Expression of Sorsby's fundus dystrophy mutations in human retinal pigment epithelial cells reduces matrix metalloproteinase inhibition and may promote angiogenesis.

Sorsby's fundus dystrophy (SFD) is an autosomal dominant degenerative disease of the macula caused by mutations in the tissue inhibitor of metalloproteinase-3 (TIMP-3) gene. Choroidal neovascularization is a hallmark of this disease, which closely resembles the exudative form of age-related macular degeneration. However, the mechanism by which TIMP-3 mutations induce the disease phenotype in SFD remains unknown. To address this question we established human retinal pigment epithelial cell lines expressing wild type or S156C (Ser(156) changed to cysteine) mutant TIMP-3. S156C TIMP-3 had reduced matrix metalloproteinase (MMP) inhibitory activity in retinal pigment epithelial cells and resulted in increased secretion and activation of gelatinase A and B. The conditioned medium from these cells induced angiogenesis in "in vivo" chick chorioallantoic membrane assays that could be reversed with recombinant wild type TIMP-3. Our data indicate that the choroidal neovascularization in SFD may be a result of increased MMP activity, which could lead to the stimulation of angiogenesis. These results also suggest the potential therapeutic use of TIMP-3 or synthetic MMP inhibitors in this disease.

Sorsby's fundus dystrophy (SFD), 1 a fully penetrant, autosomal dominant, degenerative disease of the macula (1), is manifested by symptoms of night blindness or sudden loss of acuity usually in the third to fourth decades of life due to submacular neovascularization (2)(3)(4)(5). Clinically, early mid-peripheral drusen and color vision deficits are found (4,5). SFD is a relatively rare disease, but it has generated significant interest because it closely resembles the exudative or "wet" form of age-related macular degeneration, the most common cause of blindness in the elderly population of the western world. SFD is characterized by accumulation of extracellular deposits (drusen) in Bruch's membrane, the five-layered sheet of connective tissue that separates the retinal pigment epithelium (RPE) from the choriocapillaris (6). The predominant histopathological feature in the eyes of a 63-year-old patient with SFD was a confluent 30-m thick, lipid-containing amorphous deposit found between the basement membrane of the RPE and the inner collagenous layer of Bruch's membrane (6). This is distinct from the basal linear deposits seen in age-related macular degeneration that consist of filamentous fine granular material and may represent a thickened basement membrane of the RPE (6). The subretinal deposits in both SFD and age-related macular degeneration have been shown to be rich in tissue inhibitor of metalloproteinase-3 (TIMP-3) (7,8). A serious complication of SFD is the invasion of the thickened Bruch's membrane by newly formed, thin-walled vessels derived from the choriocapillaris. These vessels grow into the subretinal space causing exudative detachment of the RPE and loss of photoreceptors (5). SFD has been linked with mutations in the TIMP-3 gene (10 -12) with five different missense mutations and a splice site mutation having been identified to date.
TIMP-3 is a member of a family of endogenous matrix metalloproteinase (MMP) inhibitors of which there are currently four members (TIMPs 1-4). By virtue of their MMP inhibitory activity, TIMP family members play a potentially important role in regulating matrix composition and thereby affect a wide range of physiological processes that include cell growth, invasion, migration, angiogenesis, transformation, and apoptosis. Structurally TIMPs have a two-domain structure with each domain folded into three loops held together by three disulfide bonds (13). The N-terminal domain contains the highly conserved CXC motif that is responsible for MMP inhibition. The C-terminal domain confers specific functions such as binding to pro-MMPs and, in the case of TIMP-3, has been shown to be partly responsible for the ability to bind to the extracellular matrix (ECM) (14). However, more recently the ECM binding function has also been attributed to the N-terminal region (15). With one recent exception, all mutations associated with SFD change residues in exon 5 (which encodes the C-terminal domain of the molecule) into cysteines. The exception involves a truncation that also results in an unpaired cysteine residue in the C-terminal domain (10). The role of these mutations in generating the SFD phenotype is unclear.
To study the effects of these mutations in retinal pigment epithelium, we generated stable RPE cell lines expressing wild type or S156C (Ser 156 changed to cysteine) mutant TIMP-3 and investigated the effects on RPE functions.

EXPERIMENTAL PROCEDURES
Cell Culture-ARPE-19 cells (American Type Culture Collection) were maintained in T75 flasks in filter medium consisting of Dulbecco's modified Eagle's medium/Ham's F-12 plus 10% fetal bovine serum (Hyclone). All cultures of ARPE-19 grown on tissue culture plastic were fed weekly. Cultures were routinely passaged by dissociation with 0.05% (w/v) trypsin and 0.02% (w/v) EDTA, tetrasodium salt, followed by replating at a split ratio ranging from 1:3 to 1:6.
