelium-derived factor (PEDF; not shown). To determine whether MMPs alter the secretion of VEGF, RPE cells were cultured in the presence of MMP-2 or MMP-9 for 6 hours, and the content of VEGF-A₁₆₅ in the cultured media was analyzed by ELISA. Cells constitutively released VEGF into the culture media. Exogenous MMP-2 caused a slightly significant (P < 0.05) decrease in the VEGF content of the cultured media (Fig. 4A) . In contrast, stimulation of the cells with MMP-9 caused a significant (P < 0.01) increase in the secretion of VEGF. The stimulatory effect of MMP-9 showed dose dependence at concentrations above 1 ng/mL (Fig. 4B) . The data indicated that MMP-9 stimulated the gene and protein expression of VEGF-A in RPE cells, whereas MMP-2 reduced the secretion of VEGF. Simultaneous administration of both MMPs revealed that the reducing effect of MMP-2 on the secretion of VEGF was overwhelmed by MMP-9 (Fig. 4C) . 

To determine further physiological roles of MMPs in RPE cells, the effects of MMP-2 and MMP-9 on cellular proliferation and chemotaxis were examined. Exogenous MMP-2 or MMP-9 in quantities up to 100 ng/mL did not alter the rate of RPE cell proliferation (not shown). On the other hand, both MMPs dose-dependently stimulated the chemotaxis of RPE cells. MMP-9 had a significantly stronger effect than MMP-2 at higher doses (P < 0.01 at 10 ng/mL) (Fig. 5A) . We found that the motogenic effect of MMP-9 on RPE cells was mediated by activation of the p38 mitogen-activated protein kinase, the phosphatidylinositol-3 kinase (PI3K)-Akt, and c-Jun N-terminal kinase (JNK) signal transduction pathways but was apparently independent on activation of the extracellular signal-regulated kinases 1 and 2 (ERK1/2; Fig. 5B ). AG1478, an inhibitor of the epidermal growth factor (EGF) receptor tyrosine kinase, was unable to alter the rate of cell migration, suggesting that transactivation of EGF receptors did not occur. The inhibitor of ERK1/2 activation, PD98059, stimulated cell migration under control conditions (Fig. 5B) . A similar effect was observed with another inhibitor of ERK1/2 activation, UO126 (4 μM; not shown). These data suggest the presence of an inhibitory cross-talk between p38 and ERK1/2 in RPE cells. It has been shown in various cell systems that pharmacologic inhibition of ERK1/2 activation causes an increase in cell migration. ^{34 35}  

It is known that the angiostatic steroid triamcinolone acetonide reduces the secretion of VEGF by RPE cells. ³¹ We found that triamcinolone significantly decreased the secretion of VEGF by RPE cells at control conditions and after stimulation with MMP-2 and MMP-9 (Fig. 4A) . The effect of triamcinolone was dose dependent, with a half-maximal effect at a concentration below 10 nM (Fig. 4D) . Triamcinolone also prevented the MMP-9–evoked increase in the gene expression of VEGF-A (Fig. 3A) . Furthermore, triamcinolone inhibited the hypoxic increase in the gene expression of MMP-2 and MMP-9 (Figs. 1C 1D)and the VEGF-evoked increase in the gene expression of MMP-9 (Fig. 1D) . The secretion of MMP-9 evoked by chemical hypoxia (Fig. 2C)or VEGF (Fig. 2D)was prevented in the presence of triamcinolone. The VEGF-evoked secretion of MMP-9 was fully inhibited at a triamcinolone concentration of 1 μM (Fig. 2D) . It is concluded that triamcinolone inhibits the positive feedback regulation between MMP-9 and VEGF by blockade of the gene expression and secretion of both factors. In contrast, the motogenic effect of MMP-9 was not inhibited by triamcinolone (Fig. 5B) . 

It has been shown that RPE cells in situ secrete VEGF from its basolateral surface to keep the choroidal vessels fenestrated and PEDF from its apical surface to inhibit vascularization of the photoreceptor layer. ^{36 37 38} Both factors are also survival factors for the respective target cells. Upregulation of VEGF expression by RPE cells under hypoxic conditions is crucially implicated in the pathologic neovascularization that occurs in blinding diseases such as age-related macular degeneration. ⁷ However, the factors that cause overproduction of VEGF by RPE cells under pathologic conditions are not well characterized. Here we describe the presence of a positive feedback regulation between MMP-9 and VEGF in RPE cells that may contribute to the hypoxic expression of both factors. MMP-9 increases the expression of VEGF in RPE cells by stimulation of the gene transcription and secretion. Conversely, VEGF increases the gene expression and secretion of MMP-9. In contrast to MMP-9, MMP-2 reduces the secretion of VEGF-A (Fig. 4A) . 

We found that MMP-2 does not regulate the expression of VEGF at the transcriptional level (Fig. 3) , suggesting that the reducing effect of MMP-2 on the secretion of VEGF is likely mediated by posttranscriptional mechanisms. The presence of posttranscriptional or posttranslational regulation is also suggested by the observation that the increases in mRNA expression of MMP-9 and VEGF were generally larger than the increases in protein secretion. It has been shown in different cell systems that the expression of MMP-9 is regulated at various levels, including transcription, mRNA stability, translational efficiency, vesicle transport, and secretion. ^{39 40 41} A decrease in MMP-9 mRNA stability, for example, has been described to be induced by nitrosative stress. ⁴² In fibroblasts, hypoxia-induced MMP-9 expression shows an imbalance between mRNA and protein level, which may be subjected to an inhibited translation of mRNA. ⁴³ The cause of the apparent difference in the gene and protein expression of MMP-9 in RPE cells remains to be determined in future experiments. 