Generation and Analysis of Timp-3 Transfectant RPE Cell Lines-Wild type (WT)-Timp-3 and S156C-Timp-3 inserts from human cDNA clones were fused in-frame with FLAG epitope DYKDDDK at their C-terminal end and cloned into expression vector pcDNA3.1 (CLON-TECH, Palo Alto, CA). ARPE-19 cell lines were transfected by lipofection (LipofectAMINE reagent, Invitrogen) and selected in medium containing neomycin (0.5 mg/ml). As a control, cells were transfected with the vector alone. Conditioned medium, lysate, and ECM samples were prepared as described previously (16). Expression of transfected genes was confirmed by Western blot analysis with a polyclonal antibody to TIMP-3 (a kind gift of Suneel Apte, Cleveland, OH) or monoclonal anti-FLAG (M2) antibody (Sigma).
Preparation of ECM-Cells were plated onto six-well culture plates (Costar). When confluent, the cells were removed from the culture dishes following a 15-min incubation in Ca 2ϩ -, Mg 2ϩ -free phosphatebuffered saline (PBS) containing 5 mM EGTA and 1 mM phenylmethylsulfonyl fluoride. After several rinses in PBS and water the ECM was scraped in a small volume of electrophoresis sample buffer without reducing agent. For protein estimations, ECM was scraped in PBS containing 1 mM phenylmethylsulfonyl fluoride. Protein concentrations were estimated using the Bio-Rad protein assay reagent.
Zymography and Reverse Zymography-Equal amounts of protein were loaded on a 7.5% polyacrylamide gel with 1 mg/ml gelatin for zymography or a 12% gel with 1 mg/ml gelatin plus baby hamster kidney cell-conditioned medium as a source of MMPs for reverse zymography. Following electrophoresis, gels were processed as described previously (16). Briefly, gels were agitated in a solution of 25 mg/ml Triton X-100 to remove SDS and to promote renaturation of proteases and inhibitors. The Triton was washed off with water, and the gels were then incubated for 16 h in 50 mM Tris-HCl (pH 7.5) containing 5 mM CaCl 2 and 0.2 mg/ml sodium azide at 37°C. Gels were stained with 5 mg/ml Coomassie Blue R-250 in acetic acid/methanol/water (1:3:6) for 2 h and destained with acetic acid/methanol/water (1:3:6). Gelatinase activity was also determined using a gelatinase activity assay kit (Chemicon International, Temecula, CA) according to the manufacturer's instructions.
Cell Adhesion Assays-Human laminin (Collaborative Biomedical Inc., Bedford, MA) was immobilized on 96-well nontissue culturetreated plates. Wells were washed and incubated with 3% bovine serum albumin in PBS for 1 h at 37°C. Subconfluent RPE cells were harvested, washed, resuspended in culture medium (1 ϫ 10 5 cells/ml) and allowed to attach to the wells for 30 min at 37°C. Nonadherent cells were removed by washing, and the attached cells were stained for 10 min with crystal violet. The wells were washed three times with PBS, and cell associated crystal violet was eluted by addition of 100 l of 10% acetic acid. Cell adhesion was quantified by measuring the optical density of eluted crystal violet at a wavelength of 600 nm.
Cell Migration Assay-A modified Boyden chamber assay was carried out as described previously (17). 8.0-m pore polyvinylpyrrolidonefree polycarbonate membranes were precoated with collagen type I (100 g/ml). For migration of RPE cell transfectants, laminin (5 g/ml) or 10% fetal bovine serum was placed in the lower wells, and RPE cells (1.0 ϫ 10 5 ) were placed in the upper wells. For endothelial cell chemotaxis, human dermal microvascular endothelial cells (Clonetics, San Diego, CA) (3.0 ϫ 10 5 ) were placed in the upper wells and allowed to migrate toward conditioned medium placed in the lower wells. Chambers were incubated for 4 h at 37°C in a 5% CO 2 humidified incubator. Cells remaining on the top of the filter were removed. The bottom surface of the filter was fixed, stained, and mounted. The number of cells migrating per well was counted microscopically, and mean Ϯ S.E. was calculated from quadruplicate samples.
Matrigel and Type I Collagen Gel Invasion Assay-This assay was carried out as described previously (18) with some modifications. Boyden chambers were used, and 8.0-m polyvinylpyrrolidone-free filters were coated with matrigel or collagen type I gel (Collaborative Biomedical Inc.). The assay was carried out at 37°C for 18 -24 h, and filters were fixed, stained, and mounted. Invaded cells were counted under a microscope, and mean Ϯ S.E. was calculated from quadruplicate samples.
Endothelial Cell Proliferation Assay-This assay was carried out as described previously (17). Human dermal microvascular endothelial cells were incubated for 72 h in the presence of conditioned medium from RPE cell transfectants. Cells were trypsinized and counted using a Coulter particle counter.
Chick Chorioallantoic Membrane (CAM) Assays-The CAM assay was performed as described previously (19) with slight modifications. Fertilized 3-day-old White Leghorn eggs (Sunnyside Inc., Beaver Dam, WI) were cracked, and embryos with the yolk intact were placed in 100ϫ 20-mm glass bottom Petri dishes. After incubation for 3 days at 37°C in 3% CO 2 , CAMs were implanted with 5-mm diameter sterilized gelatin sponges (Gelfoam, Upjohn Co., Kalamazoo, MI) loaded with equal protein amounts of conditioned medium from RPE cell transfectants and photographed on day 3. CAMs implanted with sponges loaded with serum-free medium alone or with basic fibroblast growth factor or vascular endothelial growth factor were used as negative and positive controls, respectively. Recombinant TIMP-3 (a kind gift of Gillian Murphy, University of East Anglia, United Kingdom) was added to some pellets. Samples were always compared on the same CAM to avoid egg to egg variability. Responses were graded as Ͼ, ϭ, or Ͻ by independent readers.

SFD Mutant (S156C) TIMP-3 Protein Expressed in Human RPE Cells
Is an Inefficient TIMP-To mimic the RPE cells in SFD, we generated stable human RPE cell lines expressing WT or mutant (S156C) Timp-3. The stable expression of protein was confirmed by Western blot analysis using polyclonal antibodies to TIMP-3 ( Fig. 1a) and was quantified by scanning densitometry (Fig. 1b). Like the endogenous TIMP-3 in RPE cells, both wild type and mutant forms of TIMP-3 were expressed predominantly in the ECM (Fig. 1, a and b) with a small fraction being secreted into the conditioned medium (Fig.  1, c and d). Wild type and mutant TIMP-3 proteins were expressed at levels 3-5-fold over that of endogenous TIMP-3 present in ARPE-19 cells. Antibodies to the FLAG epitope also confirmed the expression of these proteins in the ECM (Fig. 1e). MMP inhibitory activity in the ECM of these cell lines was analyzed by reverse zymography (Fig. 1f) and quantified by scanning densitometry (Fig. 1, g and h). Mock-transfected RPE cells (pcDNA3.1 alone) expressed endogenous functional TIMP-3 inhibitor in the ECM with an approximate molecular mass of 24 kDa corresponding to the protein seen on Western blots (Fig. 1a) as well as the 27 kDa glycosylated form (Fig. 1f). This inhibitory activity was increased in cells that were transfected with the wild type Timp-3 construct. The glycosylated as well as the nonglycosylated forms of mutant (S156C) TIMP-3 in the ECM demonstrated reduced MMP inhibitory activity when compared with WT-TIMP-3 (Fig. 1, g and h).
We hypothesized that a reduction in TIMP activity in the ECM of RPE cells expressing mutant TIMP-3 may affect the protease activity secreted by these cells. Gelatin zymograms were used to test the gelatinase activity in the conditioned medium, lysates, and ECM of the transfected RPE cells. Endogenous 72-kDa gelatinase (pro-MMP-2) activity was observed in the conditioned medium of mock-transfected RPE cells (Fig. 2a). Densitometric quantitation (data not shown) determined a marked increase in the expression of this activity in the conditioned medium (60.4%) (Fig. 2a), cell lysates (57.2%) (Fig. 2c), and ECM (72%) (Fig. 2e) of cells expressing S156C mutant TIMP-3. Like other MMPs, gelatinase A is secreted as an inactive proenzyme, and its cellmediated activation requires binding of the C-terminal domain of the proenzyme to a membrane-associated complex of the membrane-type matrix metalloproteinase MT1-MMP and TIMP-2. Subsequent sequential proteolysis of the propeptide by MT1-MMP and gelatinase A is believed to generate the active form of gelatinase A, and this process can be regulated by TIMP-2 or TIMP-3 (20). Concanavalin A (ConA) can induce clustering of MT1-MMP, which plays a role in the acti-vation of MMP-2 in other systems (21). To study the potential role of mutant TIMP-3 in this process, RPE cells were treated with ConA, and the gelatinase activity in the conditioned medium (Fig. 2b) and cell lysate/ECM (Fig. 2d) fractions were examined by gelatin zymograms. RPE cells expressing mutant TIMP-3 showed an induction of the active forms in the conditioned medium (62 and 43 kDa) and lysate/ECM (62 kDa) (Fig. 2, b and d, respectively). MMP-9 activity was observed only in the conditioned medium of cells expressing mutant TIMP-3 (Fig. 2, a and b). Active MMP-2 in the con-ditioned medium of these cells was also quantitated with a gelatinase activity assay utilizing a biotinylated gelatinase substrate. This assay (Table I) confirmed an increase in the gelatinase activity in the conditioned medium of RPE cells expressing S156C TIMP-3 in the absence or presence of ConA.