MMPs have been implicated in neovascularization because of their support of the migration and proliferation of vascular endothelial cells through the proteolytic degradation of basement membranes and extracellular matrix components. The present finding that MMP-9 causes the upregulation of VEGF expression by RPE cells suggests that this proteinase also may facilitate pathologic neovascularization through stimulation of the production of angiogenic factors. Although hypoxia increases the gene expression of both MMPs (Figs. 1C 1D) , the feedback regulation between MMP-9 and VEGF is assumed to be implicated in the angiogenic switch that occurred in the retinal tissue under hypoxic conditions. This assumption was based on two observations: The gene expression of MMP-9, but not of MMP-2, was elevated by VEGF (Figs. 1C 1D) . The inhibitory effect of MMP-2 on the secretion of VEGF did not increase at higher concentrations, whereas the stimulatory effect of MMP-9 increased dose dependent (Fig. 4B) . Therefore, though the expression of MMPs is increased under hypoxic conditions, the stimulatory effect of MMP-9 on the secretion of VEGF may overcome the inhibitory effect of MMP-2 at higher concentrations (Fig. 4C) , resulting in a net stimulation of VEGF release. The positive feedback regulation of VEGF on the expression of MMP-9 may exacerbate this process. The different dose-dependent regulation of VEGF secretion by MMP-2 and MMP-9 represents a novel component of the angiogenic switch in the retinal tissue under hypoxic conditions. The importance of MMP-9 for pathologic angiogenesis has been shown in experimental CNV. Gene expression patterns of MMP-2 and MMP-9 differ in laser-induced CNV in mice; whereas MMP-2 is constantly expressed after laser-induced rupture of the Bruch membrane both in the choroidal neovascular membrane and in adjacent intact areas, the expression of MMP-9 coincides with the neovascular response. ²⁴ The authors suggest that inflammatory cells are a major provider of MMP-9, which contain preformed MMP-9 in intracellular granules and which are recruited into the neovascular area. ²⁴ It remains to be determined in future experiments whether MMP-9 derived from inflammatory cells may trigger the positive feedback regulation of VEGF and MMP-9 expression in RPE cells. In addition to the stimulation of VEGF release, MMP-9 secreted by RPE cells under hypoxic conditions may have further effects that contribute to pathologic processes. An increased secretion of MMP-9 may cause impairments of the interaction of RPE cells with their extracellular matrix and Bruch membrane, and may cause ruptures and enhanced hydraulic conductivity of Bruch membrane. ²⁶ Furthermore, RPE-derived MMP-9 may contribute to the leakage of retinal endothelium and pigment epithelium through the proteolytic degradation of the tight junction protein occludin ^{14 44} and to the stimulation of RPE cell migration as a component of the progression of fibrovascular membranes. 

Angiostatic steroids such as triamcinolone acetonide represent a potential treatment of ocular disorders associated with pathologic angiogenesis by inhibiting or stabilizing neovascularization. ^{45 46 47} The development of laser-induced CNV is reduced by angiostatic steroids, ^{48 49} and triamcinolone inhibits the migration and tube formation of choroidal endothelial cells. ⁵⁰ In addition to the anti-inflammatory action of triamcinolone and its reducing effect on vascular edema, the steroid may inhibit neovascularization by the downregulation of angiogenic factors. Triamcinolone reduces the vitreal level of VEGF in patients with diabetic retinopathy ⁵¹ and decreases the secretion of VEGF by RPE cells ³¹ and the transcription of the VEGF gene in vascular smooth muscle cells. ⁵² It has been shown that triamcinolone reduces the expression of MMP-2 and MMP-9 in choroidal endothelial cells and exerts its inhibitory effects on endothelial cell migration and tube formation, at least in part, by inhibiting MMP-2 activation. ⁵⁰ Here we show that triamcinolone decreases the gene expression of MMP-2 and MMP-9 and inhibits the secretion of MMP-9 by RPE cells evoked by chemical hypoxia or VEGF. In addition, triamcinolone reduces the gene expression and VEGF secretion after stimulation of the cells with MMP-9. The data suggest that this steroid inhibits the positive feedback regulation between MMP-9 and VEGF in RPE cells under hypoxic conditions through inhibition of the gene expression of both factors. On the other hand, triamcinolone does not inhibit the effect of MMP-9 on the chemotaxis of RPE cells. The observation that, under control conditions, triamcinolone decreased the secretion (Fig. 4A)but not the gene expression of VEGF (Fig. 3)is in agreement with a recent study that showed this steroid acts at the posttranscriptional level. ⁵³ In contrast, triamcinolone decreased both the gene expression (Fig. 3)and the secretion of VEGF (Fig. 4A)after stimulation with MMP-9, suggesting that transcriptional and posttranscriptional mechanisms are involved in the steroid-induced decrease in VEGF production by RPE cells. 

In summary, we show that RPE cells upregulate the expression of MMP-2 and MMP-9 under hypoxic conditions and that MMP-9 but not MMP-2 stimulates the production and secretion of VEGF. It is suggested that MMP-9, in addition to its proteolytic action on basement membranes and extracellular matrix components, facilitates neovascularization by direct stimulation of the expression of angiogenic factors by RPE cells. Inhibition of MMP-9, such as by intravitreal triamcinolone, may be a promising method to increase the efficiency of anti-VEGF therapies of ocular neovascularization. 

 

The authors thank Ute Weinbrecht for excellent technical assistance.