Thus, in terms of MMP activity, expression of S156C TIMP-3 in RPE cells induced an increase in both pro-MMP-2 and active MMP-2 in the conditioned medium, lysate, and ECM of the cells possibly as a result of attenuated MMP inhibitory activity of the expressed mutant TIMP-3 protein. Since there is no decrease in the expression of MMP-2 in cells overexpressing WT-TIMP-3, the possibility that the increase in MMP-2 activity in the cells expressing mutant TIMP-3 may be a direct induction or activation of gelatinase A by the mutant TIMP-3 protein cannot be ruled out.

Expression of SFD Mutant (S156C) TIMP-3 Protein in RPE Cells Results in Altered Cell Adhesion, Migration, and Invasion-Quiescent RPE cells rest on the inner aspect of Bruch's membrane, which contains collagen type IV and laminin.
MMPs play a key role in the regulation of a variety of cellular functions. The adhesion of RPE cells to ECM and the migratory and invasive properties of the cell are critical for the maintenance of their physiological phenotype. RPE cell attachment was measured on laminin (Fig. 3a). Overexpression of WT-TIMP-3 had no effect on the adhesion of RPE cells to laminin relative to the mock-transfected cells. However, expression of S156C TIMP-3 resulted in a small but statistically significant decrease in adhesion to laminin (Fig. 3a). Migration of S156C TIMP-3-expressing RPE cells toward laminin (Fig. 3b) as well as 10% fetal bovine serum (Fig. 3c)   type I matrix revealed an increase in the invasive potential of RPE cells expressing S156C TIMP-3 (Fig. 3, d and e, respectively). Overexpression of WT-TIMP-3 resulted in a slight decrease in invasion through both matrigel and the collagen matrix. Thus, expression of S156C TIMP-3 in the RPE resulted in attenuated adhesion to ECM with a concomitant increase in the migratory and invasive potential. S156C TIMP-3 Mutation Induces Angiogenic Potential in RPE Cells-Choroidal neovascularization (CNV) is a major complication of Sorsby's fundus dystrophy. RPE cells are believed to play a key role in the pathogenesis of CNV (22)(23)(24)). Therefore, we tested the angiogenic potential of RPE cells expressing the S156C TIMP-3 mutation. We first examined the ability of the conditioned medium of RPE cells to induce migration and proliferation of endothelial cells. Conditioned medium from RPE cells expressing S156C TIMP-3 increased the migration of microvascular endothelial cells when compared with mock-transfected as well as WT-TIMP-3-expressing RPE cells (Fig. 4a) but had no effect on their proliferation (Fig. 4b).
We also evaluated the effects of the conditioned medium on angiogenesis using the CAM assay. Interestingly conditioned medium from cells expressing S156C TIMP-3 showed an increase in the neovascularization response compared with that induced by WT-TIMP-3 (in 89% of the CAMs tested, n ϭ 18) and control transfected cells (in 65% of the CAMs tested, n ϭ 17) (Fig. 4, c, d, and e). Addition of recombinant human TIMP-3 to the S156C TIMP-3-conditioned medium inhibited the angiogenic response (Fig. 4f). These results are consistent with the data obtained from the migration experiments and suggest that the SFD mutant (S156C TIMP-3) induces a proangiogenic phenotype in RPE cells. DISCUSSION The RPE is believed to play a modulating role in the maintenance of a vascularized and highly permeable fenestrated choriocapillaris on its outer basal side. In vivo studies as well as clinical observations have determined that the absence of RPE can lead to secondary atrophy of the choriocapillaris (25)(26)(27). Polarized secretion of vascular endothelial growth factor by the RPE toward the basal surface where its receptor KDR (vascular endothelial growth factor receptor-2) is localized on the inner choriocapillaris suggests a possible mechanism for trophic paracrine signaling between RPE and choriocapillaris (28). CNV is an important pathological feature of eye diseases such as Sorsby's fundus dystrophy and the more common agerelated macular degeneration. During the course of CNV, encapsulation of growing vessels by RPE coincides with the stasis of further vascular growth (29). This suggests that the RPE might play a critical role in controlling the initiation and progression of CNV.
A precise spatial and temporal regulation of extracellular proteolysis is required for neovascularization. Recent studies have indicated that members of the MMP family play an important role in angiogenesis (30 -33). Mice deficient in MMP-2 and MMP-9 exhibit reduced angiogenesis in vivo (34,35). MMP-2 is the most widely distributed MMP and is localized on the surface of angiogenic blood vessels (36). TIMP-3, a regulator of MMPs, is deposited by RPE cells into Bruch's membrane where it is a component of ECM (8,37). We have previously demonstrated that TIMP-3 is a potent inhibitor of angiogenesis (17). Since CNV is a prominent feature of SFD, with accumulation of TIMP-3 deposits in Bruch's membrane (7), we proposed to examine the mechanisms by which mutations in the TIMP-3 gene lead to the characteristic ocular phenotype (38,39).
Expression of S156C TIMP-3 in human RPE cells results in a reduction of MMP inhibitory (TIMP) activity in the ECM. Recombinant S156C TIMP-3 generated in NSO myeloma cells has association rate constants for a range of MMPs that are 2-3-fold reduced compared with wild type TIMP-3 (40). However, the decrease in TIMP activity in RPE cells is in contrast to that reported previously with expression in COS-7 and baby hamster kidney cells (10,40). In both of these cell types, high levels of expression of mutant TIMP-3 protein gives rise to dimers, which are not present in our study with RPE cells. We expressed protein at relatively low levels in cells that expressed endogenous wild type protein to mimic the heterozygous autosomal dominant phenotype. That TIMP-3 functions may be cell type-specific has been recently suggested with the observation that TIMP-3 may have antiapoptotic activity and is a critical survival factor during mammary gland involution (38). This is in contrast to several reports that demonstrate that high levels of TIMP-3 are proapoptotic in both normal and cancer cell lines (41)(42)(43)(44)(45)(46). In our studies with RPE cells, overexpression of TIMP-3 at modest levels does not affect cell survival (data not shown). Thus the divergent effects of TIMP-3 on cell survival as well as its other potential functions may depend on the cell type, microenvironment, and amount of TIMP-3 in the microenvironment (38). This might also explain the exclusive ocular phenotype seen in patients with the SFD mutation.
Constitutive activation of gelatinase A was observed in the conditioned medium and ECM of RPE cells expressing S156C TIMP-3. This was not surprising given the finding that SFD mutant TIMP-3 was an inefficient MMP inhibitor. Unscheduled MMP-2 activation was also observed in the lungs and mammary glands of homozygous Timp-3-null mice (38,39). It has been hypothesized that, under normal physiological conditions, choroidal endothelial cells are maintained in a quiescent state because of an orchestrated balance between angiogenic inducers and inhibitors. WT-TIMP-3 deposited by RPE cells into Bruch's membrane efficiently inhibits neovascular invasion from the choriocapillaris by inhibiting MMP activity, resulting in an intact matrix, and/or binding directly to the surface of endothelial cells to inhibit endothelial cell responses. Our results suggest that, during pathological CNV as seen in SFD, expression of mutant TIMP-3 in RPE cells may result in an accentuation of MMP activity, which can contribute to an angiostimulatory phenotype. While it is clear that endothelial cell basement membrane degradation is critical for angiogenesis, very little is known about the basic mechanism responsible for this process. There are several possibilities. Intact matrix might provide a physical barrier that limits endothelial cell migration and invasion into the surrounding ECM. Breakdown of the basement membrane allows the endothelial cells to initiate the process of sprouting. Alternatively proteolysis of ECM may release sequestered angiogenic factors such as vascular endothelial growth factor and/or basic fibroblast growth factor, which promote endothelial cell proliferation and migration. Finally a third possibility that has been recently proposed is the exposure of an angiogenic cryptic site upon proteolytic cleavage of ECM molecules such as collagen type IV (9).
In this study we show that the CNV phenotype seen in SFD may be a result of reduced MMP inhibitory activity with a concomitant increase in MMP-2 activity. This appears to provide the cells with the ability to induce angiogenesis as shown by both in vitro and in vivo assays. These results might explain the pathophysiological mechanism of CNV in SFD and suggest the potential use of MMP inhibitors locally in the subretinal space as a means of therapy.